CN114815829A - Motion Trajectory Prediction Method of Intersection - Google Patents

Abstract

The method for predicting the motion trail of the intersection comprises the following steps: the road sensor analysis and calculation system acquires initial states of a moving target and a non-moving target of the intersection and transmits a processing result to the vehicle management system, the vehicle management system selects a dangerous target according to a drawn running track, then fits the dangerous target, monitors the fitted dangerous target in a key way, changes the running track of the vehicle according to the real-time moving state of the dangerous target and selects the most feasible running track of the vehicle according to the target state of the dangerous target, and the back propagation neural network training and learning module divides the predicted data and the actual running data into a training set and a test set according to the proportion so as to further improve the accuracy of the driving control data; the data processing capacity is subjected to three-level screening and reduction of road sensors, dangerous target selection and fitting processing, the data processing capacity of a vehicle-mounted system is reduced, and the effective data processing efficiency is improved.

Description

Motion Trajectory Prediction Method of Intersection

technical field

The invention relates to the technical field of intelligent driving, in particular to a method for predicting a motion trajectory of an intersection.

Background technique

The intelligent driving system usually includes modules such as perception, positioning, and regulation, among which modules such as regulation draw up the vehicle's running trajectory according to the transmission information of perception and positioning. Intelligent driving only involves vehicle-to-vehicle driving interference on ordinary roads, but at intersections, it involves a variety of detection environments, such as pedestrians, non-pedestrians (living things), moving objects, road facilities (divided into moving objects, non-moving objects) and so on. The control of the running trajectory of the intersection is a key step to realize the high safety of intelligent driving. The safety performance lies not only in the accurate calculation, but also in the fast calculation speed, which is convenient for the car to respond sensitively according to the real-time changing data.

The present invention proposes a method for predicting the motion trajectory of an intersection to solve the above-mentioned technical problems.

SUMMARY OF THE INVENTION

The motion trajectory prediction method of the intersection, the steps are as follows:

The road sensor analysis and calculation system uses the intersection to obtain the initial state of the moving target and the non-moving target;

The vehicle driving management system predicts the running trajectory according to the initial state of the moving target and the driving state of the vehicle itself, and determines the moving target whose running trajectory has an intersection or is within a dangerous distance as a dangerous target; (in the same intersection in the same time period The initial state of the internal moving target and the non-moving target is the same, and there is no need to conduct real-time monitoring for each vehicle passing through this place. The sensor analysis and calculation system set at the intersection performs the same detection and preliminary calculation, and then transmits it to the distance intersection at a specific location. Vehicles within the range can be used, avoiding the waste of resources caused by simultaneous monitoring and calculation of multiple measurements, and reserving available space for subsequent processing by the measurement and driving management system alone, improving computing efficiency)

According to the initial state of the dangerous target combined with the target state of the vehicle itself, one or more vehicle driving trajectories are obtained by planning.

Preferably, in the method for predicting the motion trajectory of the intersection, the road sensor analysis and calculation system determines and distinguishes non-moving objects such as pedestrians, non-pedestrians and living bodies, non-living moving objects, road facilities, etc. , The initial state (position, moving speed, moving acceleration, etc.) of non-pedestrians and living, non-living moving objects, road facilities and other non-moving objects is transmitted to the vehicle driving management system. Among them, non-moving objects such as road facilities are fixed, such as road signs, road edges, flower beds, etc., so there is no need for real-time monitoring of non-moving objects such as road facilities, and information can be checked regularly or/and in the case of low traffic in the road section. , update to avoid temporary displacement or damage to road facilities.

Preferably, in the method for predicting the motion trajectory of the intersection, the vehicle driving management system performs fitting processing on the dangerous target according to parameters such as position, moving speed, state maintenance time, etc., that is, the moving speed difference is within the set range and the interval distance is set Two or more pedestrians, non-pedestrians, and living or non-living moving objects within a certain range and maintaining a state that meets the lower limit of the set time are fitted to the same target whose volume contours are accumulated. Select the representative object of the state and multiple state parameters. From the same object or objects in the fit processing object. In this way, the amount of data processing will be greatly reduced in the later motion trajectory prediction process, and the common states of pedestrians pulling pets, pedestrians pushing or riding means of travel at intersections are fully considered, and multiple objects to be fitted are correlated with each other.

Preferably, in the method for predicting the movement trajectory of the intersection, the vehicle driving management system predicts the running trajectory of the dangerous target according to the initial state of the dangerous target and the position information of the non-moving target, and then compares the running trajectory of the dangerous target with the driving movement of the vehicle. The intersection point of the trajectory is set as the movement end point of the dangerous target, and the movement end point state of the dangerous target corresponding to the movement end point with the maximum probability is calculated. Track and select a feasible (considering surrounding vehicles, safety factor, etc.) a vehicle driving trajectory.

Preferably, in the method for predicting the motion trajectory of the intersection, when the vehicle driving management system selects the most feasible vehicle driving trajectory, the movement end position of the dangerous target with the maximum probability is a circle, and the preset R is The radius is a circle, and the smallest area formed by intersecting the circle is predetermined as a feasible driving trajectory of the vehicle, and the R is related to the moving speed, the moving direction, and the moving acceleration.

Preferably, in the method for predicting the motion trajectory of the intersection, the vehicle driving management system includes a back-propagation neural network training and learning module; the back-propagation neural network training and learning module is based on the initial state of the dangerous target and the predicted maximum probability The target state and the real target state are divided into two groups of data in the ratio of 6:4 to 8:2 into training set data and test set data, which are used to train the reverse neural network module.

Preferably, the method for predicting the motion trajectory of the intersection, the back-propagation neural network training and learning module is divided into two groups according to the preset radius R, the motion state when the vehicle is located directly in front of the dangerous target, and the distance from the dangerous target The data is divided into training set data and test set data in the ratio of 6:4 to 8:2 for training the reverse neural network module.

Preferably, in the method for predicting the motion trajectory of the intersection, the initial state parameters of the moving target include one or more of speed, acceleration, position, and angular velocity.

Preferably, in the method for predicting the motion trajectory of the intersection, the target state of the dangerous target includes one or more of speed, acceleration, position, and angular velocity.

A motion trajectory prediction device operating according to the motion trajectory prediction method of an intersection according to any one of claims 1 to 9.

In the method for predicting the motion trajectory of the intersection involved in the patent of the present invention, the road sensor analysis and calculation system is used to detect the moving target of the intersection in real time, and the information of the non-moving target is transmitted to the vehicle driving management system together with the information of the non-moving target. After the received information determines the dangerous target number, the dangerous target is fitted, and then pay close attention to the state of the fitted dangerous target and make one or more vehicle running trajectories accordingly, and then predict the target with the highest probability of the dangerous target. The state (not the actual target state, only the target state defined when safe driving is considered during the operation of the vehicle) is combined with the surrounding environment to select the most feasible vehicle running trajectory.

Furthermore, according to the reverse neural network learning system, by dividing the historical data into training set data and test set data in proportion, the subsequent data processing process in the vehicle driving management system can be corrected, so that the data processing of the vehicle driving management system can be improved. Resilience is improved.

The data processing volume of the above-mentioned vehicle driving management system is screened and reduced in three stages, which greatly reduces the data processing volume of the system and improves the effective processing efficiency of data.

Description of drawings

The specific embodiments are further described below in conjunction with the accompanying drawings, wherein:

Fig. 1 is the connection schematic diagram of the motion trajectory prediction method operating system of the intersection involved in the present invention;

2 is a schematic flowchart of a method for predicting a motion trajectory of an intersection according to the present invention;

The following specific embodiments will further illustrate the present invention with reference to the above drawings.

Detailed ways

Hereinafter, exemplary embodiments according to the present disclosure will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present disclosure, not all of the embodiments of the present disclosure, and it should be understood that the present disclosure is not limited by the example embodiments described herein.

It should be noted that the relative arrangement of the components and steps, the numerical expressions and numerical values set forth in these embodiments do not limit the scope of the present disclosure unless specifically stated otherwise.

Those skilled in the art can understand that terms such as "first" and "second" in the embodiments of the present disclosure are only used to distinguish different steps, devices, or modules, etc., and neither represent any specific technical meaning, nor represent any difference between them. the necessary logical order of . It should also be understood that, in the embodiments of the present disclosure, "a plurality" may refer to two or more, and "at least one" may refer to one, two or more. It should also be understood that any component, data or structure mentioned in the embodiments of the present disclosure can generally be understood as one or more in the case of no explicit definition or contrary indications given in the context.

In addition, the term "and/or" in the present disclosure is only an association relationship to describe associated objects, indicating that there can be three kinds of relationships, for example, A and/or B, it can mean that A exists alone, and A and B exist at the same time , there are three cases of B alone. In addition, the character "/" in the present disclosure generally indicates that the related objects are an "or" relationship. It should also be understood that the description of the various embodiments in the present disclosure emphasizes the differences between the various embodiments, and the same or similar points can be referred to each other, and for the sake of brevity, they will not be repeated.

As shown in FIG. 1 , the overall system of the method for predicting the motion trajectory of the intersection includes: a first sensor module, a first information receiving module, a non-moving target information updating module, a moving target information processing module, a first information sending module, a first 2. Information receiving module, running trajectory prediction module, dangerous target judging module, dangerous target fitting module, second sensor module, driving control module, inverse neural network training and learning module, the first sensor module and first information receiving module are connected , the non-moving target information updating module and the moving target information processing module are simultaneously connected to the first information receiving module and the first information sending module, and the first information sending module is connected to the second information sending module in the vehicle driving management system. The information receiving module is connected, and the second information receiving module is respectively connected with the running trajectory prediction module and the dangerous target judgment module, the dangerous target judgment module, the dangerous target fitting module, the second sensor module, and the anti-neural network training and learning The modules are connected in sequence, the running trajectory prediction module is respectively connected with the second information receiving module and the dangerous target judging module, the driving control module is respectively connected with the running trajectory prediction module and the inverse neural network training and learning module, the The inverse neural network training and learning module is connected with the running trajectory prediction module.

The motion trajectory prediction method of the intersection, the steps are as follows:

The road sensor analysis and calculation system uses the intersection to obtain the initial state of the moving target and the non-moving target;

The vehicle driving management system predicts the running trajectory according to the initial state of the moving target and the driving state of the vehicle itself, and determines the moving target whose running trajectory has an intersection or is within a dangerous distance as a dangerous target; (in the same intersection in the same time period The initial state of the internal moving target and the non-moving target is the same, and there is no need to conduct real-time monitoring for each vehicle passing through this place. The sensor analysis and calculation system set at the intersection performs the same detection and preliminary calculation, and then transmits it to the distance intersection at a specific location. Vehicles within the range can be used, avoiding the waste of resources caused by simultaneous monitoring and calculation of multiple measurements, and reserving available space for subsequent processing by the measurement and driving management system alone, improving computing efficiency)

According to the initial state of the dangerous target combined with the target state of the vehicle itself, one or more vehicle driving trajectories are obtained by planning.

The specific implementation is as follows:

The road sensor analysis and calculation system determines and distinguishes pedestrians, non-pedestrians and living bodies, non-living moving objects, road facilities and other non-moving objects (moving target information processing module) through the sensing information of each sensor (first sensor module), and divides pedestrians. , The initial state (position, moving speed, moving acceleration, etc.) of non-pedestrians and living, non-living moving objects, road facilities and other non-moving objects is transmitted to the vehicle driving management system (first information sending module). Among them, non-moving objects such as road facilities are fixed, such as road signs, road edges, flower beds, etc., so there is no need for real-time monitoring of non-moving objects such as road facilities, and information can be checked regularly or/and in the case of low traffic in the road section. , update (non-moving target information update module) to avoid temporary displacement or damage to road facilities.

After the vehicle driving management system receives the preliminary calculation data results of the moving target and the non-moving target, the running trajectory prediction module draws the running trajectory of the moving target and the dangerous target, and according to the running trajectory, the dangerous target judgment module judges the dangerous target.

The vehicle driving management system is performing fitting processing on the dangerous target according to parameters such as position, moving speed, and state maintenance time (dangerous target fitting module), that is, the moving speed difference is within the set range, and the interval distance is within the set range. Two or more pedestrians, non-pedestrians, and living or non-living moving objects whose state maintains the lower limit of the set time are fitted to the same target whose volume contours are accumulated. Select the representative object of the state and multiple state parameters from the fitting Process the same object or objects in the object.

For example, when an adult moves, he also leads a dog and pushes a car. The dog's trajectory is more variable, but the itinerary is still within a certain range of the old man's position. The distance between the car and the adult remains unchanged and Chezi's moving speed Same as the adult, at this time, the three are fitted as a dangerous target, the overall volume of the dangerous target is the accumulation of the spatial contours (including the contour of the spatial distance) of the cheetah, the dog, and the adult, and the speed of the adult can be adopted.

In this way, the amount of data processing will be greatly reduced in the later motion trajectory prediction process, and the common states of pedestrians pulling pets, pedestrians pushing or riding means of travel at intersections are fully considered, and multiple objects to be fitted are correlated with each other.

Further, after the vehicle driving management system predicts the running trajectory of the dangerous target according to the initial state of the dangerous target and the position information of the non-moving target, the intersection of the running trajectory of the dangerous target and the vehicle driving running trajectory is set as the intersection of the dangerous target. The movement end point is calculated, and the movement end point state with the maximum probability of the dangerous target corresponding to the movement end point is calculated, and the multiple vehicle driving running trajectories obtained by the vehicle driving management system are evaluated according to the movement state of the maximum probability and the feasible ones are selected (considering the surrounding vehicles , safety factor, etc.) a vehicle driving trajectory.

Optionally, when the vehicle driving management system selects the most feasible vehicle driving trajectory, the movement end position of the dangerous target with the maximum probability is used as a circle, and the preset R is the radius to make a circle, which will intersect with the circle. The formed area with the smallest area is predetermined as a feasible driving trajectory of the vehicle, and the R is related to the moving speed, the moving direction, and the moving acceleration. The more stable the moving acceleration and the smaller the change in the moving direction, the smaller the R is.

On the basis of the above-mentioned method for predicting the motion trajectory of the intersection, the following further optimization methods can also be adopted:

In the method for predicting the motion trajectory of the intersection, the vehicle driving management system includes a back-propagation neural network training and learning module; the back-propagation neural network training and learning module is based on the initial state of the dangerous target, the predicted maximum probability target state, The real target state is divided into two groups of data in a ratio of 6:4 to 8:2 into training set data and test set data, which are used to train the reverse neural network module.

Further, the back-propagation neural network training and learning module is divided into two groups of data with a ratio of 6:4 to 8:2 according to the preset radius R, the motion state when the vehicle is located directly in front of the dangerous target, and the distance from the dangerous target. The ratio is divided into training set data and test set data for training the reverse neural network module.

To sum up, the initial state parameters of the moving target include one or more of speed, acceleration, position, and angular velocity. The target state of the dangerous target includes one or more of velocity, acceleration, position, and angular velocity.

The above-mentioned embodiments only represent several embodiments of the present invention, and the descriptions thereof are specific and detailed, but should not be construed as a limitation on the scope of the patent of the present invention. It should be pointed out that for those of ordinary skill in the art, without departing from the concept of the present invention, several modifications and improvements can also be made, which all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention should be subject to the appended claims.

Claims (10)

1. The method for predicting the motion trail of the intersection is characterized by comprising the following steps: the steps are as follows,

the road sensor analysis and calculation system uses the intersection to obtain the initial states of a moving target and a non-moving target;

the vehicle driving management system predicts the running track according to the initial state of the moving target and the driving state of the vehicle, and determines the moving target with an intersection point or in a dangerous distance range as a dangerous target;

and planning to obtain one or more vehicle driving running tracks according to the initial state of the dangerous target and the target state of the vehicle.

2. The intersection motion trajectory prediction method according to claim 1, characterized in that: the road sensor analysis and calculation system determines and distinguishes non-moving objects such as pedestrians, non-pedestrians and living bodies, non-living body moving objects, road facilities and the like according to the sensing information of the sensors, and transmits the initial state of the non-moving objects such as the pedestrians, the non-pedestrians and the living bodies, the non-living body moving objects, the road facilities and the like to the vehicle driving management system.

3. The intersection motion trajectory prediction method according to claim 2, characterized in that: the vehicle driving management system carries out fitting processing on the dangerous target according to parameters such as position, moving speed, state maintaining time and the like, namely two or more pedestrian, non-pedestrian and living or non-living moving objects with moving speed difference within a set range, interval distance within the set range and state maintaining meeting a set time lower limit are fitted into the same target with accumulated volume profiles, and a plurality of state parameters are selected from the same or a plurality of objects in fitting processing objects.

4. The intersection motion trajectory prediction method according to claim 3, characterized in that: the vehicle driving management system predicts the running track of the dangerous target according to the initial state of the dangerous target and the position information of the non-moving target, sets the intersection point of the running track of the dangerous target and the vehicle driving running track as the movement terminal point of the dangerous target, calculates the movement terminal point state of the maximum probability of the dangerous target corresponding to the movement terminal point, evaluates a plurality of vehicle driving running tracks obtained by the vehicle driving management system according to the movement state of the maximum probability, and selects the most feasible vehicle driving running track.

5. The intersection motion trajectory prediction method according to claim 4, characterized in that: when the vehicle driving management system selects the most feasible vehicle driving operation track, the position of the motion end point with the maximum probability of the dangerous target is taken as a circle, a preset R is taken as a radius to make a circle, and the preset vehicle driving operation track with the smallest area formed by intersection of the circle is taken as the most feasible vehicle driving operation track, wherein the R is related to the moving speed, the moving direction and the moving acceleration.

6. The intersection motion trajectory prediction method according to claim 4, characterized in that: the vehicle driving management system comprises a back propagation neural network training learning module; the back propagation neural network training learning module is divided into training set data and test set data according to the proportion of 6: 4-8: 2 according to the initial state of the dangerous target, the predicted maximum probability target state and the real target state, and the training set data and the test set data are used for training the back neural network module.

7. The intersection motion trajectory prediction method according to claim 6, characterized in that: the back propagation neural network training learning module is divided into two groups of data according to a preset radius R, a motion state when a vehicle is positioned right in front of a dangerous target and a distance between the vehicle and the dangerous target, and the two groups of data are divided into training set data and testing set data according to a ratio of 6: 4-8: 2 and are used for training the back neural network module.

8. The intersection motion trajectory prediction method according to claim 1, characterized in that: the initial state parameters of the moving target comprise one or more of speed, acceleration, position and angular speed.

9. The intersection motion trajectory prediction method according to claim 1, characterized in that: the target state of the dangerous target comprises one or more of speed, acceleration, position and angular velocity.

10. A motion trail prediction device operated according to the motion trail prediction method for the intersection according to any one of claims 1 to 9.

Images