阅读说明：本技术 一种基于模型预测控制的混合交通流协同优化控制方法 (Mixed traffic flow cooperative optimization control method based on model predictive control ) 是由 孙湛博 高子延 李哲宜 于 2020-12-24 设计创作，主要内容包括：本发明公开了一种基于模型预测控制的混合交通流协同优化控制方法,该方法采用基于模型预测控制的双层优化模型进行混合交通流协同决策控制,适用于两条有交叉的单向单行道路且交叉路口前无车辆行驶信号指示下混合交通流中的不同交通场景；模型预测控制是指通过建立系统架构,实现实时闭环控制,用以解决实际情况中出现的随机性问题；双层优化模型包括上层模型和下层模型,上层模型是一个运用动态规划递归求解的车辆排序问题,下层模型是一个运用动态矩阵预测算法求解每一单次车辆的轨迹优化问题,下层模型中每一单次车辆的轨迹优化结果为上层模型动态规划递归求解过程中的一个输入；通过模型预测控制和双层优化模型共同保证系统车辆运行最优。(The invention discloses a mixed traffic flow collaborative optimization control method based on model predictive control, which adopts a double-layer optimization model based on model predictive control to carry out mixed traffic flow collaborative decision control and is suitable for different traffic scenes in mixed traffic flows under the condition that two crossed one-way roads are arranged and no vehicle driving signal is indicated in front of the intersection; the model predictive control means that real-time closed-loop control is realized by establishing a system architecture so as to solve the problem of randomness in actual conditions; the double-layer optimization model comprises an upper layer model and a lower layer model, wherein the upper layer model is a vehicle sequencing problem solved by dynamic programming recursion, the lower layer model is a track optimization problem solved by a dynamic matrix prediction algorithm for each single vehicle, and a track optimization result of each single vehicle in the lower layer model is an input in the dynamic programming recursion solving process of the upper layer model; and the optimal operation of the system vehicle is ensured through the model predictive control and the double-layer optimization model.)

1. A mixed traffic flow cooperative optimization control method based on model predictive control is characterized in that: the method adopts a double-layer optimization model based on model prediction control to carry out mixed traffic flow cooperative decision control, and is suitable for different traffic scenes in mixed traffic flows under the indication of two crossed one-way roads and no vehicle running signal in front of the intersection; the model predictive control means that real-time closed-loop control is realized by establishing a system architecture so as to solve the problem of randomness of driving behaviors in actual situations; the double-layer optimization model comprises an upper layer model and a lower layer model, the upper layer model is a vehicle sequencing problem solved by dynamic programming recursion, the lower layer model is a track optimization problem solved by a dynamic matrix prediction algorithm, and a track optimization result of each single vehicle in the lower layer model is used as one input in the dynamic programming recursion solving process of the upper layer model; ensuring that the operation of the system vehicle is optimal through the model predictive control and the double-layer optimization model;

the method comprises the following steps:

s1, determining road grouping optimization sections and performing time domain division;

the method comprises the following steps that an X road and a Y road are assumed to be two one-way roads with intersections, no vehicle driving signal knowledge exists before the intersections, and the road grouping optimization sections comprise an X road grouping optimization section and a Y road grouping optimization section; setting a section of road before vehicles on the X road reach the traffic flow intersection as an X road grouping optimization section, and setting a section of road before vehicles on the Y road reach the traffic flow intersection as a Y road grouping optimization section;

s2, carrying out traffic flow grouping according to the road grouping optimization section, determining vehicles running on the road grouping optimization section at the same time as vehicles in the group, and acquiring real-time state information of the vehicles in the group;

s3, predicting the running tracks of the vehicles in the group by adopting a double-layer optimization model, making an optimization control decision according to the running tracks, and performing optimization control on the vehicles according to the optimization control decision, so that the vehicle closest to the traffic flow junction in the vehicles in the group can smoothly pass through the traffic flow junction, and the running track of the vehicle in the family is ensured to be optimal;

and S4, performing optimization control on the allowable tracks of the vehicles in the group driving on the road grouping optimization section at the next moment by repeatedly executing the steps S2 and S3, and executing in a circulating mode until all the vehicles drive out of the road grouping optimization section.

2. The model predictive control-based mixed traffic flow collaborative optimization control method according to claim 1, characterized in that: step S3, predicting the movement tracks of the vehicles in the group by using a double-layer optimization model, making an optimization control decision based on the prediction, and performing optimization control on the vehicles according to the optimization control decision, so that the vehicle closest to the traffic flow intersection point in the vehicles in the group can smoothly pass through the traffic flow intersection point, and the movement track of the vehicle in the group is ensured to be optimal, which specifically includes:

s3-1, establishing the upper layer model, including:

s3-1-1, determining the division stage, the state variable and the decision variable of the upper layer model, which are as follows:

the upper layer model is divided: at the same time, n vehicles on the X road grouping optimization section need to sequentially merge into or pass through m +1 intervals among the vehicles on the Y road grouping optimization section, the behavior that each single vehicle on the X road grouping optimization section merges into or passes through the Y road is represented as a stage, the behavior that the k-th vehicle on the X road grouping optimization section merges into or passes through the Y road is recorded as the k-th stage, wherein k is 1,2,3,..., n;

state variables of the upper model: the k stage can be used for leading k to n vehicles on the X road grouping optimization section to merge into or pass through the vehicle interval number on the Y road grouping optimization section by s_kRepresents;

decision variables of the upper model: the decision made at each stage represents s of the k vehicle on the X road grouping optimization section at the k stage on the section capable of merging into or passing through the Y road grouping optimization section_kSelecting a particular xth of individual vehicle intervals_kThe vehicles are alternately converged or passed;

s3-1-2, determining a state transition equation, a cost function and an objective function of the upper layer model, wherein the details are as follows:

the state transition equation of the upper model is as follows:

setting the initial condition as s₀＝m+1；

The state transition equation of the upper model indicates that s is equal to 1 when k_k＝s₁The number of vehicle intervals for the 1 st vehicle on the X road grouping optimization section to enter or pass through the Y road grouping optimization section in the 1 st stage is m + 1; when k is 2, 3.., n, the (k-1) th vehicle on the X-road grouping optimization section selects the (X) th vehicle at the (k-1) th stage_k-1The interval of each vehicle is used as a state variable s after the vehicle is converged or passes through a Y road_kA change in (c); s₀M +1 represents that the number of vehicle intervals available for vehicles on the X road grouping optimization section to merge into or pass through the Y road grouping optimization section in the initial state is m + 1;

cost function of the upper model:

cost function D of the upper model_k(s_k，x_k) Indicating a phase index function required for making a decision at the kth phase, whereinShowing that the k-th vehicle on the X road grouping optimization section is merged into or passes through the k-th vehicle or passes through the S on the Y road grouping optimization section under the action of the collaborative optimization control strategy in the k stage_kAll possible cost costs arising from individual vehicle intervals;refers to the Y roadThe vehicles which do not directly participate in the process of vehicle convergence or vehicle confluence or vehicle crossing of the intersection on the grouping optimization section are influenced by vehicle convergence or passing of the front vehicles, so that the cost is consumed due to vehicle speed adjustment of the self vehicle due to the following safety requirement;

an objective function of the upper model:

setting the initial condition as f₀(s₀)＝0；

Objective function f of the upper model_k(s_k) Representing the cumulative cost consumption of the system vehicles from stage 1 to stage k, f₀(s₀) 0 denotes that the system cost is 0 in the initial state;

s3-2, establishing the lower layer model, including:

s3-2-1, determining a micro following model, describing a following state of the vehicle by using the micro following model, and predicting a vehicle track; the following state of the vehicle comprises the speed, acceleration and position of the vehicle;

s3-2-2, establishing a condition constraint model for judging whether the kth vehicle in the path grouping optimization section at the kth stage X can smoothly merge into or pass through the Y path;

s3-2-3, based on the predicted vehicle track of the step S3-2-1, sequentially judging whether the k vehicle in the X road grouping optimization section of the k stage can merge into or pass through the S vehicle in the Y road grouping optimization section by the condition constraint model established in the step S3-2-2_kEach vehicle interval of the vehicle intervals;

s3-2-4, determining an optimally controlled vehicle in the vehicles participating in the k stage as a target vehicle, optimizing the running track of the target vehicle, solving the optimization problem into an optimal control problem of continuous track state constraint under discrete time, and solving by using a dynamic matrix predictive control algorithm to obtain a cooperative optimization control strategy about the target vehicle;

s3-2-5, calculating the k-th vehicle merging on the X road grouping optimization section or passing through the k-th vehicle merging or passing through the S on the Y road grouping optimization section in the k stage under the action of the cooperative optimization control strategy_kAll possible cost consumptions of one vehicle interval

S3-3, solving the double-layer optimization model:

s3-3-1, solving the upper layer model, and determining the decision made by each stage of the upper layer model when the accumulated cost consumption of the system vehicle is the lowest;

s3-3-2, reversely deducing the decision made at each stage of the upper layer model determined by the step S3-3-1 to obtain the vehicle optimization track of each single vehicle merging into or passing through the Y road on the X road grouping optimization section of the lower layer model;

and S3-4, obtaining the mixed traffic flow cooperative decision of the double-layer optimization model aiming at the system optimization by solving in the step S3-3, and acting the decision on the system vehicle to control the operation of the system vehicle.

3. The model predictive control-based mixed traffic flow collaborative optimization control method according to claim 2, characterized in that: the micro following model in the step S3-2-1 comprises a deterministic micro following model and a stochastic micro following model, the vehicle can be optimally controlled to adopt the deterministic micro following model to perform action prediction, and the deterministic micro following model can be a Gipps following model or an IDM/EIDM following model; the non-optimizable control vehicle adopts the randomness micro-following model to simulate the randomness of the driving behavior, and the randomness micro-following model can carry out random following modification on the basis of the certainty micro-following model.

4. The model predictive control-based mixed traffic flow collaborative optimization control method according to claim 3, characterized in that: the stochastic micro-follow model is as follows:

v_n(t+δ)＝min(v^e+θ₁，v_safe+θ₂)，

θ₁～N(0，σ₁^2)，

θ₂～N(0，σ₂^2)，

wherein v is_safeRepresenting a safe speed in the stochastic micro-follow model; b represents the deceleration; δ represents the interval between two adjacent time slices; n denotes the number of the non-optimally controlled vehicle, n-1 denotes the number of the preceding vehicle followed by the non-optimally controlled vehicle n, v_n-1(t) represents the speed of the preceding vehicle n-1 at time t during the vehicle-following; l_n(t) represents the position of the non-optimistically controllable vehicle n during the vehicle-following at time t; l_n-1(t) represents the position of the preceding vehicle n-1 during the vehicle-following at time t; l_aRepresenting a vehicle length; l₀Represents a vehicle following minimum distance; v. of_n(t + delta) represents the speed of the non-optimistically controllable vehicle n at the moment t + delta during the following of the vehicle, taking v^e+θ₁And v_safe+θ₂The smaller of the two; v. of^eRepresenting a desired speed; theta₁、θ₂Are all constant and all follow normal distribution; σ is set in consideration of the fact that the fluctuation of the speed of the vehicle will be small when the vehicle is at a desired speed and the fluctuation of the speed will be large when the vehicle is forced to decelerate to a safe speed₁＜σ₂。

5. The model predictive control-based mixed traffic flow collaborative optimization control method according to claim 2, characterized in that: in step S3-2-2, a condition constraint model for determining whether the kth vehicle in the kth stage X road grouping optimization section can smoothly merge into or pass through the Y road is established, which specifically includes:

assuming a kth vehicle on the kth phase X road-grouping optimization segment,recording as vehicle k, merging or passing through the Y road from the interval between two vehicles of the continuous traffic flow on the Y road by k'₁And k'₂Is represented by, wherein vehicle k'₁Denotes a front vehicle, vehicle k'₂The rear vehicle is represented by:

wherein E is_k(t) a conflict cooperative utility function for reflecting whether a vehicle k on the X road can smoothly merge into or pass through the Y road; | u_k(t) | denotes an absolute value of acceleration or deceleration of the vehicle k at time t on the X-road,denotes vehicle k 'on the Y road'₂The absolute value of acceleration or deceleration at time t;denotes at time t ' vehicle k on the X road and vehicle k ' on the Y road '₁The relative distance between the two or more of them,denotes at time t ' vehicle k on the X road and vehicle k ' on the Y road '₂The relative distance therebetween; l_k(t) represents the position of the vehicle k on the X road at time t,denotes vehicle k 'on the Y road'₁At the position of the instant t of time,denotes vehicle k 'on the Y road'₂Position at time t,/_aIs the vehicle length;representing a minimum inter-vehicle distance between the optimally controllable vehicle and a vehicle ahead of the optimally controllable vehicle;representing a minimum inter-vehicle distance between the non-optimally controllable vehicle and a preceding vehicle following the non-optimally controllable vehicle; when vehicle k is ready to merge or pass through the Y road, vehicle k on the X road actually follows vehicle k 'on the Y road'₁Run, and vehicle k 'on the Y road'₂The vehicles k on the actual following X road run, and the acceleration or deceleration of the vehicles k can be calculated according to the microcosmic following model; b_safeRepresents a maximum allowable deceleration; phi_AA set of optimally controllable vehicles; phi_HA set of non-optimally controlled vehicles; eta₁Representing a safety factor; eta₂Representing a polite coefficient;

polite coefficient eta₂The expression of (a) is as follows:

wherein v is_kIs the speed of the vehicle k, v_thIs a given threshold speed, β₁And beta₂Is a constant;

m for cooperative decision_k(t+T_c) Is represented as follows:

wherein m is_k(t+T_c) A value of 1 indicates that the vehicle k on the X road is at T + T_cCan smoothly converge into or pass through a Y road at any moment; m is_k(t+T_c) A value of 0 indicates that the vehicle k on the X road is at T + T_cThe user can not smoothly converge into or pass through the Y road at any time.

6. Model-based predictive control according to claim 2The manufactured mixed traffic flow collaborative optimization control method is characterized by comprising the following steps: and in the k stage, under the action of a collaborative optimization control strategy, the k vehicle on the X road grouping optimization section is merged into or passes through the k vehicle or passes through the s on the Y road grouping optimization section_kAll possible cost costs resulting from the separation of individual vehiclesThe method specifically comprises the following steps:

defining a vehicle combination K: predicting a vehicle k on an X road, merging into or passing through a Y road an interval between two vehicles passing a continuous traffic flow on the Y road, the two vehicles on the continuous traffic flow on the Y road being respectively usedAndshow, wherein the vehicleIndicating predicted vehicle ahead and vehicleRepresenting a predicted vehicle; further defining a vehicle combination K, representing the vehicle combination directly participating in the K stage on the X road and the Y road, and K belongs to { K ∈ [ K ]₁，K₂，K₃Vehicle combinationIndicating that the vehicle participating in the k-th stage has a vehicleVehicle k and vehicleVehicle combinationThe vehicle participating in the kth stage comprises a vehicle k and a vehicleAt this time, the x-th road is positioned on the Y road_kThe interval of each vehicle has no front vehicle; vehicle combinationIndicating that the vehicle participating in the k-th stage has a vehicleVehicle k, now on the Xth road of Y_kNo rear vehicle participates at the interval of each vehicle;

thenThe expression of (a) is as follows:

wherein i represents the ith vehicle in the current vehicle combination and is recorded as a vehicle i;representing the speed of the vehicle i at the moment t in the k phase; v. of^eRepresenting a desired speed; t is t₀Represents the current time; t is_PRepresents a predicted time domain length;representing the speed of a rear vehicle on a Y road in the k-1 stage at the time t; x is the number of_k＞x_k-1+1 indicates that at least one vehicle clearance still exists between the vehicle k on the k-th stage X road and the vehicle clearance on the Y road occupied by the vehicle k-1 on the k-1 stage X road; x is the number of_k＝x_k-1+1 denotes the kth orderThe vehicle clearance on the Y road occupied by the vehicle k on the road of the segment X and the vehicle k-1 on the road of the k-1 stage X is two continuous vehicle clearances; x is the number of_k＝x_k-1It means that the vehicle clearance on the Y road selected by the vehicle k on the k-th stage X road and the vehicle k-1 on the k-1 stage X road is the same vehicle clearance.

7. The model predictive control-based mixed traffic flow collaborative optimization control method according to claim 2, characterized in that: the vehicles which do not directly participate in the process of vehicle confluence or vehicle crossing intersection on the Y road grouping optimization section are influenced by vehicle confluence or passing of the front vehicles, so that cost consumption is caused by vehicle speed adjustment of the self vehicle due to the requirement of following safetyThe expression is as follows:

wherein i represents the ith vehicle in the current vehicle combination and is recorded as a vehicle i;representing the speed of the vehicle i at the moment t in the k phase; v. of^eRepresenting a desired speed; t is t₀Represents the current time; t is_PRepresents a predicted time domain length; the vehicle combination K' represents vehicles between the rear vehicle at the K-1 stage and the front vehicle at the K stage on the Y road grouping optimization section, namely, the vehicles do not directly participate in the process of vehicle confluence or vehicle crossing of the intersection; x is the number of_k＞x_k-1+1 indicates that there is at least one more vehicle clearance between vehicle k on the k-th stage X road and the vehicle clearance on the Y road occupied by vehicle k-1 on the k-1 stage X road.

8. The model predictive control-based mixed traffic flow collaborative optimization control method according to claim 2, characterized in that: step S3-2-4, determining vehicles capable of being controlled in an optimized manner in the vehicles participating in the k stage as target vehicles, optimizing the running tracks of the target vehicles, solving the optimization problem into an optimal control problem constrained by a continuous track state in discrete time, and solving by using a dynamic matrix predictive control algorithm to obtain a cooperative optimization control strategy related to the target vehicles, wherein the cooperative optimization control strategy specifically comprises the following steps:

defining a vehicle combination K: predicting a vehicle k on an X road, merging into or passing through a Y road an interval between two vehicles passing a continuous traffic flow on the Y road, the two vehicles on the continuous traffic flow on the Y road being respectively usedAndshow, wherein the vehicleIndicating predicted vehicle ahead and vehicleRepresenting a predicted vehicle; further defining a vehicle combination K, representing the vehicle combination directly participating in the K stage on the X road and the Y road, and K belongs to { K ∈ [ K ]₁，K₂，K₃Vehicle combinationIndicating that the vehicle participating in the k-th stage has a vehicleVehicle k and vehicleVehicle combinationThe vehicle participating in the kth stage comprises a vehicle k and a vehicleAt this time, the x-th road is positioned on the Y road_kThe interval of each vehicle has no front vehicle; vehicle combinationIndicating that the vehicle participating in the k-th stage has a vehicleVehicle k, now on the Xth road of Y_kNo rear vehicle participates at the interval of each vehicle;

defining a dynamic matrix predictive control model as follows:

the above equations (1) to (4) are described:

equation (1) is an objective function of a dynamic matrix predictive control model, whereIndicating actual vehicle speed and desired speedThe difference in degrees; introduction ofThe aim is to make the vehicle acceleration as soft as possible, taking into account the comfort of the driver; gamma ray₁、γ₂Is a weight coefficient; weight coefficient gamma₂The larger the dynamic matrix predictive control model is, the more the objective function of the dynamic matrix predictive control model focuses on optimizing the comfort of the controlled vehicle; on the contrary, the objective function of the dynamic matrix predictive control model focuses on optimizing the comfort level of the whole traffic flow;

the formula (2), the formula (3) and the formula (4) are constraint conditions of the dynamic matrix prediction control model;

equation (2) describes the controlled variable constraint, where a_minAnd a_maxRepresenting an allowable maximum deceleration and a maximum acceleration, respectively, whose values are determined based on the comfort of the controlled vehicle;

equation (3) is a vehicle speed state constraint, whereRepresenting a vehicle following speed matrix function, wherein the constraint can ensure the safety of the following behavior in the control process, namely collision avoidance constraint; wherein the content of the first and second substances,indicating that vehicle i in vehicle combination K is at t₀Position matrix of time of day comprising t₀Is at the momentPredicting the position of a vehicle i in the vehicle combination K at any moment; v_i ^k(t₀) Indicating that vehicle i in vehicle combination K is at t₀Velocity matrix of time of day comprising t₀Is at the momentPredicting the speed of the vehicle i in the vehicle combination K at any moment;indicates that the preceding vehicle in the vehicle combination K is following the vehicle i at t₀Velocity matrix of time of day comprising t₀Is at the momentPredicting the speed of a vehicle before a vehicle i in the vehicle combination K at any moment;

equation (4) defines the final state constraint that vehicle k on the kth stage X road is inThe conflicting collaborative utility function value at a time may only be at or above 0Smoothly converge into or pass through a Y road at any moment;

to complete the iterative process of dynamic matrix predictive control, the state transition process is defined as follows:

in the above-mentioned formula (5),indicating that vehicle i in vehicle combination K is at t₀Velocity v of time_i(t₀) And position l_i(t₀) A set of (a);

the above equations (6) to (8) are the state transition equations of the non-optimally controlled vehicle, where K_H＝{i|α_i0 n ∈ K, and α_iIs the vehicle type, alpha if the vehicle is an optimally controllable vehicle_iIs 1, alpha if the vehicle is a non-optimally controllable vehicle_iIs 0;

equation (6) is a function of the position of the non-optimally controlled vehicle;

equation (7) is a speed function for a non-optimally controlled vehicle;

equation (8) is a general formula for a microscopic follow-up model, i.e., the acceleration of a non-optimally controlled vehicle is a function of the clearance between the vehicle and its following predecessor;

C_v＝[1 0]； (11)

C_s＝[0 1]； (12)

the above equations (9) to (12) are the state transition equations for optimally controlling the vehicle, where K_A＝{i|α_i＝1∩i∈K}，α_iIs the vehicle type, alpha if the vehicle is an optimally controllable vehicle_iIs 1, alpha if the vehicle is a non-optimally controllable vehicle_iIs 0;

the formula (9) includes a formula (9a) and a formula (9b), in whichIt is described that the controlled speed and position of the vehicle at time t can be optimally controlled, sinceThere is a recurrence relation shown by the formula (9a) and the formula (9b), and therefore at an arbitrary predicted timingCan be prepared fromAndcarrying out representation; matrix arrayMatrix arrayIn a corresponding shapeA state transition matrix;is to optimally control the acceleration decision of the vehicle at time t, i.e.Is the optimal control variable;

in the formula (10),is a set of column vectors that can optimally control the acceleration decision of the vehicle at each predicted time point;

the compound of the formula (11),is a set of column vectors that can optimally control the controlled vehicle speed of the vehicle at each predicted time point, whereinIs the corresponding state transition matrix and is,is an auxiliary matrix in the derivation process;

in the formula (12),is a set of column vectors that optimally controls the controlled vehicle position of the vehicle at each predicted time point, whereinIs the corresponding state transition matrix and is,is the auxiliary matrix in the derivation process.

Technical Field

The invention relates to a model prediction control-based hybrid traffic flow collaborative optimization control method, and belongs to the field of traffic engineering.

Background

The concept of autonomous driving since the 40's of the 20 th century was proposed, intelligent networked vehicles have experienced a development history of nearly 80 years. In recent years, with the great application of technologies such as navigation, electronic maps, sensing detection, wireless communication, automatic control, mobile interconnection and the like in the transportation and automobile industries, the development of intelligent networked vehicles enters a new stage. An autonomous vehicle (smart networked vehicle) is defined as a vehicle that is capable of sensing and communicating with the driving environment, and the operation of the vehicle (in part or in whole) may be performed without driver action. Autonomous vehicles have better controllability and synergy than the difficult synergy of conventional human-driven vehicles (conventionally driven vehicles). This may therefore provide benefits such as improved fuel/energy efficiency, traffic safety and traffic stability. Before the hundred percent popularization of the automatic driving vehicles, mixed traffic (namely the state that the traditional driving vehicles and the intelligent networked vehicles are mixed) becomes a main mode of road traffic. Under the mixed traffic flow environment, under various traffic scenes such as ramp junction convergence at a highway intersection, junction or junction confluence at a T-junction and the like, the automatic driving vehicle may collide with the traditional driving vehicle to cause certain harm. The scientific theoretical framework and the modeling method are used for carrying out mixed traffic flow cooperative decision control, so that traffic conflicts are reduced or eliminated, and the vehicle track is optimized to a certain degree to become a practical problem which needs to face in future traffic for a long time.

Existing research related to autonomous vehicles is mostly based on the assumption that the permeability of the autonomous vehicle is 100%, and is mostly studied from a macroscopic perspective. The model of the microscopic level is also mainly focused on the single-vehicle track optimization of the research automatic driving vehicle, and the optimization of the system level cannot be guaranteed. From a microscopic view point, a mixed traffic flow cooperative decision control optimization model aiming at system optimization is researched. With the development of real-time communication capability of V2V and V2I, more advanced and efficient management control systems can be developed to reduce congestion and air pollutant emission and enhance driving safety. The cooperative decision control under the mixed traffic flow environment is carried out aiming at various microscopic traffic scenes, such as crossroads, T-shaped intersections, expressway ramp confluence and the like, so that traffic conflicts are eliminated to the maximum extent, and the research on traffic operation efficiency and traffic capacity is basically not available. Uncertainty in actual following behavior, especially in human driving behavior of a conventional driving vehicle, is inevitable in practical use. Therefore, the cooperative control decision model should solve the randomness of following in the mixed traffic flow by using some more robust models.

Disclosure of Invention

The invention aims to provide a model prediction control-based hybrid traffic flow cooperative optimization control method.

The technical scheme adopted by the invention for realizing the aim is as follows:

a mixed traffic flow cooperative optimization control method based on model predictive control is characterized in that a double-layer optimization model based on model predictive control is adopted to carry out mixed traffic flow cooperative decision control, and the method is suitable for different traffic scenes (different traffic scenes comprise highway ramp vehicle convergence, intersection vehicle confluence and vehicle passing intersection) in mixed traffic flow under the condition that two crossed one-way roads are provided and no vehicle driving signal is used for indicating before the intersection; the model predictive control means that real-time closed-loop control is realized by establishing a system framework (namely, slicing is carried out according to time, the track of a vehicle before reaching a traffic flow junction is predicted at a certain moment, an optimization control decision is made according to the track, but only an adjacent control sequence is executed at the current moment; the double-layer optimization model comprises an upper layer model and a lower layer model, the upper layer model is a vehicle sequencing problem solved by dynamic programming recursion, the lower layer model is a track optimization problem solved by a dynamic matrix prediction algorithm, and a track optimization result of each single vehicle in the lower layer model is used as one input in the dynamic programming recursion solving process of the upper layer model; ensuring that the operation of the system vehicle is optimal through the model predictive control and the double-layer optimization model;

(for the whole road grouping optimization section, including the vehicles of the X road grouping optimization section and the Y road grouping optimization section, the adjacent control sequence can be executed to make all the vehicles which can be optimally controlled, namely the intelligent internet vehicles, advance to the next moment, and make the vehicle which is closest to the traffic flow junction smoothly pass through the traffic flow junction.)

The method comprises the following steps:

s1, determining road grouping optimization sections and performing time domain division;

the method comprises the following steps that an X road and a Y road are assumed to be two one-way roads with intersections, no vehicle driving signal indication exists before the intersections, and the road grouping optimization sections comprise an X road grouping optimization section and a Y road grouping optimization section; setting a section of road before vehicles on the X road reach the traffic flow intersection as an X road grouping optimization section, and setting a section of road before vehicles on the Y road reach the traffic flow intersection as a Y road grouping optimization section;

s2, carrying out traffic flow grouping according to the road grouping optimization section, determining vehicles running on the road grouping optimization section at the same time as vehicles in the group, and acquiring real-time state information of the vehicles in the group;

s3, predicting the running tracks of the vehicles in the group by adopting a double-layer optimization model, making an optimization control decision according to the running tracks, and performing optimization control on the vehicles according to the optimization control decision, so that the vehicle closest to the traffic flow junction in the vehicles in the group can smoothly pass through the traffic flow junction, and the running track of the vehicle in the group is ensured to be optimal;

and S4, performing optimization control on the running tracks of the vehicles in the group driving on the road grouping optimization section at the next moment by repeatedly executing the steps S2 and S3, and executing in a circulating mode until all the vehicles drive out of the road grouping optimization section.

(namely, slicing according to time, obtaining vehicle information in the road grouping optimization section at a certain moment, predicting a vehicle track, performing optimization control on the vehicle track by executing a double-layer optimization model, obtaining the vehicle information in the road grouping optimization section again at the next moment, predicting the vehicle track again, performing optimization control on the vehicle track by executing the double-layer optimization model, and repeating the steps till the end.)

Compared with the prior art, the invention has the following beneficial effects:

the method adopts a double-layer optimization model based on model prediction control to carry out mixed traffic flow cooperative decision control, and is suitable for different traffic scenes (including expressway ramp vehicle convergence, intersection vehicle confluence and vehicle passing intersection) in mixed traffic flow under the condition that two crossed one-way roads are provided and no vehicle driving signal is indicated in front of the intersection. Therefore, the method of the invention has certain universal applicability.

In the method, an upper layer model of a double-layer optimization model is a vehicle sequencing problem solved by dynamic programming recursion, and the system optimization of the mixed traffic flow cooperative decision control is ensured by searching the system optimal sequence; the lower layer model of the double-layer optimization model is a dynamic matrix prediction algorithm which is used for solving the track optimization problem of each single vehicle, and the track of each single vehicle is optimized and controlled, so that the vehicles can smoothly pass through a traffic flow intersection point and the track of the vehicles is optimal; and a vehicle track optimization result obtained by an objective function in a lower layer model of the double-layer optimization model is used as an input in the dynamic programming recursive solving process of an upper layer model, iteration is continuously carried out until the calculation of the current vehicle group is completed, a group of control sequences is output, and a first control decision in a vehicle execution sequence in the group is output. And based on the model prediction control mechanism, re-identifying the vehicles in the group at the next moment and re-acquiring the state information of the vehicles, and repeatedly executing the optimization process of the double-layer optimization model, wherein the process is continuously repeated until the process is finished. Finally, the model predictive control and the double-layer optimization model jointly ensure the optimal running of the system vehicles, reduce the vehicle conflict times under the mixed traffic flow environment and effectively improve the vehicle passing efficiency and the driving comfort. In addition, the control mechanism of the method can optimize vehicle sequencing in real time and give real-time control decisions so as to ensure the system to be optimal.

Further, step S3, predicting the running tracks of the vehicles in the group by using a double-layer optimization model, making an optimization control decision based on the predicted running tracks, and performing optimization control on the vehicles according to the made optimization control decision, so that the vehicle closest to the traffic flow intersection point in the vehicles in the group can smoothly pass through the traffic flow intersection point, and the running tracks of the vehicles in the group are ensured to be optimal, which specifically includes:

s3-1, establishing the upper layer model, including:

s3-1-1, determining the division stage, the state variable and the decision variable of the upper layer model, which are as follows:

the upper layer model is divided: at the same time, n vehicles on the X road grouping optimization section need to sequentially merge into or pass through m +1 intervals among the vehicles on the Y road grouping optimization section, the behavior that each single vehicle on the X road grouping optimization section merges into or passes through the Y road is represented as a stage, the behavior that the k-th vehicle on the X road grouping optimization section merges into or passes through the Y road is recorded as the k-th stage, wherein k is 1,2,3, …, n;

state variables of the upper model: the k stage can be used for leading k to n vehicles on the X road grouping optimization section to merge into or pass through the vehicle interval number on the Y road grouping optimization section by s_kRepresents;

decision variables of the upper model: the decision made at each stage represents s of the k vehicle on the X road grouping optimization section at the k stage on the section capable of merging into or passing through the Y road grouping optimization section_kSelecting a particular xth of individual vehicle intervals_kThe vehicles are alternately converged or passed;

s3-1-2, determining a state transition equation, a cost function and an objective function of the upper layer model, wherein the details are as follows:

the state transition equation of the upper model is as follows:

setting the initial condition as s₀＝m+1；

The state transition equation of the upper model indicates that s is equal to 1 when k_k＝s₁The number of vehicle intervals for the 1 st vehicle on the X road grouping optimization section to enter or pass through the Y road grouping optimization section in the 1 st stage is m + 1; when k is 2,3, …, n, the (k-1) th vehicle on the X road grouping optimization section selects the (X) th vehicle at the (k-1) th stage_k-1The interval of each vehicle is used as a state variable s after the vehicle is converged or passes through a Y road_kA change in (c); s₀M +1 represents that the number of vehicle intervals available for vehicles on the X road grouping optimization section to merge into or pass through the Y road grouping optimization section in the initial state is m + 1;

cost function of the upper model:

cost function D of the upper model_k(s_k,x_k) Indicating a phase index function required for making a decision at the kth phase, whereinShowing that the k-th vehicle on the X road grouping optimization section is merged into or passes through the k-th vehicle or passes through the S on the Y road grouping optimization section under the action of the collaborative optimization control strategy in the k stage_kAll possible cost costs arising from individual vehicle intervals;the method is characterized in that vehicles which do not directly participate in vehicle convergence or vehicle confluence or vehicle crossing intersection process on a Y road grouping optimization section are subjected to influence of vehicle convergence or vehicle passing on a front vehicle, so that cost consumption is caused by vehicle speed adjustment of the vehicle due to the requirement of following safety of the vehicle;

an objective function of the upper model:

setting the initial condition as f₀(s₀)＝0；

Objective function f of the upper model_k(s_k) Representing the cumulative cost consumption of the system vehicles from stage 1 to stage k, f₀(s₀) 0 denotes that the system cost is 0 in the initial state;

s3-2, establishing the lower layer model, including:

s3-2-1, determining a micro following model, describing a following state of the vehicle by using the micro following model, and predicting a vehicle track; the following state of the vehicle comprises the speed, acceleration and position of the vehicle;

s3-2-2, establishing a condition constraint model for judging whether the kth vehicle in the path grouping optimization section at the kth stage X can smoothly merge into or pass through the Y path;

s3-2-3, based on the predicted vehicle track of the step S3-2-1, sequentially judging whether the k vehicle in the X road grouping optimization section of the k stage can merge into or pass through the S vehicle in the Y road grouping optimization section by the condition constraint model established in the step S3-2-2_kEach vehicle interval of the vehicle intervals;

s3-2-4, determining vehicles capable of being controlled in an optimized mode (namely intelligent networked vehicles) in the vehicles participating in the k stage as target vehicles, optimizing the running tracks of the target vehicles, resolving the optimization problem into an optimal control problem of continuous track state constraint under discrete time, and solving by using a dynamic matrix predictive control algorithm to obtain a cooperative optimization control strategy related to the target vehicles;

s3-2-5, calculating the k-th vehicle merging on the X road grouping optimization section or passing through the k-th vehicle merging or passing through the S on the Y road grouping optimization section in the k stage under the action of the cooperative optimization control strategy_kAll possible cost consumptions of one vehicle interval

S3-3, solving the double-layer optimization model:

s3-3-1, solving the upper layer model, and determining the decision made by each stage of the upper layer model when the accumulated cost consumption of the system vehicle is the lowest;

s3-3-2, reversely deducing the decision made at each stage of the upper layer model determined by the step S3-3-1 to obtain the vehicle optimization track of each single vehicle merging into or passing through the Y road on the X road grouping optimization section of the lower layer model;

and S3-4, obtaining the mixed traffic flow cooperative decision of the double-layer optimization model aiming at the system optimization by solving in the step S3-3, and acting the decision on the system vehicle to control the operation of the system vehicle.

In the double-layer optimization model, the upper layer model is a sequencing problem, and the lower layer model is a track optimization problem. In combination with the trajectory cost calculated by the lower model, the upper model can ensure the global optimization of the solution, that is, although the solution of the lower problem may converge to the local optimization, the global performance of the solution can be ensured by the upper model. This is also the most important advantage of using a two-layer optimization model.

Further, the micro-following model in step S3-2-1 includes a deterministic micro-following model and a stochastic micro-following model, the optimized controlled vehicle (i.e. the intelligent networked vehicle) adopts the deterministic micro-following model to perform action prediction, and the deterministic micro-following model can be a Gipps following model or an IDM/EIDM following model; the non-optimistically controlled vehicle (i.e., a conventionally driven vehicle) employs the stochastic micro-following model to simulate randomness of driving behavior, which may be randomly modified for following based on a deterministic micro-following model.

In order to define the driving behavior of the vehicle more accurately and simulate the driving behavior, a deterministic microscopic follow-up model is adopted for the intelligent networked vehicle, and a stochastic microscopic follow-up model is adopted for the traditional driving vehicle.

Further, the stochastic micro-following model is as follows:

v_n(t+δ)＝min(v^e+θ₁,v_safe+θ₂)，

θ₁～N(0，σ₁^2)，

θ₂～N(0，σ₂^2)，

wherein v is_safeRepresenting a safe speed in the stochastic micro-follow model; b represents the deceleration; δ represents the interval between two adjacent time slices; n denotes the number of the non-optimally controlled vehicle, n-1 denotes the number of the preceding vehicle followed by the non-optimally controlled vehicle n, v_n-1(t) represents the speed of the preceding vehicle n-1 at time t during the vehicle-following; l_n(t) represents the position of the non-optimistically controllable vehicle n during the vehicle-following at time t; l_n-1(t) represents the position of the preceding vehicle n-1 during the vehicle-following at time t; l_aRepresenting a vehicle length; l₀Represents a vehicle following minimum distance; v. of_n(t + delta) represents the speed of the non-optimistically controllable vehicle n at the moment t + delta during the following of the vehicle, taking v^e+θ₁And v_safe+θ₂The smaller of the two; v. of^eRepresenting a desired speed; theta₁、θ₂Are all constant and all follow normal distribution; σ is set in consideration of the fact that the fluctuation of the speed of the vehicle will be small when the vehicle is at a desired speed and the fluctuation of the speed will be large when the vehicle is forced to decelerate to a safe speed₁<σ₂。

Further, in step S3-2-2, establishing a conditional constraint model for determining whether the kth vehicle in the kth stage X road grouping optimization section can smoothly merge into or pass through the Y road specifically includes:

assuming that the k-th vehicle on the X-way grouping optimization section in the k-th stage is recorded as a vehicle k, the vehicle is converged into or passes through a Y-way by the interval between two vehicles of the continuous traffic flow on the Y-way, wherein the two vehicles of the continuous traffic flow on the Y-way respectively use the k^′ ₁And k^′ ₂Is shown, wherein the vehicle k^′ ₁Indicating the preceding vehicle, vehicle k^′ ₂The rear vehicle is represented by:

wherein E is_k(t) a conflict cooperative utility function for reflecting whether a vehicle k on the X road can smoothly merge into or pass through the Y road; | u_k(t) | denotes an absolute value of acceleration or deceleration of the vehicle k at time t on the X-road,indicating vehicle k on Y road^′ ₂The absolute value of acceleration or deceleration at time t;indicating vehicle k on the X road and vehicle k on the Y road at time t^′ ₁The relative distance between the two or more of them,indicating vehicle k on the X road and vehicle k on the Y road at time t^′ ₂The relative distance therebetween; l_k(t) represents the position of the vehicle k on the X road at time t,indicating vehicle k on Y road^′ ₁At the position of the instant t of time,indicating vehicle k on Y road^′ ₂Position at time t,/_aIs the vehicle length;representing a minimum workshop between an optimally controllable vehicle (i.e. an intelligent networked vehicle) and its following front vehicleDistance;represents the minimum inter-vehicle distance between a non-optimally controllable vehicle (i.e., a conventionally driven vehicle) and a vehicle ahead of the vehicle; when vehicle k is about to merge into or pass through the Y road, vehicle k on the X road actually follows vehicle k on the Y road^′ ₁Running, and vehicles k on the Y-road^′ ₂The vehicles k on the actual following X road run, and the acceleration or deceleration of the vehicles k can be calculated according to the microcosmic following model; b_safeRepresents a maximum allowable deceleration; phi_AThe method comprises the steps of (1) optimizing a controllable vehicle set (namely an intelligent networking vehicle set); phi_HA non-optimally controlled vehicle set (i.e., a traditional driving vehicle set); eta₁Representing a safety factor; eta₂Representing a polite coefficient;

polite coefficient eta₂The expression of (a) is as follows:

wherein v is_kIs the speed of the vehicle k, v_thIs a given threshold speed, β₁And beta₂Is a constant;

m for cooperative decision_k(t+T_c) Is represented as follows:

wherein m is_k(t+T_c) A value of 1 indicates that the vehicle k on the X road is at T + T_cCan smoothly converge into or pass through a Y road at any moment; m is_k(t+T_c) A value of 0 indicates that the vehicle k on the X road is at T + T_cThe user can not smoothly converge into or pass through the Y road at any time.

After the gap of the Y road is selected, under the premise of ensuring safety, the courtesy problem of convergence or passing is considered, and the Y road can be adjusted according to different system emphasis points through the weight coefficient.

Further, the k-th vehicle on the X-road grouping optimization section is merged into or passes through the k-th vehicle or passes through the s on the Y-road grouping optimization section in the k-th stage under the action of the collaborative optimization control strategy_kAll possible cost costs resulting from the separation of individual vehiclesThe method specifically comprises the following steps:

defining a vehicle combination K: predicting a vehicle k on an X road, merging into or passing through a Y road an interval between two vehicles passing a continuous traffic flow on the Y road, the two vehicles on the continuous traffic flow on the Y road being respectively usedAndshow, wherein the vehicleIndicating predicted vehicle ahead and vehicleRepresenting a predicted vehicle; further defining a vehicle combination K, representing the vehicle combination directly participating in the K stage on the X road and the Y road, and K belongs to { K ∈ [ K ]₁,K₂,K₃Vehicle combinationIndicating that the vehicle participating in the k-th stage has a vehicleVehicle k and vehicleVehicle combinationThe vehicle participating in the kth stage comprises a vehicle k and a vehicleAt this time, the x-th road is positioned on the Y road_kThe interval of each vehicle has no front vehicle; vehicle combinationIndicating that the vehicle participating in the k-th stage has a vehicleVehicle k, now on the Xth road of Y_kNo rear vehicle participates at the interval of each vehicle;

thenThe expression of (a) is as follows:

wherein i represents the ith vehicle in the current vehicle combination and is recorded as a vehicle i;representing the speed of the vehicle i at the moment t in the k phase; v. of^eRepresenting a desired speed; t is t₀Represents the current time; t is_PRepresents a predicted time domain length;representing the speed of a rear vehicle on a Y road in the k-1 stage at the time t; x is the number of_k>x_k-1+1 indicates that at least one vehicle clearance still exists between the vehicle k on the k-th stage X road and the vehicle clearance on the Y road occupied by the vehicle k-1 on the k-1 stage X road; x is the number of_k＝x_k-1+1 represents that the vehicle k on the k-th stage X road and the vehicle k-1 on the k-1 stage X road occupy two consecutive vehicle gaps on the Y road; x is the number of_k＝x_k-1Indicating that vehicle k on the k stage X road and vehicle k-1 on the k-1 stage X road were selectedThe vehicle clearance on the selected Y road is the same vehicle clearance.

Further, the vehicles which do not directly participate in the process of vehicle confluence or vehicle crossing of the intersection on the Y road grouping optimization section are influenced by vehicle confluence or crossing of the front vehicles, so that cost consumption is caused by vehicle speed adjustment of the self vehicle due to the following safety requirementThe expression is as follows:

wherein i represents the ith vehicle in the current vehicle combination and is recorded as a vehicle i;representing the speed of the vehicle i at the moment t in the k phase; v. of^eRepresenting a desired speed; t is t₀Represents the current time; t is_PRepresents a predicted time domain length; vehicle combination K^′The vehicle between the rear vehicle at the k-1 stage and the front vehicle at the k stage on the Y road grouping optimization section is represented, namely the vehicle which does not directly participate in the process of vehicle confluence or vehicle crossing of the intersection; x is the number of_k>x_k-1+1 indicates that there is at least one more vehicle clearance between vehicle k on the k-th stage X road and the vehicle clearance on the Y road occupied by vehicle k-1 on the k-1 stage X road.

(once the solution to the underlying model problem is complete,will return to the upper model; then, the upper layer model problem is solved continuously, and finally the optimization control problem of multiple CAVs (vehicles capable of being controlled in an optimized mode, namely intelligent networked vehicles) in the current group is completed. )

To pairCovers the influence of vehicles directly or indirectly participating in vehicle confluence or vehicle crossing processes on the overall traffic flow. By minimizing the influence, the global optimality of the upper-layer problem solution is ensured, and the global optimality of the control decision calculated by the whole control system is further ensured.

Further, step S3-2-4, determining an optimally controllable vehicle from the vehicles participating in the k-th stage as a target vehicle, optimizing the traveling trajectory of the target vehicle, and solving the optimization problem as an optimal control problem constrained by a continuous trajectory state in discrete time, and using a dynamic matrix predictive control algorithm to obtain a collaborative optimization control strategy for the target vehicle, specifically including:

defining a vehicle combination K: predicting a vehicle k on an X road, merging into or passing through a Y road an interval between two vehicles passing a continuous traffic flow on the Y road, the two vehicles on the continuous traffic flow on the Y road being respectively usedAndshow, wherein the vehicleIndicating predicted vehicle ahead and vehicleRepresenting a predicted vehicle; further defining a vehicle combination K, representing the vehicle combination directly participating in the K stage on the X road and the Y road, and K belongs to { K ∈ [ K ]₁,K₂,K₃Vehicle combinationIndicating that the vehicle participating in the k-th stage has a vehicleVehicle k and vehicleVehicle combinationThe vehicle participating in the kth stage comprises a vehicle k and a vehicleAt this time, the x-th road is positioned on the Y road_kThe interval of each vehicle has no front vehicle; vehicle combinationIndicating that the vehicle participating in the k-th stage has a vehicleVehicle k, now on the Xth road of Y_kNo rear vehicle participates at the interval of each vehicle;

defining a dynamic matrix predictive control model as follows:

the above equations (1) to (4) are described:

equation (1) is an objective function of a dynamic matrix predictive control model, whereRepresenting the difference between the actual vehicle speed and the desired speed; introduction ofThe aim is to make the vehicle acceleration as soft as possible, taking into account the comfort of the driver; gamma ray₁、γ₂Is a weight coefficient; weight coefficient gamma₂The larger the dynamic matrix predictive control model is, the more the objective function of the dynamic matrix predictive control model focuses on optimizing the comfort of the controlled vehicle; on the contrary, the objective function of the dynamic matrix predictive control model focuses on optimizing the comfort level of the whole traffic flow;

the formula (2), the formula (3) and the formula (4) are constraint conditions of the dynamic matrix prediction control model;

equation (2) describes the controlled variable constraint, where a_minAnd a_maxRepresenting an allowable maximum deceleration and a maximum acceleration, respectively, whose values are determined based on the comfort of the controlled vehicle;

equation (3) is a vehicle speed state constraint, whereRepresenting a vehicle following speed matrix function, wherein the constraint can ensure the safety of the following behavior in the control process, namely collision avoidance constraint; wherein the content of the first and second substances,indicating that vehicle i in vehicle combination K is at t₀Position matrix of time of day comprising t₀Is at the momentPredicting the position of a vehicle i in the vehicle combination K at any moment;indicating that vehicle i in vehicle combination K is at t₀Velocity matrix of time of day comprising t₀Is at the momentTime-to-time vehicleCombining the speed predictions of vehicle i in K;indicates that the preceding vehicle in the vehicle combination K is following the vehicle i at t₀Velocity matrix of time of day comprising t₀Is at the momentPredicting the speed of a vehicle before a vehicle i in the vehicle combination K at any moment;

equation (4) defines the final state constraint that vehicle k on the kth stage X road is inThe conflicting collaborative utility function value at a time may only be at or above 0Smoothly converge into or pass through a Y road at any moment;

to complete the iterative process of dynamic matrix predictive control, the state transition process is defined as follows:

in the above-mentioned formula (5),indicating that vehicle i in vehicle combination K is at t₀Velocity v of time_i(t₀) And position l_i(t₀) A set of (a);

the above equations (6) to (8) are the state transition equations of the non-optimally controlled vehicle, where K_H＝{i|α_i0 n ∈ K, and α_iIs the vehicle type, alpha if the vehicle is an optimally controllable vehicle_iIs 1, alpha if the vehicle is a non-optimally controllable vehicle_iIs 0;

equation (6) is a function of the position of the non-optimally controlled vehicle;

equation (7) is a speed function for a non-optimally controlled vehicle;

equation (8) is a general formula for a microscopic follow-up model, i.e., the acceleration of a non-optimally controlled vehicle is a function of the clearance between the vehicle and its following predecessor;

C_v＝[1 0]； (11)

C_s＝[0 1]； (12)

the above equations (9) to (12) are the state transition equations for optimally controlling the vehicle, where K_A＝{i|α_i＝1∩i∈K}，α_iIs the vehicle type, alpha if the vehicle is an optimally controllable vehicle_iIs 1, alpha if the vehicle is a non-optimally controllable vehicle_iIs 0;

the formula (9) includes a formula (9a) and a formula (9b), in whichIt is described that the controlled speed and position of the vehicle at time t can be optimally controlled, sinceThere is a recurrence relation shown by the formula (9a) and the formula (9b), and therefore at an arbitrary predicted timingCan be prepared fromAndcarrying out representation; matrix arrayMatrix arrayA corresponding state transition matrix;is to optimally control the acceleration decision of the vehicle at time t, i.e.Is the optimal control variable;

in the formula (10),is a set of column vectors that can optimally control the acceleration decision of the vehicle at each predicted time point;

the compound of the formula (11),is a set of column vectors that can optimally control the controlled vehicle speed of the vehicle at each predicted time point, whereinIs the corresponding state transition matrix and is,is an auxiliary matrix in the derivation process;

in the formula (12),is a set of column vectors that optimally controls the controlled vehicle position of the vehicle at each predicted time point, whereinIs the corresponding state transition matrix and is,is the auxiliary matrix in the derivation process.

(due to the fact thatAndall can be composed ofThe trajectory optimization problem solved by the dynamic matrix predictive control can be regarded as an independent variableThe problem of non-linear programming of (a); the problem contains nonlinear state transition equations (6-8)) and complex state constraints (equations (3) and (4)), so numerical solution Sequential Quadratic Programming (SQP) is used to solve the problem. )

The above further describes the process of the dynamic matrix predictive control algorithm. The method does not need discretization processing on the control quantity, and only discretization processing is carried out on time. Because the method can adopt a numerical solution, the calculation efficiency is obviously improved, and the improvement is obvious along with the increase of the calculation scale, thereby being more in line with the requirements in practical use. In addition, the track optimized by the numerical solution is smoother. Although there is a possibility of convergence to a local optimum, the influence on the optimization effect of the whole system is not obvious due to the limitation of an upper layer model.

The present invention will be described in further detail with reference to the following detailed description and the accompanying drawings, which are not intended to limit the scope of the invention.

Drawings

Fig. 1 is a framework diagram of a hybrid traffic flow cooperative optimization control method based on model predictive control according to an embodiment of the present invention.

Fig. 2 is a schematic diagram of traffic flow grouping according to the road grouping optimization section in the embodiment of the invention.

Fig. 3 is a time domain division diagram of an embodiment of the present invention.

Detailed Description

The following description of the embodiments of the present invention will be made with reference to the accompanying drawings.

A mixed traffic flow cooperative optimization control method based on model predictive control adopts a double-layer optimization model based on model predictive control to carry out mixed traffic flow cooperative decision control, and is suitable for different traffic scenes in mixed traffic flows under the condition that two crossed one-way roads are arranged and no vehicle driving signal is indicated in front of the intersection; the model predictive control means that real-time closed-loop control is realized by establishing a system architecture so as to solve the problem of randomness of driving behaviors in actual situations; the double-layer optimization model comprises an upper layer model and a lower layer model, the upper layer model is a vehicle sequencing problem solved by dynamic programming recursion, the lower layer model is a track optimization problem solved by a dynamic matrix prediction algorithm, and a track optimization result of each single vehicle in the lower layer model is used as one input in the dynamic programming recursion solving process of the upper layer model; ensuring that the operation of the system vehicle is optimal through the model predictive control and the double-layer optimization model;

the method comprises the following steps:

s1, determining road grouping optimization sections and performing time domain division;

the method comprises the following steps that an X road and a Y road are assumed to be two one-way roads with intersections, no vehicle driving signal indication exists before the intersections, and the road grouping optimization sections comprise an X road grouping optimization section and a Y road grouping optimization section; setting a section of road before vehicles on the X road reach the traffic flow intersection as an X road grouping optimization section, and setting a section of road before vehicles on the Y road reach the traffic flow intersection as a Y road grouping optimization section;

s2, carrying out traffic flow grouping according to the road grouping optimization section, determining vehicles running on the road grouping optimization section at the same time as vehicles in the group, and acquiring real-time state information of the vehicles in the group;

s3, predicting the running tracks of the vehicles in the group by adopting a double-layer optimization model, making an optimization control decision according to the running tracks, and performing optimization control on the vehicles according to the optimization control decision, so that the vehicle closest to the traffic flow junction in the vehicles in the group can smoothly pass through the traffic flow junction, and the running track of the vehicle in the group is ensured to be optimal;

and S4, performing optimization control on the running tracks of the vehicles in the group driving on the road grouping optimization section at the next moment by repeatedly executing the steps S2 and S3, and executing in a circulating mode until all the vehicles drive out of the road grouping optimization section.

Examples

The method adopts a double-layer optimization model based on model prediction control to carry out mixed traffic flow cooperative decision control, is suitable for different traffic scenes in mixed traffic flow under the condition that two crossed one-way roads are provided and no vehicle driving signal is indicated in front of the intersection, and comprises ramp vehicle convergence on an expressway, intersection vehicle confluence and vehicle passing through the intersection. The method of the present invention will be further explained based on the condition of vehicle convergence on the ramp of the expressway (in this example, the ramp is an X-road, and the main road is a Y-road). Fig. 1 is a framework diagram of a hybrid traffic flow cooperative optimization control method based on model predictive control in this example. The following is described in detail in terms of the steps:

and S1, determining the road grouping optimization section and performing time domain division.

And setting a section of road before vehicles on the ramp reach the traffic flow junction as a ramp grouping optimization section, and setting a section of road before vehicles on the main road reach the traffic flow junction as a main road grouping optimization section.

Fig. 2 is a schematic diagram illustrating grouping of traffic flows according to the road grouping optimization section in the present example, and fig. 2 shows a link structure of the highway ramp vehicle convergence condition in the present example. In this example, the position of the virtual vehicle in the link configuration diagram shown in fig. 2 is 0m, and the traffic direction is determined as the positive direction.

Determining a ramp grouping optimization section: as shown in fig. 2, a road between 225m to 500m from the virtual vehicle on the ramp traffic flow upward is determined as the ramp grouping optimized section.

Determining a main trunk grouping optimization section: as shown in fig. 2, a road, which is located between 200m and 550m from the virtual vehicle in the main road traffic flow upward, is determined as the main road grouping optimization section.

FIG. 3 isThe time domain division chart in this example describes the control time domain of a vehicle from entering the control area to exiting the control areaAnd prediction time domain (T)_p) The change process diagram of (2). The abscissa in fig. 3 is time, and the ordinate is the number of ramp cars. Assuming that there are n ramp cars in total, the letter k is the currently calculated ramp car number.

Is the convergence decision time point. The time point is the time when the vehicle reaches the confluence decision point by parallel simulation under the assumption that the vehicle runs in a non-cooperative state, namely

Controlling time domainI.e. the time range in which the decision is actually performed. In the case of a ramp vehicle,is defined from the current time point t₀To the point in time when the ramp vehicle k makes a decision whether to mergeNamely, it isIn the case of a main road vehicle,is defined as Thus, for the same group K, the ramp vehicle K and the following vehicles on the corresponding arterial roadAnd the front vehicleAre identical.

Prediction time domain (T)_p) I.e. the time frame for which the prediction is made for each vehicle. T is_pIs defined from the data acquisition time point t₀Time point of convergence decision made by the last ramp vehicleNamely, it is

Executing a time domain: the coverage time range of the proximity control sequence, which in this example is 1 second.

At a point in time t₀Time domain of control of each vehicle on the rampAre all different, but the predicted time domain (T) of each vehicle on the ramp_p) Are consistent. At a point in time t₀+ τ, as the vehicle approaches the point of convergence decisionControl time domain of each vehicleAre being shortened.

And S2, carrying out traffic flow grouping according to the road grouping optimization section, determining vehicles running on the road grouping optimization section at the same time as vehicles in the group, and acquiring real-time state information of the vehicles in the group.

Fig. 2 is a schematic diagram showing traffic flow grouping according to the road grouping optimization section in this example. An example of a packet is shown in fig. 2 (a). The point 250m in fig. 2(a) is the starting point of the merging action, that is, if it is judged that the ramp cars can perform the merging, the action is started from the point 250 m.

M(t)＝{k；225<l_k(t)<500} (1-1)

The formulas (1-1) and (1-2) define a grouping rule, wherein M (t) represents a main road set at the time t, and R (t) represents a ramp set at the time t. Since the vehicle position is obtained again every second, the grouping information is updated every second, i.e., m (t), r (t) are changed according to different specific time in the simulation process. As can be seen from fig. 2, the division range of the main road train set is significantly larger than that of the ramp train set, which is mainly to ensure that the first vehicle in the ramp group can bring the vehicle in front of the main road into the calculation range, and the last vehicle in the ramp group can bring the vehicle behind the main road into the calculation range, so as to amplify the optimization effect as much as possible. In fig. 2, the cars 3 and 4 are t ═ t₀Out-of-packet state at time (as shown in fig. 2 (a)) to when t equals t₀The process of entering a packet at time +1 (as shown in fig. 2 (b)) is an example of the above-described scrolling process.

S3, predicting the running tracks of the vehicles in the group by adopting a double-layer optimization model, making an optimization control decision according to the running tracks, and performing optimization control on the vehicles according to the optimization control decision, so that the vehicles closest to the traffic flow junction in the vehicles in the group can smoothly pass through the traffic flow junction, and the running tracks of the vehicles in the group are ensured to be optimal. The method specifically comprises the following steps:

s3-1, establishing an upper layer model, including:

s3-1-1, determining the division stage, the state variable and the decision variable of the upper layer model, which are as follows:

and (3) dividing the upper layer model: at the same moment, n vehicles on the ramp grouping optimization section need to sequentially merge into m +1 intervals among the vehicles on the main road grouping optimization section, the behavior that each single vehicle on the ramp grouping optimization section merges into the main road is represented as a stage, and the behavior that the kth vehicle on the ramp grouping optimization section merges into the main road is recorded as the kth stage, wherein k is 1,2,3, …, n;

state variables of the upper model: in the k stage, the number of vehicle intervals for leading the k to n vehicles on the ramp grouping optimization section to merge into the main road grouping optimization section is s_kRepresents;

decision variables of the upper model: the decision made at each stage represents s of the k vehicle on the kth ramp grouping optimization section on the afflux main road grouping optimization section_kSelecting a particular xth of individual vehicle intervals_kThe vehicles are alternately merged.

S3-1-2, determining a state transition equation, a cost function and an objective function of the upper layer model, wherein the details are as follows:

the state transition equation of the upper model:

setting the initial condition as s₀＝m+1；

The state transition equation of the upper layer model shows that when k is 1, s_k＝s₁The number of vehicle intervals for 1 vehicle on the ramp grouping optimization section to converge into the main road grouping optimization section in the 1 st stage is m + 1; when k is 2,3, …, n, the k-1 vehicle on the ramp grouping optimization section at the k-1 stage selects the x-th vehicle_k-1The interval of each vehicle is used as a state variable s after being converged into the main road_kA change in (c); s₀M +1 represents that the number of vehicle intervals for vehicles on the ramp grouping optimization section to merge into the main road grouping optimization section in the initial state is m + 1;

cost function of upper model:

cost function D of upper model_k(s_k,x_k) Indicating a phase index function required for making a decision at the kth phase, whereinS represents that the k-th vehicle on the ramp grouping optimization section is merged into the grouping optimization section for the k-th vehicle to be merged into the main road grouping optimization section under the action of the cooperative optimization control strategy in the k-th stage_kAll possible cost costs arising from individual vehicle intervals;the method is characterized in that the method is that vehicles which do not directly participate in a vehicle convergence process on a main road grouping optimization section are influenced by vehicle convergence of a front vehicle, so that cost consumption is caused by vehicle speed adjustment of the self vehicle due to the following safety requirement;

objective function of upper model:

setting the initial condition as f₀(s₀)＝0；

Objective function f of upper model_k(s_k) Representing the cumulative cost consumption of the system vehicles from stage 1 to stage k, f₀(s₀) 0 indicates that the system cost is 0 in the initial state.

In the kth stage, under the action of a collaborative optimization control strategy, the kth vehicle on the ramp grouping optimization section is converged into the s section which can be converged into the main road by the kth vehicle_kAll possible cost costs resulting from the separation of individual vehiclesThe method specifically comprises the following steps:

defining a vehicle combination K: predicting vehicles k on the ramp, merging the interval between two vehicles passing through the continuous traffic flow on the main road into the main roadFor two vehicles in a continuous flowAndshow, wherein the vehicleIndicating predicted vehicle ahead and vehicleRepresenting a predicted vehicle; further defining a vehicle combination K which represents the vehicle combination directly participating in the K stage on the ramp and the main road, and K belongs to { K ∈ [ K ]₁,K₂,K₃Vehicle combinationIndicating that the vehicle participating in the k-th stage has a vehicleVehicle k and vehicleVehicle combinationThe vehicle participating in the kth stage comprises a vehicle k and a vehicleNow located on the x-th main road_kThe interval of each vehicle has no front vehicle; vehicle combinationIndicating that the vehicle participating in the k-th stage has a vehicleVehicle k, now on the x-th road_kThe vehicles have no rear vehicle at the interval；

ThenThe expression of (a) is as follows:

wherein i represents the ith vehicle in the current vehicle combination and is recorded as a vehicle i;representing the speed of the vehicle i at the moment t in the k phase; v. of^eRepresenting a desired speed; t is t₀Represents the current time; t is_PRepresents a predicted time domain length;representing the speed of a rear vehicle on the main road at the k-1 stage at the time t; x is the number of_k>x_k-1+1 indicates that at least one vehicle gap still exists between the vehicle k on the kth stage ramp and the vehicle gap on the main road occupied by the vehicle k-1 on the k-1 stage ramp; x is the number of_k＝x_k-1+1 represents that the vehicle clearance on the main road occupied by the vehicle k on the kth stage ramp and the vehicle k-1 on the kth-1 stage ramp is two consecutive vehicle clearances; x is the number of_k＝x_k-1It is indicated that the vehicle gap on the selected main road of the vehicle k on the kth stage ramp and the vehicle k-1 on the kth stage ramp is the same vehicle gap.

Secondly, vehicles which do not directly participate in the vehicle convergence process in the grouped and optimized section of the main road are influenced by the vehicle convergence of the front vehicles, so that the cost consumption is caused by the vehicle speed adjustment of the self vehicle due to the following safety requirementThe expression is as follows:

wherein i represents the ith vehicle in the current vehicle combination and is recorded as a vehicle i;representing the speed of the vehicle i at the moment t in the k phase; v. of^eRepresenting a desired speed; t is t₀Represents the current time; t is_PRepresents a predicted time domain length; vehicle combination K^′Representing vehicles between a rear vehicle at the k-1 stage and a front vehicle at the k stage on the main road grouping optimization section, namely vehicles which do not directly participate in the vehicle confluence process; x is the number of_k>x_k-1+1 indicates that there is at least one more vehicle clearance between vehicle k on the kth stage ramp and the vehicle clearance on the arterial road occupied by vehicle k-1 on the k-1 stage ramp.

S3-2, establishing the lower layer model, including:

s3-2-1, determining a microscopic follow-up model, describing a follow-up state of the vehicle by using the microscopic follow-up model, and predicting a vehicle track; the following state of the vehicle includes the speed, acceleration, and position of the vehicle.

The micro-following model comprises a deterministic micro-following model and a stochastic micro-following model, the vehicle can be optimally controlled to adopt the deterministic micro-following model to predict the action, and the deterministic micro-following model can be a Gipps following model or an IDM/EIDM following model; the unoptimizable control vehicle adopts a stochastic micro-following model to simulate the randomness of the driving behavior, and the stochastic micro-following model can carry out stochastic following modification on the basis of the deterministic micro-following model.

The stochastic microscopic follow model was as follows:

v_n(t+δ)＝min(v^e+θ₁,v_safe+θ₂)，

θ₁～N(0，σ₁^2)，

θ₂～N(0，σ₂^2)，

wherein v is_safeRepresenting a safe speed in a stochastic micro-following model; b represents the deceleration; δ represents the interval between two adjacent time slices; n denotes the number of the non-optimally controlled vehicle, n-1 denotes the number of the preceding vehicle followed by the non-optimally controlled vehicle n, v_n-1(t) represents the speed of the preceding vehicle n-1 at time t during the vehicle-following; l_n(t) represents the position of the non-optimistically controllable vehicle n during the vehicle-following at time t; l_n-1(t) represents the position of the preceding vehicle n-1 during the vehicle-following at time t; l_aRepresenting a vehicle length; l₀Represents a vehicle following minimum distance; v. of_n(t + delta) represents the speed of the non-optimistically controllable vehicle n at the moment t + delta during the following of the vehicle, taking v^e+θ₁And v_safe+θ₂The smaller of the two; v. of^eRepresenting a desired speed; theta₁、θ₂Are all constant and all follow normal distribution; σ is set in consideration of the fact that the fluctuation of the speed of the vehicle will be small when the vehicle is at a desired speed and the fluctuation of the speed will be large when the vehicle is forced to decelerate to a safe speed₁<σ₂。

S3-2-2, establishing a condition constraint model for judging whether the kth vehicle on the kth stage ramp grouping optimization section can smoothly merge into the main road. (i.e., the deterministic conflux decision model is established as described in FIG. 1.) specifically includes:

assuming that the kth vehicle on the kth stage ramp grouping optimization section is recorded as a vehicle k, merging the interval between two vehicles of continuous traffic flow on the main road into the main road, wherein the two vehicles of the continuous traffic flow on the main road respectively use the k^′ ₁And k^′ ₂Is shown, wherein the vehicle k^′ ₁Indicating the preceding vehicle, vehicle k^′ ₂The rear vehicle is represented by:

wherein E is_k(t) a conflict cooperative utility function is represented to reflect whether a vehicle k on the ramp can smoothly merge into the main road; | u_k(t) | represents the absolute value of the acceleration or deceleration of the vehicle k on the ramp at time t,indicating vehicles k on the arterial road^′ ₂The absolute value of acceleration or deceleration at time t;denotes at time t, vehicle k on the ramp and vehicle k on the main road^′ ₁The relative distance between the two or more of them,denotes at time t, vehicle k on the ramp and vehicle k on the main road^′ ₂The relative distance therebetween; l_k(t) represents the position of vehicle k on the ramp at time t,indicating vehicles k on the arterial road^′ ₁At the position of the instant t of time,indicating vehicles k on the arterial road^′ ₂Position at time t,/_aIs the vehicle length;representing a minimum inter-vehicle distance between the optimally controllable vehicle and a vehicle ahead of the optimally controllable vehicle;representing a minimum inter-vehicle distance between the non-optimally controllable vehicle and a preceding vehicle following the non-optimally controllable vehicle; when vehicle k is ready to merge into the main road, vehicle k on the ramp actually follows vehicle k on the main road^′ ₁Running while the vehicle k on the main road^′ ₂K-way for vehicles on actual following rampAnd the acceleration or deceleration of the wheels can be calculated according to a microscopic follow-up model; b_safeRepresents a maximum allowable deceleration; phi_AA set of optimally controllable vehicles; phi_HA set of non-optimally controlled vehicles; eta₁Representing a safety factor; eta₂Representing a polite coefficient;

polite coefficient eta₂The expression of (a) is as follows:

wherein v is_kIs the speed of the vehicle k, v_thIs a given threshold speed, β₁And beta₂Is a constant;

m for cooperative decision_k(t+T_c) Is represented as follows:

wherein m is_k(t+T_c) A value of 1 indicates that k is T + T for the vehicle on the ramp_cCan be smoothly converged into the main trunk road at any time; m is_k(t+T_c) A value of 0 indicates that the vehicle k on the ramp is at T + T_cCan not be smoothly converged into the main trunk road at any time.

S3-2-3, based on the vehicle track predicted in the step S3-2-1, sequentially judging whether the kth vehicle in the kth stage ramp grouping optimization section can be converged into the S-branch road grouping optimization section for being converged into the main road grouping optimization section by the condition constraint model established in the step S3-2-2_kEach vehicle interval of the individual vehicle intervals.

S3-2-4, determining vehicles capable of being controlled in an optimized mode in the vehicles participating in the k stage as target vehicles, optimizing the running tracks of the target vehicles, solving the optimization problem into an optimal control problem of continuous track state constraint in discrete time, and solving by using a dynamic matrix predictive control algorithm to obtain a collaborative optimization control strategy about the target vehicles. (i.e., the establishment of the model for optimizing the continuous trajectory in discrete time based on dynamic matrix predictive control, which is described in fig. 1) specifically includes:

defining a dynamic matrix predictive control model as follows:

the above equations (1) to (4) are described:

equation (1) is an objective function of a dynamic matrix predictive control model, whereRepresenting the difference between the actual vehicle speed and the desired speed; introduction ofThe aim is to make the vehicle acceleration as soft as possible, taking into account the comfort of the driver; gamma ray₁、γ₂Is a weight coefficient; weight coefficient gamma₂The larger the dynamic matrix predictive control model is, the more the objective function of the dynamic matrix predictive control model focuses on optimizing the comfort of the controlled vehicle; on the contrary, the objective function of the dynamic matrix predictive control model focuses on optimizing the comfort level of the whole traffic flow;

the formula (2), the formula (3) and the formula (4) are constraint conditions of the dynamic matrix prediction control model;

equation (2) describes the controlled variable constraint, where a_minAnd a_maxRepresenting respectively the maximum deceleration and the maximum acceleration allowable,its value is determined based on the comfort of the controlled vehicle;

equation (3) is a vehicle speed state constraint, whereRepresenting a vehicle following speed matrix function, wherein the constraint can ensure the safety of the following behavior in the control process, namely collision avoidance constraint; wherein the content of the first and second substances,indicating that vehicle i in vehicle combination K is at t₀Position matrix of time of day comprising t₀Is at the momentPredicting the position of a vehicle i in the vehicle combination K at any moment;indicating that vehicle i in vehicle combination K is at t₀Velocity matrix of time of day comprising t₀Is at the momentPredicting the speed of the vehicle i in the vehicle combination K at any moment;indicates that the preceding vehicle in the vehicle combination K is following the vehicle i at t₀Velocity matrix of time of day comprising t₀Is at the momentPredicting the speed of a vehicle before a vehicle i in the vehicle combination K at any moment;

equation (4) defines the final state constraint that vehicle k on the kth stage ramp is onThe conflicting collaborative utility function value at a time may only be at or above 0Smoothly merging into a main road all the time;

to complete the iterative process of dynamic matrix predictive control, the state transition process is defined as follows:

in the above-mentioned formula (5),indicating that vehicle i in vehicle combination K is at t₀Velocity v of time_i(t₀) And position l_i(t₀) A set of (a);

the above equations (6) to (8) are the state transition equations of the non-optimally controlled vehicle, where K_H＝{i|α_i0 n ∈ K, and α_iIs the vehicle type, alpha if the vehicle is an optimally controllable vehicle_αIs 1, alpha if the vehicle is a non-optimally controllable vehicle_iIs 0;

equation (6) is a function of the position of the non-optimally controlled vehicle;

equation (7) is a speed function for a non-optimally controlled vehicle;

equation (8) is a general formula for a microscopic follow-up model, i.e., the acceleration of a non-optimally controlled vehicle is a function of the clearance between the vehicle and its following predecessor;

C_v＝[1 0]； (11)

C_s＝[0 1]； (12)

the above equations (9) to (12) are the state transition equations for optimally controlling the vehicle, where K_A＝{i|α_i＝1∩i∈K}，α_iIs the vehicle type, alpha if the vehicle is an optimally controllable vehicle_iIs 1, alpha if the vehicle is a non-optimally controllable vehicle_iIs 0;

the formula (9) includes a formula (9a) and a formula (9b), in whichIt is described that the controlled speed and position of the vehicle at time t can be optimally controlled, sinceThere is a recurrence relation shown by the formula (9a) and the formula (9b), and therefore at an arbitrary predicted timingCan be prepared fromAndcarrying out representation; matrix arrayMatrix arrayA corresponding state transition matrix;is to optimally control the acceleration decision of the vehicle at time t, i.e.Is the optimal control variable;

in the formula (10),is a set of column vectors that can optimally control the acceleration decision of the vehicle at each predicted time point;

the compound of the formula (11),is a set of column vectors that can optimally control the controlled vehicle speed of the vehicle at each predicted time point, whereinIs the corresponding state transition matrix and is,is an auxiliary matrix in the derivation process;

in the formula (12),is a set of column vectors that optimally controls the controlled vehicle position of the vehicle at each predicted time point, whereinIs the corresponding state transition matrix and is,is the auxiliary matrix in the derivation process.

S3-2-5, calculating S for the k-th vehicle on the ramp grouping optimization section to merge into the main road grouping optimization section under the action of the cooperative optimization control strategy in the k-th stage_kAll possible cost consumptions of one vehicle interval

S3-3, solving a double-layer optimization model:

s3-3-1, solving the upper layer model, and determining the decision made by each stage of the upper layer model when the accumulated cost consumption of the system vehicle is the lowest.

S3-3-2, and the decision made at each stage of the upper layer model determined by the step S3-3-1 is reversely deduced to obtain the vehicle optimization track of each single vehicle merging into or passing through the main road on the ramp grouping optimization section of the lower layer model.

And S3-4, obtaining a mixed traffic flow cooperative decision of the double-layer optimization model by solving in the step S3-3, aiming at system optimization, acting the decision on the system vehicle, and controlling the operation of the system vehicle.

And S4, performing optimization control on the running tracks of the vehicles in the group driving on the road grouping optimization section at the next moment by repeatedly executing the steps S2 and S3, and executing in a circulating mode until all the vehicles drive out of the road grouping optimization section.

The main optimization means of the technical scheme of the embodiment is to execute the optimization control of the multiple intelligent agents (the vehicles can be optimally controlled by the intelligent networked vehicles) in groups. To ensure that the ramp vehicles can successfully start the merging action at the merging decision point (250 m in fig. 2 (a)), the collaborative optimization algorithm calculates all control sequences of all optimally controllable vehicles before leaving the optimization area. The upper layer model is a vehicle sequencing problem solved by dynamic programming recursion, and optimizes the convergence order of the vehicles. The lower layer model is used for solving the track optimization problem of each single vehicle by using a dynamic matrix prediction algorithm and optimizing the confluence track of each single vehicle. The lower model aims to optimize the track of a specific vehicle, calculate the corresponding system cost and return to the upper model. The stage variable k is determined by the number of ramp vehicles, namely, each ramp vehicle corresponds to a stage of an upper-layer problem. The rolling calculation process will continue until the last vehicle exits the optimization zone.

The next second of the optimized control sequence will be transferred to each of the optimally controllable vehicles and executed. After the decision is executed, the state of the vehicle is changed, and the execution of the double-layer optimization model is triggered to form a closed loop, so that a model prediction control mechanism is formed. The mechanism is suitable for various microcosmic mixed traffic flow scenes.

The mixed traffic flow cooperative optimization control method based on model predictive control of the embodiment is subjected to simulation analysis by constructing a microscopic traffic flow simulation environment. The simulation operation result shows that the control mechanism of the method can optimize the vehicle convergence sequence, solve the problem of randomness possibly generated by driving vehicles in the prior art, reduce vehicle conflicts in the mixed traffic flow environment, effectively improve the vehicle passing efficiency and the driving comfort, and can realize real-time control.