Method for selecting a limited or empty set of assumptions of a possible position of a vehicle
阅读说明:本技术 用于选择出车辆的可能位置的有限或空假设集合的方法 (Method for selecting a limited or empty set of assumptions of a possible position of a vehicle ) 是由 P·博尼法特 F·李 J·伊巴涅斯-古茨曼 于 2019-02-12 设计创作,主要内容包括:本发明涉及一种用于从多个假设中选择出车辆的可能位置的有限或空假设集合(Z1至Z8)的方法。本发明涉及:-借助于地理定位系统获取该车辆的至少一个地理定位位置(P<Sub>0</Sub>)的步骤,-获取该车辆的可能位置的多个假设的步骤,-确定该车辆的地理定位位置的协方差和每个所获取的假设的协方差的步骤,-针对每个所获取的假设,计算马氏距离的步骤,以及-从所获取的假设中选择出有限或空假设集合的步骤。(The invention relates to a method for selecting a limited or empty set of hypotheses (Z1-Z8) of possible positions of a vehicle from a plurality of hypotheses. The present invention relates to: -acquiring at least one geolocation position (P) of the vehicle by means of a geolocation system 0 ) -a step of acquiring a plurality of hypotheses of possible positions of the vehicle, -a step of determining the covariance of the geo-located position of the vehicle and the covariance of each acquired hypothesis, -a step of calculating the mahalanobis distance for each acquired hypothesis, and-a step of selecting from the acquired hypothesesA step of selecting a limited or empty set of assumptions.)
1. A method for selecting a limited or empty set of hypotheses for a possible position of a vehicle (10) from a plurality of hypotheses, characterized in that it comprises:
-acquiring a geo-located position (P) of the vehicle (10) by means of a geo-locating system0) In the step (2) of (a),
-a step of obtaining a plurality of hypotheses of possible positions of the vehicle (10),
-determining a geo-located position (P) of the vehicle (10)0) And the covariance of each acquired hypothesis,
-for each acquired hypothesis, according to the geolocation position (P) of the vehicle (10)0) And the assumed covariance to calculate the mahalanobis distance (D)Mj) A step of
-from each calculated mahalanobis distance (D)Mj) Selecting a limited or empty set of assumptions from the obtained assumptions.
2. Selection method according to the preceding claim, wherein at the end of the selection, the availability or unavailability of each selected hypothesis is determined based on the number of selected hypotheses.
3. Selection method according to one of the preceding claims, wherein, in the selection step, for each mahalanobis distance (D)Mj) And (5) carrying out chi-square test.
4. Selection method according to one of the preceding claims, wherein, before the step of obtaining each hypothesis, the following operations are performed:
-prepositioning the vehicle (10) on the map in its geo-located position (P)0) At the position of the air compressor, the air compressor is started,
-surrounding the geo-located position (P) on the map0) Distribution of particles (P)i) Each particle corresponding to a possible position of the vehicle (10),
-in particular by means of a particle (P) for each particlei) Assign weight (w)i) To the particles (P)i) Using particle filters, and
-particles (P) to be derived from the particle filteri) Combined into a limited number of hypotheses, each associated with a lane of travel memorized in the map.
5. Selection method according to the preceding claim, wherein in the selection step only hypotheses are selected: for this hypothesis, the method is based on the particles (P) constituting this hypothesisi) Weight (w) ofi) And the calculated indicator is greater than the determined threshold.
6. Selection method according to one of the preceding claims, wherein in the determination step, the geo-located position (P) of the vehicle (10) is calculated0) And the covariance matrix of each acquired hypothesis.
7. Selection method according to one of the preceding claims, wherein a step of transmitting an alarm is provided if the set of selected hypotheses is empty and/or if several hypotheses remain.
8. A vehicle (10) comprising:
-means for memorizing a map,
-a geographical positioning system, and
-a computer adapted to pre-position the vehicle (10) on the map,
characterized in that the computer is adapted to implement the selection method according to one of the preceding claims.
Technical Field
The present invention relates generally to the field of map construction.
More specifically, the present invention relates to a method for selecting a limited or empty set of hypotheses for a possible location of a vehicle from a plurality of hypotheses.
The invention also relates to a vehicle comprising:
-means for memorizing a map,
-a geographical positioning system, and
-a computer adapted to pre-position the vehicle on the map and to implement the above selection method.
Background
To ensure the safety of autonomous and partially automated vehicles, it is necessary to have a deep knowledge of the environment in which these vehicles are moving around.
In fact, the perception of the vehicle of its environment is done in two different ways, namely:
-a geographical positioning device by using a map and a vehicle, and
by using an external sensitivity sensor (camera, radar sensor or lidar sensor, etc.).
Companies that create maps are currently studying so-called "high-definition" maps that make it possible to obtain very detailed information about road network characteristics (lane width, ground markings, signal transmission plates, etc.).
These maps are embedded in vehicles equipped with a geolocation device, allowing them to be located at estimated locations in longitude and latitude on the map.
Unfortunately, it is found that this position is not always very accurate and reliable, which is then reflected by the vehicle being positioned outside the path actually taken. This problem has proven to be particularly dangerous in the case of autonomous vehicles that use this information to determine their direction.
In order to solve this problem, a first technical solution consists in: several possible positions of the vehicle are determined, given the road taken and the geographical location position, and the most likely position is selected.
The main disadvantage of this solution is that only one position can be selected in the end, so that in the event of an error, the motor vehicle cannot recognize the error, which proves to be very dangerous.
Another solution is also known from document EP 1729145, which consists in: the geolocation signals received from the satellites are processed to reduce errors associated with poor propagation of the signals to the vehicle.
The main drawback of this solution is that it uses external sensitivity sensors (gyroscopes, accelerometers, etc.) to determine the precise position of the vehicle, which proves to be costly and makes the reliability of the solution subject to the reliability of the sensors used.
Disclosure of Invention
In order to remedy the above-mentioned drawbacks of the prior art, the invention proposes to take into account the assumptions of the possible positions of the vehicle and to carry out a conformance test on these assumptions in order to minimize the number of assumptions and to reduce as far as possible to a single solution.
More specifically, according to the invention, a method for selecting hypotheses is proposed as defined in the introduction, in which method the following are provided:
-a step of acquiring at least one geolocation position of the vehicle by means of a geolocation system,
-a step of obtaining a plurality of hypotheses of possible positions of the vehicle,
-a step of determining the covariance of the geo-located position of the vehicle and the covariance of each acquired hypothesis,
-for each acquired hypothesis, a step of calculating the mahalanobis distance based on the covariance of the geo-located position of the vehicle and the covariance of said hypothesis, and
-a step of selecting a limited or empty set of hypotheses from the acquired hypotheses, based on each calculated mahalanobis distance.
This method therefore involves an arbitration method which makes it possible to check each hypothesis for correspondence with the vehicle's geo-located position, thus being able to indicate whether the chosen hypothesis is available.
The advantage of the claimed solution is that it does not evaluate the accuracy of the received geo-location data, but rather the consistency of the operation of the algorithm using these data. Thus, this solution does not require the use of additional sensors, making it cheap and very reliable.
More generally, the proposed solution makes it possible to drop down the available data and to judge whether this information is available in the context of the control of the autonomous vehicle by solving the problem of knowing whether the chosen hypothesis can be fully trusted with great reliability.
Another advantage of the proposed solution is that, since it is possible to indicate whether the data used in the algorithm are consistent, it is possible not only to determine whether the assumption is correct, but also to diagnose errors in the geolocation system. This advantage will appear more clearly on reading the rest of the present explanation.
Other advantageous and non-limiting characteristics of the process according to the invention are as follows:
-at the end of the selection, determining the availability or unavailability of each selected hypothesis based on the number of selected hypotheses;
-in the selecting step, performing a chi-squared test for each mahalanobis distance;
-before the step of acquiring each hypothesis, performing the following operations:
prepositioning the vehicle on the map at its geo-located location,
distributing particles on the map around the geo-located location, each particle corresponding to a possible location of the vehicle,
applying particle filters to the particles, in particular by assigning a weight to each particle, and
combining the particles (Pi) derived from the particle filter into a limited number of hypotheses, each hypothesis being associated with a lane of travel memorized in the map;
in this selection step, only hypotheses are selected: for the hypothesis, an index calculated from the weights of the particles that make up the hypothesis is greater than the determined threshold;
-in the determining step, calculating a covariance matrix of the geo-located position of the vehicle and a covariance matrix of each acquired hypothesis;
-providing a step of transmitting an alert if the set of selected hypotheses is empty and/or if several hypotheses remain.
The invention also relates to a vehicle comprising:
-means for memorizing a map,
-a geographical positioning system, and
-a computer adapted to pre-position the vehicle on the map and to implement the above selection method.
Drawings
The following description, given as a non-limiting example with reference to the accompanying drawings, will give a good understanding of what the invention consists of and how it may be produced.
In the drawings:
figure 1 is a diagram illustrating the different steps of the method according to the invention;
figure 2 is a plan view of a vehicle travelling on a road;
figure 3 is a schematic view of particles distributed on a map;
FIG. 4 is a schematic view of two particles positioned along two consecutive road sections, an
Fig. 5 is a schematic view of four particles positioned beside four road sections.
Detailed Description
In fig. 2, a
In the following of the description, more particular attention will be paid to the positioning of this
The
The computer is connected to so-called "proprioceptive" sensors that can accurately measure the speed of the vehicle and the yaw rate of the vehicle.
Preferably, the computer is also connected to "externality" sensors that can sense the direct environment of the motor vehicle 10 (these sensors can be cameras, radar sensors, lidar sensors, etc.).
The computer is also connected to a geolocation system that can evaluate the geolocation position P, here defined by latitude and longitude, of the
It will be considered here that the geolocation system is also suitable for transmitting to the computer data called "horizontal protection level HPL". This data, well known to the person skilled in the art, corresponds to the geo-located position P0The uncertainty of the measurement of. The value of which varies, for example, according to the number of satellites from which the geolocation system receives data, the quality of the signal reception, the quality of the geolocation system used, etc.
Along the same lines, it will also be considered here that the geolocation system is adapted to transmit to the computer a covariance matrix associated with this same uncertainty.
The
However, in the following of the present explanation, it will be considered that the
The computer then comprises a computer memory which stores data used in the context of autonomous control of the vehicle and records data used in the context of the method described below in particular.
The computer memory especially memorizes computer applications consisting of computer programs comprising instructions which are executed by a processor to allow the method described below to be implemented by a computer.
The computer memory also memorizes so-called "high definition" topographic maps.
The map stores a large amount of data.
Which first includes information about the topography of the road. Here, this topography is memorized in the form of road sections (or "links"). Here, each road section is defined as a portion of a single traffic lane of the road, which portion is characterized by being constant over its entire length (the shape of the ground markings along this road section is the same, the width of this road section is constant, etc.).
The map also stores other data characterizing each road section, including the width of the roadway, the shape of markers positioned on the ground on either side of the roadway, the position and shape of each plate at the road section that defines the road, the identifiers of the previous and subsequent road sections, and the like.
In order to estimate the precise position P of the
Assume that the
Thus, the first step will describe the first
This operation is performed recursively, that is to say in a cyclic manner and at fixed time steps.
This operation comprises three main steps.
The
Thus, the computer obtains the geo-location position P of the
The computer also obtains data relating to the dynamics of the
The
The
Each particle PiCan be defined by:
two coordinates x making it possible to define the position of the particle in a cartesian reference systemi、yi(these coordinates are linked to the acquired longitude and latitude),
a yaw angle such that the angle of the particles with respect to a given direction (such as north) can be defined, an
-particles PiAn identifier of the map section associated therewith.
In figures 3 to 5 particles P in the form of isosceles triangles are representediEach triangle having a center M corresponding to the position of the particle on the mapiAnd with particles on mapsYaw angle corresponds to orientation.
As shown in fig. 1, the third
The
This situation can then be adopted, in which case no particles have yet been generated.
The
For this purpose, particles P are introducediDistributed over geographically located locations P with
These particles are in particular distributed helically with a constant angular difference. Based on the particle P to be generatediThe number of the helix and the particle P are selectediThe angular difference between them.
This number is greater than 100 and preferably about 1000. In a manner determined to obtain sufficient accuracy without overloading the computer in any way.
At this stage, the particle P has not yet been alignediOrientation is performed.
Each particle P is considered to be in the presence of errors affecting the geolocation systemiCorresponding to possible locations the vehicle may have.
As can be seen in fig. 3, some particles are located outside the road. This demonstrates the following fact: the particles are not constrained on the map and they can move around in two dimensions. The filter is thus very flexible and makes it possible to consider a very large number of different solutions initially, the most unreasonable of which will then be eliminated by the particle filter.
During the
The method chosen here is a "point-to-curve" type of method. The method consists in making each particle PiAssociated with the closest road segment in the sense of euclidean distance.
As an illustrative example, it can thus be observed in fig. 4 that the particles P1Associated with the road section AB.
At this stage, the computer may cause the particle P toiIn particular based on the orientation of the road section with which each particle is associated (and possibly also based on the dynamic characteristics of the vehicle).
The method then proceeds to substep 116, which will be described later.
As mentioned above, the
Now it can be considered that this is not the case and that the procedure has been initiated previously.
In this case, during
To this end, the particles P are caused toiAll move on the map based on information about the dynamic characteristics of the vehicle.
The two data items, i.e. the speed V of the vehicle and the yaw rate Ψ of the
It should be noted that this sub-step does not use the geolocation position P of the motor vehicle this time0。
During the
More specifically, the computer determines which particles P areiA new road section must be associated and it identifies this new road section.
To understand how a computer works, reference may be made to figure 4,wherein two particles P centered around points M1, M2 are shown1、P2And also a road section AB.
It is considered here that the two particles P are present at a previous time step1、P2Are associated with the same road section AB and they are shifted during
The computer then targets each particle PiThe ratio r is determined in order to know whether each particle should be associated with a new road section.
This ratio r is calculated according to the following equation:
in the case where the ratio r is between 0 and 1, the particle P must not be alterediAssociation with its original road section. For the particle P1This is true here.
In the case where this ratio is negative, the particle P must be alterediAssociation with its road section. More specifically, this particle should be associated with one of the previous road section or the preceding road section.
In the case where this ratio is strictly greater than 1, the particle P must be modifiediAssociation with its road section. More specifically, this particle should be associated with one of the next road section or the following road section.
Thus, several situations may be encountered.
In case the road section AB in fig. 4 comprises only a single successor BB', the particle P2Associated with this successor (since the ratio r for this new road section is between 0 and 1, otherwise another successor is considered).
In case the road section AB in fig. 5 comprises a plurality of successors BC, BD, BE, the particle P considered at the previous time step2Cloned as many particles P as there are successors BC, BD, BE21、P22、P23。
Given the dynamic nature of the vehicle, the number of particle clones may be less if certain successors cannot be considered.
In another case, not represented in the figure, it may be necessary to associate the particle with another road section parallel to the road section with which it was associated in the previous time step (this would occur in particular when a certain vehicle laterally changes lanes, for example in order to pass another vehicle). This is possible because the particles are not constrained to move only on the same road segment. This can be detected given the new position of the particle and the data stored in the map (information about ground markers, lane width, etc.). In a variant, it is also conceivable to use a camera embedded in the vehicle to detect this.
During
Here, the likelihood of a particle is given by its weight wiTo indicate. The greater the weight of a particle, the more likely it is that the particle under consideration corresponds to the exact location of the
This weight may be calculated in different ways.
In the first embodiment, each particle P is calculated based on only data derived from a mapiWeight w ofi。
More specifically, the weight is determined based on the euclidean distance of the considered particle from the road section with which it is associated (this weight is for example inversely equal to this distance).
In a second embodiment, each particle P is also calculated based on the data derived from the external susceptibility sensor, provided that these data are considered reliableiWeight w ofi。
In practice, it is conceivable to increase or decrease the weight of the particles under consideration based on lateral information originating from the camera CAM of the vehicle. These cameras are actually able to detect the ground marker lines and return them to the computer in the form of a polynomial model. The computer may then check whether the shape of these lines corresponds to the shape of the ground markings stored in the map, and adjust the weight of the particles accordingly.
It may be noted that the ground mark is not always detected by the camera. This may be due to conditions to which the sensor is difficult to adapt, such as for example insufficient light, wet road, marks being deleted, etc. In these particular cases, the camera indicates a low confidence level to the computer, and then calculates the weights based only on the data supplied by the map, as described in the first embodiment.
Regardless of which method is used, the method proceeds to sub-step 117: selecting a limited number of particles PiTo eliminate the instantaneous geolocation position P from the
To perform this substep, the computer obtains a new geographic position P of the
If this distance is greater than the horizontal protection level HPL, the corresponding particle P will be presentiWeight w ofiSet to zero, which will allow this particle to be automatically eliminated thereafter.
Otherwise, the corresponding particle P is not modifiediWeight w ofi。
During the
For this purpose, the computer uses the index NeffThe index is based on the particle PiWeight w ofiAnd particles PiIs calculated from the number of the first and second electrodes.
If the index is NeffFalls below a predetermined threshold (which is stored in the computer's read-only memory), the computer detects the particles P on the mapiResampling is performed. Otherwise, let the particle PiRemaining in its state.
As is well known, resampling consists in: particles (hereinafter referred to as original particles) are considered in their set, and new particles are extracted from this original set.
To align particlesThe subsampling is done and the computer can use a conventional method during which the computer will extract the original particle P fromiRandomly extracting a predefined number of new particles from the set to each particle PiProbability of and the particle PiWeight w ofiIs in direct proportion. However, this approach typically results in particle depletion because very high weight particles are always drawn.
Preferably, in contrast, the computer uses a resampling method referred to herein as "low variance" (referred to as "low variance resampling). In fact, this approach is advantageous to maintain a good distribution of particles on the map. The method comprises the following steps: from the primary particle P randomlyiExtracting a predefined number of new particles from the set, to each particle PiHas a probability that the particle P isiWeight w ofiBut this time not in proportion to this weight.
At this stage, the computer may simply resume the execution of the sub-steps 114 to 118 cycle until particles are obtained that are all positioned around the same point that will be considered to correspond to the exact position P of the
However, this is not the option chosen here. Thus, as already explained previously, the
This
The
During a
To this end, the computer takes into account the particles P in the different setsiWithin each set, the particles are associated with the same traffic lane (or, as a variant, with the same road section).
The benefit of studying the hypotheses is that it will then be possible to select all the most likely hypotheses, so that it will be possible to retain good hypotheses from the selected hypotheses on the one hand and to verify the validity of each selected hypothesis on the other hand.
The assumption may be expressed in the form of an assertion, such as "the vehicle is on a lane of travel referenced … …".
In order to obtain a better understanding of what the hypotheses correspond to within the meaning of the present explanation, the particles are grouped together in fig. 3 into eight sets Z each corresponding to one hypothesis1、Z2、Z3、Z4、Z5、Z6、Z7、Z8。
As an example, set Z1Particle in (b) corresponds to the assumption that "the vehicle is located on the road R1On the right traffic lane ".
Set Z2Particle in (b) corresponds to the assumption that "the vehicle is located on the road R1Left lane of traffic ".
Set Z3Particle in (b) corresponds to the assumption that "the vehicle is located on the road R2Left lane of traffic ".
Set Z4Corresponds to the assumption that "the vehicle is located in its place with the road R1And R2At the ring intersection between the junctions of ".
Considering that a number "J" of hypotheses is found (in fig. 3, J is 8), each hypothesis may also be represented as a mean coordinate vectorIs expressed (the components of the vector correspond to the particle P in this hypothesisiAnd by the weight w of these particles) of the particleiAnd (4) weighting.
The computer may assign a "confidence index" to each hypothesis, the confidence index being equal to the particle P of that hypothesisiWeight w ofiAnd (4) summing.
During a
Indeed, operating such a covariance matrix may characterize the uncertainties associated with each hypothesis and with the geolocation position P supplied by the geolocation system0An associated uncertainty.
As explained above, the geo-location position P with the
With respect to the covariance matrix associated with each hypothesis
This is based on a set of particles P associated with this hypothesisiWeight w ofiIs calculated. The covariance matrix is also a 2 × 2 matrix, whose expression is as follows:
wherein the content of the first and second substances,
and is
Then, at a given geographic positioning position P supplied by the geographic positioning system0And to take into account the error associated with the measurement of this geo-located position, it is necessary to determine the degree of "agreement" for each hypothesis.
For this purpose, during
wherein, XGNSSCorresponding to a "geolocation position P0”。
In fact, mahalanobis distance is an object that can take into account the covariance of the variables (that is, the uncertainty associated with each variable) to evaluate the consistency between two uncertain cases.
Then, during
For this purpose, for each mahalanobis distance DMjProceed with chi fang (X)2) And (6) testing.
In practice, each mahalanobis distance D will be used hereMjIs compared with a critical threshold for a given risk of false detection to determine whether the hypothesis in question is related to the geolocation position P0And (5) the consistency is achieved.
If the assumption considered is related to the geographical positioning position P0This assumption is retained if it is consistent in the sense of chi-squared test.
Otherwise, if the assumption considered is associated with the geolocation position P0Disagreement in the sense of chi-squared tests, the hypothesis is rejected.
It will be noted here that if a certain assumption is retained, this does not necessarily mean that the assumption is true. In fact, at this stage, several assumptions may be retained.
On the other hand, if a certain hypothesis is rejected, this does not necessarily mean that this hypothesis is false. In fact, there may be a large error impact on the geolocation position P0The case of the measurement of (1). In this case, the assumption that is true may be rejected. As will be clearly apparent in the explanation below, this does not in any way affect the reliability of the method proposed here.
Then, during a
It will be noted here that this second selection can be applied before the first selection without affecting the progress of the method in any way.
This second option consists in: only the "possible" hypothesis is retained, for which the particles P constituting this hypothesisiWeight w ofiThe associated indicator is greater than the determined threshold. In fact, the goal is to eliminate the assumption that the chi-square conformance test has been met but is unlikely.
For this purpose, the computer eliminates the need for a confidence index (which, as will be appreciated, is equal to the particle P in the hypothesis under consideration)iWeight w ofiSum) below the determined threshold. Here, this threshold is unchanged and memorized in the read only memory of the computer.
At the end of these two hypothesis selection steps, the computer has retained a number N of not only consistent but also possible hypotheses.
During
Three cases can then be envisaged.
The first case is the case where the number N is equal to 1. In this case, since a single hypothesis is retained, the hypothesis is considered fair and can be used to generate the driving set points for the autonomous vehicle. If this assumption is followed, the computer may rely on the assumption. In this case, the computer may then consider the mean position of the particles in this hypothesis to correspond to the precise position P of the
The second case is the case where the number N is strictly greater than 1. In this case, since several assumptions remain, none of the assumptions is considered to be useful for generating the driving set point of the autonomous vehicle.
The last case is the case where the number N is equal to 0. In this case, as in the previous case, no position is considered available for generating the driving set point of the autonomous vehicle, since no assumptions remain. Furthermore, the computer may advantageously deduce from this that there is an inconsistency between the measurements performed by the geolocation system and the assumptions obtained, which may be due to problems affecting the geolocation system. In this possible case, a
- 上一篇:一种医用注射器针头装配设备
- 下一篇:将现有路线延伸至目的地的远程操作