阅读说明：本技术 自适应的gps错误观测值识别方法 (Self-adaptive GPS error observed value identification method ) 是由 彭国旗 黄友 张国龙 张放 李晓飞 张德兆 王肖 霍舒豪 于 2020-04-24 设计创作，主要内容包括：本发明提供了一种自适应的GPS错误观测值识别方法,包括：从GPS传感器中获取车辆的定位信息,从定位信息中提取第一观测值数据。获取车辆的姿态信息和速度信息,根据姿态信息和速度信息计算得到车辆的航位推算轨迹数据。根据第一观测值数据中的数据状态值数据、航向有效位数据、收星数数据及水平精度因子数据剔除错误的GPS观测值得到第二观测值数据。根据第二观测值数据构建位姿图数据,并计算得到处理结果信息,对处理结果信息分析优化,剔除不在预设的代价函数阈值内的错误的GPS观测值,得到第三观测值数据。根据第三观测值数据及三维场景地图数据构建高精度地图,实现高精度实时定位。由此,降低了GPS错误观测值带来的安全问题。(The invention provides a self-adaptive GPS error observed value identification method, which comprises the following steps: positioning information of the vehicle is acquired from the GPS sensor, and first observation value data is extracted from the positioning information. And acquiring attitude information and speed information of the vehicle, and calculating dead reckoning track data of the vehicle according to the attitude information and the speed information. And obtaining second observation value data according to the data state value data, the course effective bit data, the satellite number data and the GPS observation value with the horizontal precision factor data rejection error in the first observation value data. And constructing pose graph data according to the second observation value data, calculating to obtain processing result information, analyzing and optimizing the processing result information, and eliminating wrong GPS observation values which are not within a preset cost function threshold value to obtain third observation value data. And constructing a high-precision map according to the third observation value data and the three-dimensional scene map data, and realizing high-precision real-time positioning. Therefore, the safety problem caused by the GPS error observed value is reduced.)

1. An adaptive GPS error observation identification method, the method comprising:

acquiring continuous positioning information of the vehicle from a GPS sensor;

extracting first observation value data from the positioning information, wherein the first observation value data comprises data state value data, heading effective bit data, star data and horizontal precision factor data;

acquiring attitude information of the vehicle and speed information of the vehicle;

calculating according to the attitude information and the speed information to obtain dead reckoning track data of the vehicle;

when the data state value data of the first observation value data is equal to a preset data state expected value, the course effective bit data is effective, the satellite number data is greater than a preset satellite number threshold value, and the horizontal precision factor data is less than a preset horizontal precision factor threshold value, marking the grade of the first observation value data as a first grade;

when the data state value data of the first observed value data is equal to the data state expected value, the course valid bit data is invalid, the satellite number data is greater than the satellite number threshold value, and the horizontal precision factor data is less than the horizontal precision factor threshold value, marking the level of the first observed value data as a second level;

acquiring a level set of the first observed value data of a preset continuous frame number before the current first observed value data;

when all the levels of the first observed value data in the level set are the first levels, marking the current first observed value data as second observed value data;

when the levels of the first observed value data in the level set only comprise the first level and the second level, and the number of the first observed value data of the second level is smaller than a preset second level number threshold, marking the current first observed value data as the second observed value data;

setting a covariance matrix and a kernel function by taking a position sensor matching position of the vehicle as a vertex and adding one second observation value data to the vertex, wherein a first relative observation quantity of the position sensor at a first continuous moment and a second relative observation quantity of the dead reckoning trajectory data at a second continuous moment are taken as constraint edges, so as to construct pose graph data;

calculating according to the pose graph data to obtain processing result information;

judging whether the cost functions of all the second observation value data exceed a preset cost function threshold value or not according to the processing result information; when the number of the second observed value data exceeding the cost function threshold is smaller than a preset second number threshold, marking the second observed value data not larger than the cost function threshold as third observed value data;

and processing according to the third observation value data and the three-dimensional scene map data to obtain real-time positioning data.

2. The method of claim 1, wherein the rating of the first observation data is labeled a third rating when the data state value data of the first observation data is not equal to the data state expected value or the number of stars data is not greater than the number of stars threshold or the horizontal dilution of precision data is not less than the horizontal dilution of precision threshold.

3. The method of claim 1, wherein the attitude information is calculated by an inertial measurement unit of the vehicle;

the speed information is calculated from a wheel speed sensor of the vehicle.

4. The method according to claim 1, wherein after determining whether the cost functions of all the second observed value data exceed a preset cost function threshold value according to the processing result information, the method further comprises:

if the number of the second observation value data exceeding the cost function threshold is not smaller than the second number threshold, expanding the cost function threshold in a multiple mode to obtain a second cost function threshold;

judging whether the cost functions of all the second observed value data exceed the second cost function threshold value or not according to the processing result information; when the number of the second observed value data exceeding the second cost function threshold is less than the second number threshold, the second observed value data not greater than the second cost function threshold is labeled as third observed value data.

Technical Field

The invention relates to the technical field of vehicle real-time positioning, in particular to a self-adaptive GPS error observed value identification method.

Background

The automatic driving technology is a hot topic in recent years, and the automatic driving brings subversive changes in the fields of relieving traffic jam, improving road safety, reducing air pollution and the like. In the automatic driving commercialization process, unmanned cleaning vehicles, unmanned express delivery vehicles, unmanned taxies and the like in a limited area scene provide a specific application scene for landing of the automatic driving technology. Due to the aggravation of aging of population, the rising of labor cost in China and the increase of workload of people caused by heavy repetitive physical labor, such as environmental sanitation cleaning, express delivery, taxi trip and the like, the automatic driving technology is out of gear to replace repetitive work.

The unmanned vehicle can realize automatic driving, can not leave a high-precision map and a high-robustness positioning method. In an outdoor unmanned application scene, a gps (global positioning system) is used as an input sensor for data, regardless of three-dimensional scene reconstruction or vehicle real-time positioning. If the error observation value input by the GPS is not correctly identified and rejected, the accuracy of the constructed three-dimensional scene map is reduced, and the unremoved error value causes the positioning result to be seriously deviated from the real position, so that the safety and reliability of automatic driving of the vehicle are influenced.

At present, the identification methods of GPS error observation values which are applied more can be roughly divided into two types, namely an identification method based on the GPS observation value and an identification method based on the fusion of the GPS observation value and an accelerometer.

The identification method based on the GPS observation value mainly comprises a Kalman filtering method and a method combining classical probability and Leidete discrimination. First, a data processing model, such as a prediction model of kalman filtering, an error probability distribution model, or the like, is acquired from sample data. Then, selecting a reasonable processing model and a parameter threshold value, analyzing the real-time observed GPS data, and finally judging whether the GPS data is an error value according to an analysis result. The identification method based on the GPS observation value needs a pre-planned movement track and sample data to obtain a corresponding model, the GPS observation value has certain difference due to different working time, and the vehicle can not be completely guaranteed to travel according to a preset track in the movement process, so that the error of the GPS observation value is difficult to be accurately analyzed by using a processing model.

The method comprises the steps of acquiring speed and position variation of an accelerometer and a GPS observation value in preset time, obtaining a predicted value of a GPS according to the data variation of the accelerometer, judging whether absolute deviation between the GPS observation value and the predicted value is smaller than a certain specific value or not, and if not, determining that the GPS observation value at the moment is an error value. The identification method based on the fusion of the GPS observation value and the accelerometer can well identify the error value of GPS jump, but in the automatic driving of the vehicle, the GPS is influenced by the multipath effect, the GPS observation value can generate gradual change errors and step change errors, and the error of the observation value at the moment can not be identified by the recursion of the accelerometer in the preset time and the GPS fusion.

Disclosure of Invention

The invention aims to provide a self-adaptive GPS error observation value identification method aiming at the defects in the prior art.

In order to achieve the above object, the present invention provides an adaptive GPS error observation value identification method, including:

acquiring continuous positioning information of the vehicle from a GPS sensor;

extracting first observation value data from the positioning information, wherein the first observation value data comprises data state value data, heading effective bit data, star data and horizontal precision factor data;

acquiring attitude information of the vehicle and speed information of the vehicle;

calculating according to the attitude information and the speed information to obtain dead reckoning track data of the vehicle;

when the data state value data of the first observation value data is equal to a preset data state expected value, the course effective bit data is effective, the satellite number data is greater than a preset satellite number threshold value, and the horizontal precision factor data is less than a preset horizontal precision factor threshold value, marking the grade of the first observation value data as a first grade;

when the data state value data of the first observed value data is equal to the data state expected value, the course valid bit data is invalid, the satellite number data is greater than the satellite number threshold value, and the horizontal precision factor data is less than the horizontal precision factor threshold value, marking the level of the first observed value data as a second level;

acquiring a level set of the first observed value data of a preset continuous frame number before the current first observed value data;

when all the levels of the first observed value data in the level set are the first levels, marking the current first observed value data as second observed value data;

when the levels of the first observed value data in the level set only comprise the first level and the second level, and the number of the first observed value data of the second level is smaller than a preset second level number threshold, marking the current first observed value data as the second observed value data;

setting a covariance matrix and a kernel function by taking a position sensor matching position of the vehicle as a vertex and adding one second observation value data to the vertex, wherein a first relative observation quantity of the position sensor at a first continuous moment and a second relative observation quantity of the dead reckoning trajectory data at a second continuous moment are taken as constraint edges, so as to construct pose graph data;

calculating according to the pose graph data to obtain processing result information;

judging whether the cost functions of all the second observation value data exceed a preset cost function threshold value or not according to the processing result information; when the number of the second observed value data exceeding the cost function threshold is smaller than a preset second number threshold, marking the second observed value data not larger than the cost function threshold as third observed value data;

and processing according to the third observation value data and the three-dimensional scene map data to obtain real-time positioning data.

Further, when the data state value data of the first observed value data is not equal to the data state expected value or the star count data is not greater than the star count threshold or the horizontal precision factor data is not less than the horizontal precision factor threshold, the rank of the first observed value data is marked as a third rank.

Further, the attitude information is calculated by an inertial measurement unit of the vehicle;

the speed information is calculated from a wheel speed sensor of the vehicle.

Further, after determining whether the cost functions of all the second observed value data exceed a preset cost function threshold according to the processing result information, the method further includes:

if the number of the second observation value data exceeding the cost function threshold is not smaller than the second number threshold, expanding the cost function threshold in a multiple mode to obtain a second cost function threshold;

judging whether the cost functions of all the second observed value data exceed the second cost function threshold value or not according to the processing result information; when the number of the second observed value data exceeding the second cost function threshold is less than the second number threshold, the second observed value data not greater than the second cost function threshold is labeled as third observed value data.

By applying the self-adaptive GPS error observation value identification method provided by the invention, firstly, the error observation value is quickly identified and eliminated through the related marker bits based on the differential GPS data, and then secondary identification based on the problem of the position and orientation map is carried out on the rest observation values, so that the error observation value is further eliminated, the rest correct GPS observation value is output to the three-dimensional scene map construction, the high-precision map construction is completed, and the high-precision real-time positioning with high robustness and safety is ensured. Therefore, the probability of safety problems caused by the GPS error observation value is reduced, and the three-dimensional map reconstruction precision and the vehicle real-time positioning precision in the complex outdoor environment are improved.

Drawings

Fig. 1 is a schematic flow chart of a method for identifying an adaptive GPS error observation value according to an embodiment of the present invention.

Detailed Description

The technical solution of the present invention is further described in detail by the accompanying drawings and embodiments.

The self-adaptive GPS error observation value identification method provided by the embodiment of the invention is applied to a vehicle-mounted server of an intelligent vehicle, wherein the intelligent vehicle can be understood as an unmanned automatic driving vehicle, the vehicle-mounted server is equivalent to the brain of the automatic driving vehicle, and can acquire and process data of a plurality of sensors on the vehicle, so that the state information of the vehicle and the environmental information around the vehicle are obtained, an environmental map is constructed, the vehicle is positioned in real time, and the automatic driving of the vehicle is realized.

The first, second, etc. numbering is used only for distinguishing and has no other meaning.

Fig. 1 is a schematic flow chart of a method for identifying a GPS erroneous observation value in an adaptive manner according to an embodiment of the present invention, where the method is applied to a vehicle-mounted server of an intelligent vehicle, and an application scenario of the method is a scenario in which an autonomous vehicle processes GPS data and identifies a wrong GPS observation value when the autonomous vehicle is autonomously driven in an outdoor environment. As shown in fig. 1, the main execution subject of the method is an on-board server of an intelligent vehicle, and the adaptive GPS error observation value identification method includes:

step 101, acquiring continuous positioning information of a vehicle from a GPS sensor; first observed value data is extracted from the positioning information.

The autonomous vehicle travels in an outdoor environment and the vehicle's onboard server obtains the vehicle's continuous positioning information from the vehicle's GPS sensor. Extracting the positioning information each time to obtain first observation value data required by the identification method, wherein the first observation value data comprises: data state value data, course effective bit data, star number data and horizontal precision factor data.

And 102, acquiring attitude information and speed information of the vehicle, and calculating according to the attitude information and the speed information to obtain dead reckoning track data of the vehicle.

The vehicle-mounted server acquires attitude information of a vehicle from an IMU (Inertial Measurement Unit) sensor of the vehicle, acquires speed information of the vehicle from a wheel speed sensor of the vehicle, and calculates according to the attitude information and the speed information to obtain dead reckoning track data of the vehicle.

103, when the data state value data of the first observation value data is equal to a preset data state expected value, the course effective bit data is effective, the star number data is greater than a preset star number threshold value, and the horizontal precision factor data is less than a preset horizontal precision factor threshold value, marking the grade of the first observation value data as a first grade; and when the data state value data of the first observation value data is equal to a preset data state expected value, the course effective bit data is invalid, the star number data is greater than a preset star number threshold value, and the horizontal precision factor data is less than a preset horizontal precision factor threshold value, marking the level of the first observation value data as a second level.

And judging and processing the data state value data, the heading effective bit data, the satellite number data and the horizontal precision factor data of the first observation value data acquired in the step 101. The judgment processing basis is a preset reasonable data state expected value, a star collection number threshold value and a horizontal precision factor threshold value.

And when the data state value data in the first observation value data is equal to a preset data state expected value, the course effective bit data is effective, the star number data is greater than a preset star number threshold value, and the horizontal precision factor data is less than a preset horizontal precision factor threshold value, marking the grade of the first observation value data as a first grade.

And when the data state value data in the first observation value data is equal to a preset data state expected value, the course valid bit is invalid, the star number data is greater than a preset star number threshold value, and the horizontal precision factor data is less than a preset horizontal precision factor threshold value, marking the grade of the first observation value data as a second grade.

Meanwhile, the remaining levels of the first observed value data that do not belong to either the first level or the second level are labeled as third levels.

And 104, acquiring a level set of the first observed value data of preset continuous frames before the current first observed value data.

A continuous number of frames L is set, where 3< L <20, and a set of levels of first observed value data for the continuous L frames prior to the current first observed value data is acquired.

Step 105, when all the levels of the first observed value data in the level set are first levels, marking the current first observed value data as second observed value data; and when the levels of the first observed value data in the level set only comprise the first level and the second level, and the number of the first observed value data of the second level is less than a preset second level number threshold, marking the current first observed value data as second observed value data.

Judging according to the level set of the first observed value data of the continuous L frames acquired in the step 104:

1. and when all the levels of the first observed value data in the level set are the first level, the current first observed value data is considered to be in a good state.

2. And when the levels of the first observed value data in the level set only comprise the first level and the second level, and the number of the first observed value data of the second level is less than a preset second level number threshold, the current state of the first observed value data is considered to be general.

And regarding the current first observed value data with the two conditions not satisfied, the state of the first observed value data is considered to be poor. And eliminating the first observed value data with poor state, only keeping the first observed value data with good state and common state, and marking the first observed value data into second observed value data for being used as the input quantity of the next secondary recognition.

And 106, setting a covariance matrix and a kernel function by taking the matched position of the position sensor of the vehicle as a vertex and adding a second observation value data at the vertex, and taking the first relative observation quantity of the position sensor at the first continuous moment and the second relative observation quantity of the dead reckoning trajectory data at the second continuous moment as constraint edges, thereby constructing the pose graph data.

The build Pose Graph problem identifies the second observed value data retained after the identification in step 105 a second time.

The matching position of the position sensor of the vehicle is used as a vertex point, wherein the position sensor of the vehicle can be a laser sensor of the vehicle or a camera sensor of the vehicle. In order to ensure smoothness and no distortion of the local trajectory shape, one second observed value data is added for each vertex. And setting a corresponding covariance matrix and a kernel function according to the attributes of various sensors by taking the first relative observed quantity of the first continuous time position sensor and the second relative observed quantity of the second continuous time dead reckoning track data as constraint edges, thereby constructing the pose graph data.

Step 107, calculating according to the pose graph data to obtain processing result information; judging whether the cost functions of all the second observation value data exceed a preset cost function threshold value or not according to the processing result information; and when the number of the second observed value data exceeding the cost function threshold is smaller than a preset second number threshold, marking the second observed value data not larger than the cost function threshold as third observed value data.

And calculating the pose graph data constructed in the step 106 to obtain processing result information. And analyzing and optimizing the processing result information, and counting whether the cost functions of all the second observation value data input in the step 106 exceed a preset cost function threshold value.

In one example, when the number of the second observed value data exceeding the cost function threshold is smaller than a preset second number threshold, the third observed value data not larger than the cost function threshold is labeled as the third observed value data. And the second observed value data larger than the cost function threshold value is an error GPS observed value and should be removed.

In another example, when the number of the second observed value data exceeding the cost function threshold is not less than the preset second number threshold, the cost function threshold needs to be expanded in a multiple manner to obtain the second cost function threshold, so that the method can automatically adapt to a more complex outdoor environment. And continuously analyzing and optimizing the processing result information by using the second cost function, and counting whether the cost functions of all the second observation value data input in the step 106 exceed a second cost function threshold value. When the number of the second observed value data exceeding the second cost function threshold value is smaller than a preset second number threshold value, the second observed value data not larger than the second cost function threshold value is marked as a third observed value. And the second observation value data larger than the second cost function threshold value is an error GPS observation value and is to be eliminated.

And step 108, processing according to the third observation value data and the three-dimensional scene map data to obtain real-time positioning data.

The third observed value data generated in step 107 is a correct GPS observed value from which an erroneous observed value is eliminated by performing secondary recognition, and the third observed value data is output to the three-dimensional scene map, and high-precision map construction is completed in combination with the three-dimensional scene map data, thereby realizing high-precision real-time positioning of the vehicle.

By applying the self-adaptive GPS error observation value identification method provided by the embodiment of the invention, firstly, the error observation value is quickly identified and eliminated through the related marker bit based on the differential GPS data, and then secondary identification based on the problem of the position and posture diagram is carried out on the rest observation values, so that the error observation value is further eliminated, the rest correct GPS observation value is output to the three-dimensional scene map construction, the high-precision map construction is completed, and the high-precision real-time positioning with high robustness and safety is ensured. Therefore, the probability of safety problems caused by the GPS error observation value is reduced, and the three-dimensional map reconstruction precision and the vehicle real-time positioning precision in the complex outdoor environment are improved.

Those of skill would further appreciate that the various illustrative components and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or combinations of both, and that the various illustrative components and steps have been described above generally in terms of their functionality in order to clearly illustrate this interchangeability of hardware and software. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention.

The steps of a method or algorithm described in connection with the embodiments disclosed herein may be embodied in hardware, a software module executed by a processor, or a combination of the two. A software module may reside in Random Access Memory (RAM), memory, Read Only Memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

The above embodiments are provided to further explain the objects, technical solutions and advantages of the present invention in detail, it should be understood that the above embodiments are merely exemplary embodiments of the present invention and are not intended to limit the scope of the present invention, and any modifications, equivalents, improvements and the like made within the spirit and principle of the present invention should be included in the scope of the present invention.