Continuous planning method for dynamic obstacle avoidance path of intelligent bicycle sports

Abstract

The invention relates to the technical field of obstacle avoidance path planning, and discloses an intelligent sports vehicle dynamic obstacle avoidance path continuous planning method comprising the following steps: S1, obtaining a racing track pre-stored map, and constructing a racing track reference coordinate system; relates to the technical field of obstacle avoidance path planning, and the method comprises the steps: predicting the probability distribution of the future position of an obstacle through employing a recursive Bayesian filtering algorithm based on the historical observation data of the obstacle, carrying out the modeling through multivariate normal distribution, converting the modeling into a geometric confidence region, and forming a confidence occupancy set covering the uncertainty of the obstacle; a space-time safety channel is generated through set operation by combining the intelligent vehicle appearance envelope sequence and the racing track boundary; the method does not need to depend on a single prediction track, and by depicting the uncertainty of obstacle movement, even if the actual position of the obstacle deviates from prediction, the intelligent vehicle can still run in the safety channel, so that the problems of planning failure and conflict between the vehicle and the obstacle caused by deterministic prediction of the actual position of the obstacle in the prior art are effectively relieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of obstacle avoidance path planning, in particular to a dynamic obstacle avoidance path continuous planning method for intelligent vehicle racing. BACKGROUND

[0002] In the field of intelligent vehicle racing, the dynamic obstacle avoidance capability of the race track is a core factor affecting the driving safety and racing performance of the intelligent vehicle, and the key lies in the accurate prediction of dynamic obstacles and robust path planning.

[0003] On the one hand, the trajectory prediction of dynamic obstacles in the prior art generally relies on simple deterministic models, such as uniform extrapolation model, uniform acceleration extrapolation model, etc. Such models can only linearly predict the future trajectory based on the historical motion state of the obstacle, completely ignoring the possible sudden change in speed, temporary pause, and nonlinear turning motion of the obstacle in the intelligent vehicle racing track. Due to the lack of effective characterization of the motion uncertainty of the obstacle, when the actual motion pattern of the obstacle deviates from the linear law preset by the model, the prediction result cannot accurately reflect the actual position of the obstacle at the future time, thereby causing the subsequent obstacle avoidance planning based on the prediction result to lose reference value and causing planning failure problems; On the other hand, the existing obstacle avoidance planning scheme is usually based on a single deterministic predicted trajectory for path optimization design. Once the actual position of the obstacle deviates from the predicted trajectory, there is a spatial conflict between the intelligent vehicle and the obstacle when the intelligent vehicle drives according to the planned path. At the same time, due to the lack of modeling ability for the uncertainty of the obstacle, the prior art often adopts an overly conservative obstacle avoidance strategy, such as increasing the detour distance and reducing the driving speed, which causes the intelligent vehicle to be unable to drive along the near-optimal path, and the global racing performance is limited. SUMMARY

[0004] To solve the technical problems in the background art, the present application provides a dynamic obstacle avoidance path continuous planning method for intelligent vehicle racing.

[0005] The dynamic obstacle avoidance path continuous planning method for intelligent vehicle racing provided by the present application comprises the following steps: S1, obtaining a pre-stored track map to construct a track reference coordinate system; S2, constructing a coordinate conversion matrix of the intelligent vehicle local coordinate system and the track reference coordinate system; S3, collecting track geometry data and dynamic obstacle data through sensors to generate an observation set; S4, generating a confidence occupancy set according to the observation set; S5, generating a spatiotemporal safety channel according to the confidence occupancy set; S6, generating a driving path of the intelligent vehicle in the spatiotemporal safety channel.

[0006] Preferably, in S1, a pre-stored track map is acquired, and a track reference coordinate system is constructed, as follows: According to the pre-stored track map, a track reference coordinate system is established with the midpoint of the starting line of the track as the origin, the tangent direction of the design center line of the track as the longitudinal coordinate axis, and the direction perpendicular to the longitudinal coordinate axis and pointing to the inside of the track as the transverse coordinate axis. As an illustration, the right-hand rule refers to the thumb pointing to the positive direction of the X-axis, the index finger pointing to the positive direction of the Y-axis, and the middle finger pointing to the positive direction of the Z-axis perpendicular to the ground.

[0007] Preferably, in S2, a coordinate conversion matrix between the intelligent vehicle local coordinate system and the track reference coordinate system is constructed, as follows: GNSS / IMU fusion data of the intelligent vehicle is acquired, including intelligent vehicle positioning data and motion state data; the motion state data includes the real-time heading angle and speed of the intelligent vehicle; The position coordinates of the intelligent vehicle in the track reference coordinate system are obtained through the positioning data; The angle between the longitudinal coordinate axis of the intelligent vehicle local coordinate system and the track reference coordinate system is obtained through the real-time heading angle of the intelligent vehicle motion state data; According to the position coordinates of the intelligent vehicle in the track reference coordinate system and the angle between the longitudinal coordinate axis of the intelligent vehicle local coordinate system and the track reference coordinate system, the coordinate conversion matrix between the intelligent vehicle local coordinate system and the track reference coordinate system is obtained through the Euclidean transformation algorithm. As an illustration: the intelligent vehicle local coordinate system refers to the coordinate system with the intelligent vehicle centroid as the origin and the vehicle head direction as the X-axis; the Euclidean transformation algorithm is an existing algorithm for implementing translation and rotation combination operation.

[0008] Preferably, in S3, track geometry data and dynamic obstacle data are collected through sensors to generate an observation set, as follows: Track geometry data and dynamic obstacle data are collected through sensors; The sensors include laser radar, camera, and millimeter wave radar; the track geometry data includes track boundary data and track design center line curvature data; The GNSS / IMU fusion data, track geometry data, and dynamic obstacle data are fused through multi-sensor data fusion technology to obtain a first perception data set; The first perception data set is converted to the track reference coordinate system through the coordinate conversion matrix to obtain a second perception data set, and the second perception data set is processed through Kalman filtering to obtain a third perception data set; According to the third perception data set, intelligent vehicle reference center line, track boundary, and obstacle historical observation data are generated, and the intelligent vehicle reference center line, track boundary, and obstacle historical observation data form an observation set.

[0009] Preferably, in S3, according to the third perception dataset, the intelligent vehicle reference center line, the track boundary and the obstacle historical observation data are generated as follows: According to the third perception dataset; The intelligent vehicle reference center line is generated by fitting the track geometry data in the third perception dataset and the GNSS / IMU fusion data; The track boundary data of the track geometry data in the third perception dataset is extracted as the track boundary; The dynamic obstacle data in the third perception dataset is sorted by timestamp to form the obstacle historical observation data.

[0010] Preferably, in S4, according to the observation set, the confidence occupancy set is generated as follows: According to the obstacle historical observation data in the observation set, the recursive Bayesian filtering algorithm is used to predict the position probability distribution of the dynamic obstacle at each time in the future unit time; The position probability distribution at all times in the future unit time forms a position probability distribution set, and the position probability distribution set is taken as a probability occupancy area; According to the position probability distribution of the dynamic obstacle at each time in the future unit time, a multivariate normal distribution model S k , k=1, 2,..., T, T is the total time of all times in the future unit time; The multivariate normal distribution model S k includes the position mean vector and the 2x2 position covariance matrix corresponding to the time; According to the preset confidence interval and the chi-square distribution critical value corresponding to the confidence interval, the multivariate normal distribution model S k at each time is converted into a geometric confidence area by a confidence interval calculation method of the multivariate normal distribution; The geometric confidence areas of all times form a confidence occupancy set; As an illustration: the future unit time can be set to 10ms; the preset confidence interval can be set to 95%; The geometric confidence area can be understood as follows: taking the position mean vector corresponding to the time as the center, decomposing the 2x2 position covariance matrix by the analytical solution method to obtain two eigenvalues λ1 and λ2, wherein λ1≥λ2, taking the square root of λ1 multiplied by the chi-square distribution critical value as the long semi-axis, and taking the square root of λ2 multiplied by the chi-square distribution critical value as the short semi-axis, to obtain an elliptical area as the geometric confidence area; if λ1=λ2, a circular area is formed as the geometric confidence area, and at this time the center of the circle is the position mean vector corresponding to the time, and the radius is λ1 or λ2 multiplied by the square root of the chi-square distribution critical value; The confidence interval calculation method of multivariate normal distribution is the prior art, and the confidence interval calculation method of multivariate normal distribution can refer to the existing public literature “Probabilistic Robotics” (Sebastian Thrun) section 5.3, and a person skilled in the art can directly reproduce it.

[0011] Preferably, in S5, the spatiotemporal safety channel is generated according to the confidence occupancy set, as follows: Obtaining the geometric shape parameters of the intelligent vehicle in the track reference coordinate system; According to the GNSS / IMU fusion data in the third perception data set, the position and heading angle of the intelligent vehicle in the track reference coordinate system at each time in the future unit time are generated through a machine learning algorithm; According to the position and heading angle of the intelligent vehicle in the track reference coordinate system at each time in the future unit time, the envelope region of the intelligent vehicle at each time in the future unit time is generated in the track reference coordinate system through the geometric shape parameters of the intelligent vehicle, forming a vehicle envelope sequence; The vehicle envelope region is used to represent the space range occupied by the intelligent vehicle at a certain time in the track reference coordinate system; The confidence occupancy set and the vehicle envelope sequence are set to operate, and the repulsion region of the dynamic obstacle to the intelligent vehicle in the future unit time is obtained; As an illustration: the confidence occupancy set and the vehicle envelope sequence are set to operate, which means that for each time, the geometric confidence region at that time is combined with the envelope region of the intelligent vehicle at the corresponding time to obtain the joint no-entry zone at that time. The joint no-entry zone at all times is sorted in time sequence, which is the repulsion region of the dynamic obstacle to the intelligent vehicle in the future unit time. Such design is because the planned trajectory of the intelligent vehicle cannot enter the space that will be occupied at a certain time in the future. For example, the planned trajectory is exactly the position of the vehicle body 100 ms later, and at that time the vehicle body will overlap with its own trajectory, which is equivalent to scratching itself. In actual driving, problems such as steering lag and vehicle body deviation may occur; According to the track boundary in the observation set, the repulsion region of the dynamic obstacle to the intelligent vehicle in the future unit time and the track boundary region are set to operate, and the spatiotemporal safety channel is formed in the track reference coordinate system; The spatiotemporal safety channel is used to represent the drivable planning space of the intelligent vehicle within the track boundary range under the condition of obstacle uncertainty, which satisfies the obstacle avoidance constraint; As an illustration, according to the track boundary in the observation set, the repulsion region of the dynamic obstacle to the intelligent vehicle in the future unit time and the track boundary region are set to operate, which means set difference operation, to obtain the collision-free drivable space at each time in the future unit time. By concatenating all the collision-free drivable spaces in time sequence, the spatiotemporal safety channel can be formed in the track reference coordinate system.

[0012] Preferably, in S6, the driving path of the intelligent vehicle is generated in the spatiotemporal safety channel as follows: The preset curvature continuity constraint is acquired, the intelligent vehicle reference center line in the observation set is taken as the path reference in the spatiotemporal safety channel, the intelligent vehicle reference center line is searched for local path points by using the A* algorithm, a discrete obstacle avoidance path point sequence in a future unit time is generated, the discrete obstacle avoidance path point sequence in the future unit time is smoothed and fitted by using the cubic spline interpolation method, a curvature-continuous smooth path segment is generated, the curvature-continuous smooth path segment satisfies the preset curvature continuity constraint, and the driving path of the intelligent vehicle is obtained. As an illustration, when the intelligent vehicle reference center line is searched for local path points by using the A* algorithm, the search range is limited in the spatiotemporal safety channel.

[0013] Preferably, in S6, the driving path of the intelligent vehicle is generated in the spatiotemporal safety channel as follows: The preset curvature continuity constraint is acquired, the intelligent vehicle reference center line in the observation set is taken as the path reference in the spatiotemporal safety channel, the intelligent vehicle reference center line is searched for local path points by using the A* algorithm, a discrete obstacle avoidance path point sequence in a future unit time is generated, the discrete obstacle avoidance path point sequence in the future unit time is smoothed and fitted by using the cubic spline interpolation method, a curvature-continuous smooth path segment is generated, the curvature-continuous smooth path segment satisfies the preset curvature continuity constraint, and the driving path of the intelligent vehicle is obtained. The preset vehicle speed is set, the distance between path points of each candidate path point sequence is acquired, the total path time T is acquired according to the preset vehicle speed and the distance between path points of the candidate path point sequence. The sum L of adjacent point distances of each candidate path point sequence is acquired. T and L are normalized to obtain T1 and L1. For each candidate path point sequence, the comprehensive cost J is calculated, and the comprehensive cost J formula is: J=α×T1+β×L1, α+β=1, and α and β are preset weights. The candidate path point sequence with the minimum comprehensive cost J is selected as the selected path point sequence. The selected path point sequence is smoothed and fitted by using the cubic spline interpolation, a curvature-continuous smooth path segment is generated, the curvature-continuous smooth path segment satisfies the preset curvature continuity constraint, and the driving path of the intelligent vehicle is obtained.

[0014] In the application, the intelligent vehicle competitive dynamic obstacle avoidance path continuous planning method has the following beneficial technical effects: 1.The application first acquires a pre-stored track map to construct a track reference coordinate system, providing a unified spatial reference for data; then, combined with GNSS / IMU fusion data of the intelligent vehicle, the conversion matrix of the local coordinate system and the track reference coordinate system is generated through the Euclidean transformation algorithm by obstacle uncertainty modeling and space-time safety boundary construction, eliminating coordinate deviation; after collecting track geometric data and dynamic obstacle data, the observation set containing the reference center line, track boundary and obstacle historical observation data is obtained through multi-sensor fusion, coordinate conversion and Kalman filter processing; next, based on the obstacle historical observation data, the recursive Bayesian filtering algorithm is used to predict the future position probability distribution of the obstacle, which is converted into a geometric confidence region through multivariate normal distribution modeling, forming a confidence occupancy set covering the uncertainty of the obstacle; combined with the shape envelope sequence of the intelligent vehicle and the track boundary, the space-time safety channel is generated through set operation; this design does not rely on a single predicted trajectory, and by depicting the uncertainty of the obstacle motion, even if there is a deviation between the actual position of the obstacle and the prediction, the intelligent vehicle can still drive in the safety channel, improving the planning robustness, effectively alleviating the planning failure and vehicle-obstacle conflict problems caused by the determination of the actual position of the obstacle in the prior art, and improving the reliability of the intelligent vehicle in track obstacle avoidance.

[0015] 2.In the preset constraint range of the space-time safety channel, the intelligent vehicle reference center line in the observation set is taken as the path reference, and the discrete obstacle avoidance path point sequence in the future unit time is generated through the A* algorithm, and the driving path of the intelligent vehicle is obtained by using the cubic spline interpolation method, which relieves the problem of excessive detour and sudden speed drop caused by the lack of clear safety boundary in traditional obstacle avoidance.

[0016] Additional aspects and advantages of the application will be in part apparent and in part pointed out hereinafter. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 The flowchart of the method of the application. DETAILED DESCRIPTION

[0018] Embodiments of the application are described in detail below with reference to the attached drawings, in which the same or similar elements are denoted by the same or similar reference signs throughout. The embodiments described below with reference to the attached drawings are exemplary and are intended only to explain the application, and cannot be understood as limiting the application.

[0019] As Figure 1 shown in a kind of intelligent vehicle competitive dynamic obstacle avoidance path continuous planning method, comprising the following steps: S1, acquires a pre-stored track map, constructs a track reference coordinate system; In an optional embodiment, in S1, a pre-stored track map is acquired, and a track reference coordinate system is constructed as follows: According to the pre-stored track map, a track reference coordinate system is established with the midpoint of the starting line of the track as the origin, the tangent direction of the design center line of the track as the longitudinal coordinate axis, and the direction perpendicular to the longitudinal coordinate axis and pointing to the inside of the track as the transverse coordinate axis; As an illustration, the right-hand rule refers to the thumb pointing to the positive direction of the X-axis, the index finger pointing to the positive direction of the Y-axis, and the middle finger pointing to the positive direction of the Z-axis perpendicular to the ground; In S2, a coordinate conversion matrix of the local coordinate system of the intelligent vehicle and the track reference coordinate system is constructed as follows: In an optional embodiment, in S2, a coordinate conversion matrix of the local coordinate system of the intelligent vehicle and the track reference coordinate system is constructed as follows: GNSS / IMU fusion data of the intelligent vehicle is acquired, including intelligent vehicle positioning data and motion state data; the motion state data includes the real-time heading angle (vehicle head direction) and speed of the intelligent vehicle; The position coordinates of the intelligent vehicle in the track reference coordinate system are obtained through the positioning data; The angle between the longitudinal coordinate axis of the local coordinate system of the intelligent vehicle and the track reference coordinate system is obtained through the real-time heading angle of the intelligent vehicle from the motion state data; According to the position coordinates of the intelligent vehicle in the track reference coordinate system and the angle between the longitudinal coordinate axis of the local coordinate system of the intelligent vehicle and the track reference coordinate system, the coordinate conversion matrix of the local coordinate system of the intelligent vehicle and the track reference coordinate system is obtained through the Euclidean transformation algorithm; As an illustration, the local coordinate system of the intelligent vehicle refers to the coordinate system with the center of mass of the intelligent vehicle as the origin and the vehicle head direction as the X-axis; the Euclidean transformation algorithm is an existing algorithm for realizing the combination of translation and rotation operation; In S3, track geometry data and dynamic obstacle data are collected through sensors to generate an observation set; In an optional embodiment, in S3, track geometry data and dynamic obstacle data are collected through sensors to generate an observation set as follows: Track geometry data and dynamic obstacle data are collected through sensors; The sensors include lidar, camera, and millimeter wave radar; the track geometry data includes track boundary data and track design center line curvature data; The GNSS / IMU fusion data, track geometry data, and dynamic obstacle data are fused through multi-sensor data fusion technology to obtain a first perception data set; The first perception data set is converted to the track reference coordinate system through the coordinate conversion matrix to obtain a second perception data set, and the second perception data set is processed through Kalman filtering to obtain a third perception data set; According to the third perception dataset, the intelligent vehicle reference center line, the track boundary and the obstacle historical observation data are generated, and the intelligent vehicle reference center line, the track boundary and the obstacle historical observation data form an observation set; In an optional embodiment, in S3, according to the third perception dataset, the intelligent vehicle reference center line, the track boundary and the obstacle historical observation data are generated as follows: According to the third perception dataset; The intelligent vehicle reference center line is generated by fitting the track geometry data in the third perception dataset and the GNSS / IMU fusion data; The track boundary data of the track geometry data in the third perception dataset are extracted as the track boundary; The dynamic obstacle data in the third perception dataset are sorted according to the time stamp to form the obstacle historical observation data; S4, according to the observation set, a confidence occupancy set is generated; In an optional embodiment, in S4, according to the observation set, a confidence occupancy set is generated as follows: According to the obstacle historical observation data in the observation set, the recursive Bayesian filtering algorithm is used on the obstacle historical observation data to predict the position probability distribution of the dynamic obstacle at each time in a future unit time; The position probability distribution at all times in the future unit time forms a position probability distribution set, and the position probability distribution set is taken as a probability occupancy area; According to the position probability distribution of the dynamic obstacle at each time in the future unit time, a multivariate normal distribution is used to model each time position probability distribution to obtain a multivariate normal distribution model S k , k = 1, 2,..., T, T is the total time of all times in the future unit time; The multivariate normal distribution model S k includes the position mean vector and the 2*2 position covariance matrix corresponding to the time; According to the preset confidence interval and the critical value of the chi-square distribution corresponding to the confidence interval, the multivariate normal distribution model S k at each time is converted into a geometric confidence area by a confidence interval calculation method of the multivariate normal distribution; The geometric confidence areas of all times form a confidence occupancy set; As an illustration: the future unit time can be set to 10ms; the preset confidence interval can be set to 95%; The geometric confidence region can be understood as follows: taking the position mean vector at the corresponding moment as the center, decomposing the 2x2 position covariance matrix by an analytical solution to obtain two eigenvalues λ1 and λ2, where λ1≥λ2, taking the square root of λ1 multiplied by the critical value of the chi-square distribution as the long semi-axis, and taking the square root of λ2 multiplied by the critical value of the chi-square distribution as the short semi-axis, and taking the obtained elliptical region as the geometric confidence region; if λ1=λ2, a circular region is formed as the geometric confidence region, and the center of the circle is the position mean vector at the corresponding moment, and the radius is the square root of λ1 or λ2 multiplied by the critical value of the chi-square distribution; The confidence interval calculation method of multivariate normal distribution is prior art, and the confidence interval calculation method of multivariate normal distribution can refer to the existing public literature “Probabilistic Robotics” (Sebastian Thrun) Section 5.3, and a person skilled in the art can directly reproduce it; S5, generating a space-time safe channel according to the confidence occupancy set; In an optional embodiment, in S5, a space-time safe channel is generated according to the confidence occupancy set as follows: Obtaining the geometric shape parameters of the intelligent vehicle in the track reference coordinate system; According to the GNSS / IMU fusion data in the third perception data set, the position and heading angle of the intelligent vehicle in the track reference coordinate system at each moment in the future unit time are generated by a machine learning algorithm; According to the position and heading angle of the intelligent vehicle in the track reference coordinate system at each moment in the future unit time, the envelope region of the intelligent vehicle at each moment in the future unit time is generated in the track reference coordinate system by the geometric shape parameters of the intelligent vehicle, and a vehicle envelope sequence is formed; The vehicle envelope region is used to represent the space range occupied by the intelligent vehicle at a moment in the track reference coordinate system; The confidence occupancy set and the vehicle envelope sequence are set to operate to obtain the repulsion region of the dynamic obstacle to the intelligent vehicle in the future unit time; As an illustration: the set operation of the confidence occupancy set and the vehicle envelope sequence means that for each moment, the geometric confidence region at the moment is combined with the envelope region of the intelligent vehicle at the corresponding moment to obtain a joint no-entry zone at the moment; the joint no-entry zones at all moments are sorted in time sequence, which is the repulsion region of the dynamic obstacle to the intelligent vehicle in the future unit time; such a design is because the planned trajectory of the intelligent vehicle cannot enter the space to be occupied at a moment in the future, for example, the planned trajectory is exactly the position of the vehicle body 100 ms later, and at that time the vehicle body will overlap with the trajectory, which is equivalent to scratching oneself, and in actual driving, the problem of steering jam and vehicle body deviation will occur; According to the track boundary in the observation set, the exclusion region of the dynamic obstacle to the intelligent vehicle in the future unit time is set operation with the track boundary region, and a time-space safety channel is formed in the track reference coordinate system; The time-space safety channel is used to represent the drivable planning space of the intelligent vehicle in the track boundary range under the condition that the obstacle uncertainty exists and the obstacle avoidance constraint is met; As an illustration, according to the track boundary in the observation set, the exclusion region of the dynamic obstacle to the intelligent vehicle in the future unit time is set operation with the track boundary region, which is set difference operation, and the collision-free drivable space at each time in the future unit time is obtained. By concatenating all the collision-free drivable spaces in time sequence, the time-space safety channel is formed in the track reference coordinate system.

[0020] The application first acquires a pre-stored track map to construct a track reference coordinate system, which provides a unified spatial reference for data. Then, combined with the GNSS / IMU fusion data of the intelligent vehicle, the conversion matrix of the local coordinate system and the track reference coordinate system is generated by the Euclidean transformation algorithm through obstacle uncertainty modeling and time-space safety boundary construction, which eliminates the coordinate deviation. After collecting the track geometry data and dynamic obstacle data, the observation set containing the reference center line, track boundary and obstacle historical observation data is obtained through multi-sensor fusion, coordinate conversion and Kalman filter processing. Next, the recursive Bayesian filtering algorithm is used to predict the future position probability distribution of the obstacle based on the obstacle historical observation data, which is converted into a geometric confidence region through multivariate normal distribution modeling to form a confidence occupancy set covering the obstacle uncertainty. Combined with the shape envelope sequence of the intelligent vehicle and the track boundary, the time-space safety channel is generated through set operation. This design does not rely on a single predicted trajectory. By depicting the uncertainty of the obstacle motion, even if there is a deviation between the actual position of the obstacle and the prediction, the intelligent vehicle can still drive in the safety channel, improving the planning robustness and effectively alleviating the planning failure and vehicle-obstacle conflict problems caused by the determination of the actual position of the obstacle in the prior art, thereby improving the reliability of the intelligent vehicle in track obstacle avoidance.

[0021] S6, generating a driving path of the intelligent vehicle in the time-space safety channel.

[0022] In an optional embodiment, in S6, a driving path of the intelligent vehicle is generated in the time-space safety channel, as follows: A preset curvature continuity constraint is obtained, and a local path point search is performed on the intelligent vehicle reference center line in the time-space safety channel by using the A* algorithm, to generate a discrete obstacle avoidance path point sequence in the future unit time. A cubic spline interpolation method is used to smooth and fit the discrete obstacle avoidance path point sequence in the future unit time, to generate a curvature-continuous smooth path segment, and the curvature-continuous smooth path segment meets the preset curvature continuity constraint, thereby obtaining the driving path of the intelligent vehicle. As an illustration, when the intelligent vehicle reference center line is searched by the A* algorithm, the search range is limited in the space-time safe channel; In an optional embodiment, in S6, the driving path of the intelligent vehicle is generated in the space-time safe channel as follows: A preset curvature continuity constraint is obtained, the intelligent vehicle reference center line is searched by the A* algorithm in the space-time safe channel with the intelligent vehicle reference center line in the observation set as the path reference, and a plurality of groups of discrete obstacle avoidance path point sequences in a unit time in the future are generated as candidate path point sequences; A preset vehicle speed is set, the distance between path points of each candidate path point sequence is obtained, and the total path time T is obtained according to the preset vehicle speed and the distance between path points of the candidate path point sequence; The sum L of the distances between adjacent points of each candidate path point sequence is obtained; T and L are normalized to obtain T1 and L1; For each candidate path point sequence, the comprehensive cost J is calculated, and the comprehensive cost J formula is: J = α × T1 + β × L1, α + β = 1, and α and β are preset weights; The candidate path point sequence with the minimum comprehensive cost J is selected as the selected path point sequence; The selected path point sequence is smoothed by using a cubic spline interpolation, a curvature-continuous smooth path segment is generated, the curvature-continuous smooth path segment meets the preset curvature continuity constraint, and the driving path of the intelligent vehicle is obtained.

[0023] In the space-time safe channel, the discrete obstacle avoidance path point sequence in a unit time in the future is generated by the A* algorithm with the intelligent vehicle reference center line in the observation set as the path reference, the driving path of the intelligent vehicle is obtained by fitting through the cubic spline interpolation method, and the problems of excessive detour and sudden speed drop caused by no clear safe boundary in traditional obstacle avoidance are relieved.

[0024] Meanwhile, the contents not described in detail in the specification all belong to the prior art known by those skilled in the art.

[0025] In the embodiments provided by the present application, it should be understood that the disclosed system or method can be implemented in other ways. For example, the above-described embodiments of the application are merely illustrative, for example, the division of modules is merely a logical function division, and there can be another division manner in actual implementation.

[0026] The modules described as separate components may or may not be physically separate, and the components displayed as modules may or may not be physical modules, and may be located in one place or distributed to multiple network modules. Part or all of the modules can be selected according to actual needs to achieve the purpose of the embodiment.

[0027] In addition, each functional module in each embodiment of the present application can be integrated into one processing module, or each module can exist physically alone, or two or more modules can be integrated into one module. The integrated module can be realized in the form of hardware or in the form of hardware plus software functional module.

[0028] For those skilled in the art, it is obvious that the present application is not limited to the details of the above exemplary embodiments, and the present application can be implemented in other specific forms without departing from the essential characteristics of the present application.

[0029] The above is only the preferred specific implementation of the present application, but the protection scope of the present application is not limited to this. Any skilled person in the art can make equivalent replacements or changes according to the technical solution and the inventive concept of the present application within the technical range disclosed by the present application, which should be covered within the protection scope of the present application.

Claims

1. A method for continuous planning of dynamic obstacle avoidance paths for intelligent vehicles, characterized in that: The following steps are involved: S1. Obtain a pre-stored track map and construct a track reference coordinate system; S2, constructing a coordinate transformation matrix between the smart car's local coordinate system and the track's reference coordinate system; S3, collect track geometry data and dynamic obstacle data through sensors to generate an observation set; S4. Generate a confidence occupancy set based on the observation set; S5. Generate a space-time secure channel based on the confidence occupation set; S6. Generate a driving path for the smart car within the space-time safety channel.

2. The method for continuous planning of dynamic obstacle avoidance paths for intelligent vehicles according to claim 1, characterized in that: In S1, the pre-stored map of the track is obtained and the track reference coordinate system is constructed as follows: According to the pre-stored map of the track, the midpoint of the track starting line is taken as the origin, the tangent direction along the designed center line of the track is the longitudinal coordinate axis, and the transverse coordinate axis perpendicular to the longitudinal coordinate axis and pointing to the inside of the track is the transverse coordinate axis. The track reference coordinate system is established through the right-hand system.

3. The method for continuous planning of dynamic obstacle avoidance paths for intelligent vehicles according to claim 1, characterized in that: In S2, the coordinate transformation matrix between the local coordinate system of the smart car and the track reference coordinate system is constructed as follows: Obtain the GNSS / IMU fusion data of the smart car. The GNSS / IMU fusion data includes the smart car's positioning data and motion status data; the motion status data includes the smart car's real-time heading angle and speed; The position coordinates of the smart car in the track reference coordinate system are obtained through positioning data; The angle between the longitudinal coordinate axis of the local coordinate system of the smart car and the track reference coordinate system is obtained through the real-time heading angle of the smart car in the motion state data; According to the position coordinates of the smart car in the track reference coordinate system and the angle between the longitudinal coordinate axes of the smart car local coordinate system and the track reference coordinate system, the coordinate transformation matrix between the smart car local coordinate system and the track reference coordinate system is obtained through the Euclidean transformation algorithm.

4. The method for continuous planning of dynamic obstacle avoidance paths for intelligent vehicles according to claim 3, characterized in that: In S3, track geometry data and dynamic obstacle data are collected through sensors to generate an observation set as follows: Collect track geometry data and dynamic obstacle data through sensors; Sensors include lidar, cameras, and millimeter-wave radar; track geometry data includes track boundary data and track design centerline curvature data; The GNSS / IMU fusion data, track geometry data, and dynamic obstacle data are integrated through multi-sensor data fusion technology to obtain the first perception data set; The first perception data set is converted into a track reference coordinate system by using a coordinate conversion matrix to obtain a second perception data set, and the second perception data set is processed by a Kalman filter to obtain a third perception data set; Based on the third perception data set, the intelligent vehicle reference center line, track boundary and obstacle historical observation data are generated, and the intelligent vehicle reference center line, track boundary and obstacle historical observation data form an observation set.

5. The method for continuous planning of dynamic obstacle avoidance paths for intelligent vehicles according to claim 4, characterized in that: In S3, based on the third perception dataset, the intelligent vehicle reference centerline, track boundary, and obstacle historical observation data are generated as follows; According to the third perception dataset; Generate the intelligent vehicle reference centerline by fitting the track geometry data in the third-party perception dataset with the GNSS / IMU fusion data; extracting track boundary data of the track geometry data in the third perception dataset as a track boundary; The dynamic obstacle data in the third perception dataset are sorted by timestamp to form obstacle history observation data.

6. The method for continuous planning of dynamic obstacle avoidance paths for intelligent vehicles according to claim 5, characterized in that: In S4, based on the observation set, the confidence occupancy set is generated as follows: Based on the historical observation data of obstacles in the observation set, the recursive Bayesian filtering algorithm is used to predict the historical observation data of obstacles to obtain the position probability distribution of dynamic obstacles at each moment in the future unit time; The position probability distribution of all moments in the future unit time forms a position probability distribution set, and the position probability distribution set is used as the probability occupancy area; According to the position probability distribution of dynamic obstacles at each moment in the future unit time, the multivariate normal distribution model S of the position probability distribution at each moment is obtained. k , k=1, 2, ..., T, T is the total number of moments in the future unit time; Multivariate normal distribution model S k Including the position mean vector and 2×2 position covariance matrix at the corresponding moment; According to the preset confidence interval and the chi-square distribution critical value corresponding to the confidence interval, the multivariate normal distribution model S at each moment is calculated by the confidence interval calculation method of the multivariate normal distribution. k Convert to geometric confidence region; The geometrized confidence regions at all times form the confidence occupancy set.

7. The method for continuous planning of dynamic obstacle avoidance paths for intelligent vehicles according to claim 6, characterized in that: In S5, a spatiotemporal secure channel is generated based on the confidence occupancy set as follows: Obtaining the geometrical parameters of the smart car in the track reference coordinate system; Based on the GNSS / IMU fusion data in the third perception dataset, a machine learning algorithm is used to generate the position and heading angle of the smart car in the track reference coordinate system at each moment in the future unit time. Based on the position and heading angle of the smart car at each moment in the future unit time in the track reference coordinate system, the shape envelope area of ​​the smart car at each moment in the future unit time is generated in the track reference coordinate system using the smart car's geometric shape parameters to form a vehicle shape envelope sequence; Perform set operations on the confidence occupancy set and the vehicle shape envelope sequence to obtain the exclusion area of ​​the smart car caused by dynamic obstacles in the future unit time. According to the track boundary in the observation set, the exclusion area of ​​the smart car caused by dynamic obstacles in the future unit time and the track boundary area are collectively calculated to form a spatiotemporal safe channel in the track reference coordinate system.

8. The method for continuous planning of dynamic obstacle avoidance paths for intelligent vehicles according to claim 7, characterized in that: In S6, the driving path of the smart car is generated in the spatiotemporal safety channel as follows: The preset curvature continuity constraint is obtained. Within the spatiotemporal safety channel, the reference center line of the smart car in the observation set is used as the path reference. The A* algorithm is used to search for local path points on the reference center line of the smart car to generate a discrete obstacle avoidance path point sequence in the future unit time. The discrete obstacle avoidance path point sequence in the future unit time is smoothly fitted using the cubic spline interpolation method to generate a smooth path segment with continuous curvature. The smooth path segment with continuous curvature meets the preset curvature continuity constraint, and the driving path of the smart car is obtained.

9. The method for continuous planning of dynamic obstacle avoidance paths for intelligent vehicles according to claim 7, characterized in that: In S6, the driving path of the smart car is generated in the spatiotemporal safety channel as follows: Obtain the preset curvature continuity constraint. Within the spatiotemporal safety channel, use the intelligent vehicle reference centerline in the observation set as the path reference. Use the A* algorithm to search for local path points on the intelligent vehicle reference centerline, and generate multiple sets of discrete obstacle avoidance path point sequences within the future unit time as candidate path point sequences. Set a preset vehicle speed, obtain the distance between the path points of each set of candidate path point sequences, and obtain the total path time T based on the preset vehicle speed and the distance between the path points of the candidate path point sequences; Obtain the sum L of the distances between adjacent points in each set of candidate path point sequences; Normalize T and L to get T1 and L1; For each set of candidate path point sequences, calculate the comprehensive cost J. The formula for the comprehensive cost J is: J=α×T1+β×L1, α+β=1, α and β are preset weights; Filter out the candidate path point sequence with the smallest comprehensive cost J as the selected path point sequence; Cubic spline interpolation is used to smoothly fit the selected path point sequence to generate a smooth path segment with continuous curvature. The smooth path segment with continuous curvature meets the preset curvature continuity constraint, and the driving path of the smart car is obtained.

Images