CBF-based automatic driving generalization safety decision control correction method and device

Abstract

The invention relates to the technical field of automatic driving decision control, in particular to a CBF-based automatic driving generalization safety decision control correction method and device, and the method comprises the steps: obtaining a traffic scene and various types of perception information related to a vehicle, carrying out the unified conversion processing of the various types of perception information, and obtaining a processing result, and on the basis of the processing result and the initial decision control result, constructing a kinetic model, an obstacle constraint model and a road boundary constraint model of the vehicle to generate an optimization problem normal form for control correction, and obtaining control quantity output after safety correction is finally considered by using the optimization problem normal form for control correction so as to carry out safety control correction. Therefore, the problems that in related technologies, a method based on simple rules cannot strictly guarantee safety and lacks generalization ability, and a method based on CBF has mathematical completeness, but heterogeneous perception input formats are incompatible and poor in adaptability to complex scenes are solved.

Description

Technical Field

[0001] This application relates to the technical field of autonomous driving decision control, and particularly to a method and device for correcting autonomous driving generalization safety decision control based on CBF (Control Barrier Function). Background Art

[0002] In related technologies, there are mainly two types of methods for autonomous driving decision correction: methods based on simple rules and methods based on CBF (Control Barrier Function).

[0003] Methods based on simple rules mainly correct the obtained trajectories, control quantities, etc. based on specific physical rules. Common rules include: path curvature, collision detection, and control feasibility detection, etc. Subsequently, the output quantities that do not meet the rules are corrected to the nearest results that meet the rule conditions. However, although such methods can make the control output more in line with the requirements of the real traffic scenario to a certain extent, they cannot strictly guarantee their safety in mathematics. Secondly, the designed rules still follow the scenario-driven idea and cannot meet the generalization and expansion in various scenarios.

[0004] For methods based on CBF, traffic obstacles are modeled as inequality constraints. Based on the mathematical framework derived from the Lyapunov function, CBF can absolutely guarantee the safety of the decision output as long as the inequality constraints of the barrier function are satisfied under the premise of defining the safety set and the action space, and has good mathematical completeness. However, although CBF has been proven to be effective in safety control correction, there are problems of incompatible heterogeneous perception input formats and poor adaptability to complex scenarios. Summary of the Invention

[0005] This application provides a method and device for correcting autonomous driving generalization safety decision control based on CBF to solve the problems in related technologies that methods based on simple rules cannot strictly guarantee safety and lack generalization ability, and methods based on CBF have mathematical completeness but have problems of incompatible heterogeneous perception input formats and poor adaptability to complex scenarios.

[0006] The first aspect embodiment of this application provides a method for correcting autonomous driving generalization safety decision control based on CBF, including the following steps: obtaining various types of perception information related to the traffic scenario and the host vehicle; performing unified conversion processing on the various types of perception information to obtain a processing result; based on the processing result and the initial decision control result, constructing a dynamic model, an obstacle constraint model, and a road boundary constraint model of the host vehicle to generate an optimization problem paradigm for control correction, and using the optimization problem paradigm for control correction to obtain the final control quantity output after considering safety correction for safety control correction.

[0007] Through the above technical solution, the embodiments of the present application can comprehensively master the vehicle driving environment and its own state by obtaining various types of perception information of the traffic scene and the host vehicle, providing rich basis for decision-making. By uniformly converting and processing this information, the problem of incompatible heterogeneous data formats is solved, the generalization of the decision correction module is improved, and subsequent processing becomes more efficient and accurate. Based on the processing results and the initial decision control results, a host vehicle dynamics model, an obstacle constraint model, and a road boundary constraint model are constructed, comprehensively and quantitatively considering various traffic risk sources, and comprehensively improving the ability to ensure driving safety. An optimization problem paradigm for control correction is generated and the final control quantity output after safety correction is obtained, effectively avoiding dangerous behaviors, enhancing the safety and reliability of autonomous driving decision control, ensuring the vehicle's safe driving in complex and changeable traffic scenes, and promoting the development of autonomous driving technology.

[0008] Optionally, in an embodiment of the present application, the various types of perception information include a grid map and a vectorized bounding box. Among them, the unified conversion and processing of the various types of perception information to obtain a processing result includes: filtering the grid map and clustering the remaining occupied grids to obtain a clustering result; for each cluster of the clustering result, solving the minimum covering convex polygon, and based on the minimum covering convex polygon, solving the minimum covering rectangle and calculating its vectorized representation to obtain the processing result.

[0009] Through the above technical solution, the embodiments of the present application can achieve unified conversion for two common perception data, namely the grid map and the vectorized bounding box, through a series of operations. Filtering the grid map to remove the information that coincides with the vectorized bounding box effectively improves the perception conversion efficiency and avoids data redundancy. Clustering the remaining occupied grids can accurately identify the individual of the irregular obstacle, providing a clear classification basis for subsequent processing. Solving the minimum covering convex polygon for each cluster, and further solving the minimum covering rectangle and obtaining the vectorized representation, enables the heterogeneous grid map and vectorized bounding box data to finally be presented in a unified format. This unified format of data is more easily understood and processed by the autonomous driving system, enhancing the system's perception ability of obstacles in complex traffic scenes, and providing a solid and reliable data basis for subsequent safety decision control correction based on these data.

[0010] Optionally, in an embodiment of the present application, using the optimization problem paradigm of the control correction to obtain the final control quantity output after considering safety correction for safety control correction includes: taking the final control quantity output after considering safety correction as the acceleration and the steering angle to generate a control instruction; sending the control instruction to the controller terminal of the host vehicle.

[0011] Through the above technical solutions, the embodiments of the present application can directly convert the optimized and solved control quantity output into acceleration and rotation angle to generate control instructions, and send them to the vehicle controller terminal. This method makes the execution link of safety control correction simple and efficient, and can quickly convert safety decisions into actual vehicle control instructions. Directly using acceleration and rotation angle as the core of the control instructions, it accurately acts on the power and steering systems of the vehicle, ensuring that the vehicle responds in a timely manner and adjusts its driving state, effectively avoiding dangers and ensuring driving safety.

[0012] Optionally, in an embodiment of the present application, the optimization problem paradigm is:

[0013]

[0014] s.t.L f h Fi +L g h Fi u+β i h Fi ≥0

[0015] L f h rj +L g h rj u+γ i h rj ≥0

[0016] Wherein, u is the control input; u o is the initial decision control result received; L f is the partial derivative of the obstacle function h with respect to g(x); L g is the partial derivative of the obstacle function h with respect to f(x); Q is a constant matrix; h Fi is the constraint corresponding to obstacle i; h rj is the constraint corresponding to road boundary j; β i and γ i are constants.

[0017] Through the above technical solutions, the embodiments of the present application can, on the premise of meeting safety constraints, make the corrected control quantity as close as possible to the initial decision control result, ensuring both safety and maximizing the retention of the original decision-making intention, and avoiding adverse effects on vehicle driving stability and comfort caused by excessive correction.

[0018] In the second aspect of the embodiments of the present application, a CBF-based generalized safety decision control correction device for autonomous driving is provided, including: an acquisition module, configured to acquire various types of perception information related to the traffic scene and the host vehicle; a conversion module, configured to perform unified conversion processing on the various types of perception information to obtain a processing result; a correction module, configured to construct a dynamic model, an obstacle constraint model, and a road boundary constraint model of the host vehicle based on the processing result and the initial decision control result, to generate an optimization problem paradigm for control correction, and use the optimization problem paradigm for control correction to obtain the output of the control amount after finally considering safety correction for safety control correction.

[0019] Through the above technical solution, the embodiments of the present application can comprehensively master the vehicle driving environment and its own state by acquiring the traffic scene and various types of perception information of the host vehicle, providing a rich basis for decision-making. Performing unified conversion processing on this information can solve the problem of incompatible heterogeneous data formats, improve the generalization of the decision correction module, and make subsequent processing more efficient and accurate. Constructing a dynamic model, an obstacle constraint model, and a road boundary constraint model of the host vehicle based on the processing result and the initial decision control result comprehensively and quantitatively considers various traffic risk sources, comprehensively improving the ability to guarantee driving safety. Generating an optimization problem paradigm for control correction and obtaining the output of the control amount after final safety correction can effectively avoid dangerous behaviors, enhance the safety and reliability of autonomous driving decision control, ensure the vehicle's safe driving in complex and changeable traffic scenes, and promote the development of autonomous driving technology.

[0020] Optionally, in an embodiment of the present application, the various types of perception information include a grid map and a vectorized bounding box, wherein the conversion module includes: a filtering and clustering unit, configured to filter the grid map and cluster the remaining occupied grids to obtain a clustering result; a solving unit, configured to solve the minimum covering convex polygon for each cluster of the clustering result, and solve the minimum covering rectangle according to the minimum covering convex polygon and calculate its vectorized representation to obtain the processing result.

[0021] Through the above technical solutions, the embodiments of the present application can achieve unified conversion for two common types of perception data, namely raster maps and vectorized bounding boxes, through a series of operations. Filter the raster map to remove information that coincides with the vectorized bounding box, effectively improving the perception conversion efficiency and avoiding data redundancy. Cluster the remaining occupied grids to accurately identify individual abnormal obstacles, providing a clear classification basis for subsequent processing. Solve the minimum covering convex polygon for each cluster, and further solve the minimum covering rectangle to obtain the vectorized representation, so that the heterogeneous raster map and vectorized bounding box data are finally presented in a unified format. This unified format of data is easier to be understood and processed by the autonomous driving system, enhancing the system's perception ability of obstacles in complex traffic scenarios, and providing a solid and reliable data basis for subsequent safety decision control correction based on these data.

[0022] Optionally, in an embodiment of the present application, the correction module includes: a generation unit, configured to output the finally considered safety-corrected control quantity as acceleration and steering angle to generate a control instruction; a sending unit, configured to send the control instruction to the controller terminal of the host vehicle.

[0023] Through the above technical solutions, the embodiments of the present application can directly convert the output of the optimized control quantity into acceleration and steering angle to generate a control instruction, and send it to the controller terminal of the host vehicle. This method makes the execution link of safety control correction simple and efficient, and can quickly convert safety decisions into actual vehicle control instructions. Directly using acceleration and steering angle as the core of the control instruction, accurately acting on the power and steering systems of the vehicle, ensuring that the vehicle responds in a timely manner and adjusts its driving state, effectively avoiding danger and ensuring driving safety.

[0024] Optionally, in an embodiment of the present application, the correction module includes: The optimization problem paradigm is:

[0025]

[0026] s.t.L f h Fi +L g h Fi y+β i h Fi ≥0

[0027] L f h rj +L g h rj u+γ i h rj ≥0

[0028] wherein, u is the control input; u ois the received initial decision control result; L f is the partial derivative of the barrier function h with respect to g(x); L g is the partial derivative of the barrier function h with respect to f(x); Q is a constant matrix; h Fi is the constraint corresponding to obstacle i; h rj is the constraint corresponding to road boundary j; β i and γ i are constants.

[0029] Through the above technical solution, the embodiment of the present application can make the corrected control amount as close as possible to the initial decision control result on the premise of meeting safety constraints, ensuring both safety and maximizing the retention of the original decision-making intention, and avoiding adverse effects on vehicle driving stability and comfort caused by excessive correction.

[0030] The embodiment of the third aspect of the present application provides a vehicle, including: a memory, a processor, and a computer program stored on the memory and executable on the processor, and the processor executes the program to implement the CBF-based autonomous driving generalization safety decision control correction method as described in the above embodiment.

[0031] The embodiment of the fourth aspect of the present application provides a computer-readable storage medium, and the computer-readable storage medium stores a computer program, and when the program is executed by a processor, it implements the above CBF-based autonomous driving generalization safety decision control correction method.

[0032] The embodiment of the fifth aspect of the present application provides a computer program product, including a computer program, characterized in that the computer program is executed to be used to implement the above CBF-based autonomous driving generalization safety decision control correction method.

[0033] The embodiment of the present application can obtain various types of perception information of the traffic scene and the host vehicle, comprehensively master the driving environment and its own state, and provide a basis for decision-making. For common grid maps and vectorized bounding box data, through operations such as filtering, clustering, solving convex polygons and rectangles, etc., the unified conversion of heterogeneous data is realized, the perception efficiency is improved, and the perception ability of obstacles is enhanced, laying a solid data foundation for subsequent processing. Then, various models are constructed based on the processing results and the initial decision control results, comprehensively considering traffic risk sources, generating an optimization problem paradigm to solve the output of the safety-corrected control amount, and effectively avoiding dangerous behaviors. Finally, the output of the control amount is converted into acceleration and steering angle to generate control instructions and sent to the host vehicle controller terminal, which is executed simply and efficiently, accurately controls the vehicle, and while ensuring safety, retains the original decision-making intention, avoiding excessive correction from affecting driving stability and comfort, and strongly promoting the development of autonomous driving technology.

[0034] Additional aspects and advantages of the present application will be given in part in the following description, become apparent in part from the following description, or be learned through the practice of the present application. Description of the Drawings

[0035] The above-mentioned and / or additional aspects and advantages of the present application will become apparent and be readily understood from the following description of embodiments in conjunction with the drawings, where:

[0036] Figure 1 is a flowchart of a method for correcting autonomous driving generalization safety decision control based on CBF according to an embodiment of the present application;

[0037] Figure 2 is a schematic diagram of a vectorized bounding box perception failure case according to an embodiment of the present application;

[0038] Figure 3 is an example diagram of an occupancy grid according to an embodiment of the present application;

[0039] Figure 4 is a schematic diagram of filtering the grid map perception result according to an embodiment of the present application;

[0040] Figure 5 is a schematic diagram of occupancy grid clustering according to an embodiment of the present application;

[0041] Figure 6 is a schematic diagram of solving the minimum covering convex polygon according to an embodiment of the present application;

[0042] Figure 7 is a schematic diagram of the vectorized representation of solving the minimum covering rectangle according to an embodiment of the present application;

[0043] Figure 8 is a schematic diagram of the effect without safety correction on the CARLA simulation platform according to a specific embodiment of the present application;

[0044] Figure 9 is a schematic diagram of the effect on the CARLA simulation platform considering only obstacle constraints according to a specific embodiment of the present application;

[0045] Figure 10 is a schematic diagram of the effect of adopting the method for correcting autonomous driving generalization safety decision control based on CBF on the CARLA simulation platform according to a specific embodiment of the present application;

[0046] Figure 11 is a schematic diagram of the effect of the physical sand table without using safety correction according to a specific embodiment of the present application;

[0047] Figure 12 is a schematic diagram of the effect of the physical sand table considering only obstacle constraints according to a specific embodiment of the present application;

[0048] Figure 13 Schematic diagram of the effect of a physical sand table adopting a CBF-based autonomous driving generalization safety decision control correction method according to a specific embodiment of the present application;

[0049] Figure 14 Schematic diagram of the structure of a CBF-based autonomous driving generalization safety decision control correction device provided according to an embodiment of the present application;

[0050] Figure 15 Schematic diagram of the structure of a vehicle provided according to an embodiment of the present application. Detailed implementation manners

[0051] The embodiments of the present application will be described in detail below. The examples of the embodiments are shown in the accompanying drawings, where the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to explain the present application, and should not be construed as a limitation to the present application.

[0052] The CBF-based autonomous driving generalization safety decision control correction method and device according to the embodiments of the present application will be described below with reference to the accompanying drawings. In the related technologies mentioned in the above background art, the method based on simple rules cannot strictly guarantee safety and lacks generalization ability. Although the CBF-based method has mathematical completeness, there are problems of incompatible heterogeneous perception input formats and poor adaptability to complex scenarios. The present application provides a CBF-based autonomous driving generalization safety decision control correction method. In this method, various types of perception information of the traffic scene and the host vehicle can be obtained to comprehensively master the vehicle driving environment and its own state, providing rich basis for decision-making. These information are uniformly transformed and processed to solve the problem of incompatible heterogeneous data formats, improve the generalization of the decision correction module, and make subsequent processing more efficient and accurate. Based on the processing results and the initial decision control results, a vehicle dynamics model, an obstacle constraint model, and a road boundary constraint model of the host vehicle are constructed, comprehensively and quantitatively considering various traffic risk sources, and comprehensively improving the ability to guarantee driving safety. An optimization problem paradigm for control correction is generated and the final safely corrected control quantity output is obtained, effectively avoiding dangerous behaviors, enhancing the safety and reliability of autonomous driving decision control, ensuring the vehicle to drive safely in complex and changeable traffic scenarios, and promoting the development of autonomous driving technology. Thus, the problems in the related technologies that the method based on simple rules cannot strictly guarantee safety and lacks generalization ability, and although the CBF-based method has mathematical completeness, there are problems of incompatible heterogeneous perception input formats and poor adaptability to complex scenarios are solved.

[0053] Specifically, Figure 1Schematic flow diagram of a CBF-based generalized safety decision control correction method provided by an embodiment of the present application.

[0054] As Figure 1 shown, the CBF-based generalized safety decision control correction method includes the following steps:

[0055] In step S101, various types of perception information related to the traffic scenario and the host vehicle are acquired.

[0056] It can be understood that the traffic scenario includes, but is not limited to, basic road topology information, such as lane width, number of lanes, lane boundaries, etc., and the lane boundaries are represented by line segments connected end to end; physical information of other surrounding intelligent agents, including but not limited to the current position, speed, heading angle, length and width, etc. of the intelligent agents. Various types of perception information related to the host vehicle include but are not limited to the position, speed, acceleration, length and width, etc. of the host vehicle.

[0057] The embodiment of the present application can acquire comprehensive and key data, providing data support for the subsequent autonomous driving system to make more scientific and accurate decisions in a complex traffic environment.

[0058] In step S102, the various types of perception information are uniformly transformed and processed to obtain a processing result.

[0059] It should be noted that the heterogeneous perception result format is very important for enhancing the practicability of the safety correction module. Currently, the autonomous driving perception output can be mainly divided into two categories: VBB (Vectorized Bounding Box) and OGM (occupancy grid map), and the specific output format depends on the sensors used for data collection.

[0060] On the one hand, for example, the perception algorithm based on a camera mainly generates a vectorized bounding box, that is, the obstacle is represented by a vector, as shown in formula (1), and formula (1) is as follows:

[0061] x obs = [[x 1 y 1 l 1 w 1 θ 1 ,…](1)

[0062] Where, (x i , y i ) is the center coordinate of the i-th obstacle; l i is the length; w i is the width; θ i is the orientation angle.

[0063] Specifically, one problem with vision-based perception output is that the defined bounding box is a standard cube, making it difficult to represent irregularly shaped obstacles, such as Figure 2 as shown Figure 2 is a failure example where the bounding box cannot represent irregularly shaped obstacles. In this scenario, a parked car has its door open, but the perception method can still only recognize the main body of the car.

[0064] On the other hand, one type of perception data is OGM. Perception systems using millimeter-wave radar or lidar as sensors mainly use OGM as the perception output data format. OGM uses a two-dimensional or three-dimensional grid map to represent the real world. The map consists of several small squares. If there is an obstacle in a square, its state is set to "occupied"; otherwise, it is marked as "free". As Figure 3 shown Figure 3 gives an example of a grid map.

[0065] In the actual process, unified conversion processing is performed on multiple types of perception information to obtain a processing result, including: filtering the grid map and clustering the remaining occupied grids to obtain a clustering result; for each cluster in the clustering result, solving the minimum covering convex polygon and then solving the minimum covering rectangle based on the minimum covering convex polygon and calculating its vectorized representation to obtain the processing result.

[0066] Specifically, although the vectorized bounding box and the occupied grid each have their own advantages, most of the perception results they feedback overlap. Therefore, to improve the perception conversion efficiency, first, in the grid map, filter out the obstacles that have been perceived by the vectorized bounding box, and select the obstacle grids not included in the vectorized bounding box. As Figure 4 shown, set the occupied grids within the rectangular area represented by the vectorized bounding box to free, and the remaining occupied grids are the obtained filtering result.

[0067] Furthermore, the selected grids are the irregularly shaped obstacles that the vectorized bounding box fails to perceive. Classify these obstacles to identify which independent individuals they belong to respectively. The genetic clustering algorithm can be used. Set the maximum adjacent interval d, then all occupied grids with an interval distance less than d will be identified as one class, as Figure 5 shown.

[0068] After obtaining the clustering result, for each cluster, it should represent a connected obstacle. Therefore, the Graham scan algorithm can be used to solve the minimum convex polygon for each cluster, as Figure 6 shown.

[0069] Finally, after obtaining the minimum covering convex polygon of each obstacle cluster, the minimum covering rectangle of each convex polygon can be further calculated. As shown in Figure 7 shown, it is then expressed as a five-tuple as shown in Equation (1), and then merged with the original vectorized bounding box perception result, so that the two types of heterogeneous perception data are expressed in a unified manner and can be understood by the CBF system.

[0070] Through the above steps, the two different perception output formats of VBB and OGM can be adjusted to a unified format, so that the subsequent control correction module can comprehensively consider various traffic constraints.

[0071] The embodiment of the present application can perform unified conversion processing on the two heterogeneous perception data formats of VBB and OGM in autonomous driving. By filtering the grid map and removing the perception information that overlaps with VBB, the perception conversion efficiency is greatly improved, and data redundancy is avoided. Using the genetic clustering algorithm to cluster the remaining grids can accurately identify individual irregular obstacles and provide a clear classification for subsequent processing. The Graham scan algorithm is used to solve the minimum covering convex polygon, and then the vectorized representation of the minimum covering rectangle is further obtained, so that the two types of data are finally unified into a format that can be understood by the CBF system. This unified format allows the subsequent control correction module to comprehensively consider various traffic constraints, effectively improving the generalization of the autonomous driving decision correction module and enhancing the system's adaptability to complex traffic scenarios, laying a solid data foundation for the autonomous driving system to accurately perceive the environment and make safe decisions.

[0072] In step S103, based on the processing result and the initial decision control result, a dynamic model, an obstacle constraint model, and a road boundary constraint model of the ego vehicle are constructed to generate an optimization problem paradigm for control correction, and the control quantity output after finally considering safety correction is obtained by using the optimization problem paradigm for control correction to perform safety control correction.

[0073] Specifically, based on the CBF system, the present application uniformly models various traffic scenario safety constraints as obstacle functions, and then solves the control correction amount based on the optimization algorithm. In addition to receiving the unified perception result of step S102, it is also necessary to receive the initial decision control result of the autonomous driving system. The present application has no specific requirements for the decision control module of the ego vehicle. Due to its generalization performance, the control quantity output by any decision and control algorithm can be guaranteed the safety of autonomous driving control after being corrected by the present application.

[0074] It should be noted that the safety constraints of traffic scenarios can be roughly divided into two categories. One is the obstacle constraints brought about by the movement of agents, and the other is the road boundary constraints brought about by the road topology. Before modeling the two types of constraints, it is first necessary to construct a dynamic model of the ego vehicle so that the constraints can interact with the ego vehicle.

[0075] The formula used in this application for modeling the dynamics of the host vehicle is as follows:

[0076]

[0077] where is the state vector; u = [atan(δ f )] is the control input; x g is the abscissa of the center of mass of the host vehicle; y g is the ordinate of the center of mass of the host vehicle; v is the longitudinal vehicle speed; a is the longitudinal acceleration; is the orientation angle of the host vehicle; δ f is the front wheel steering angle; L is the wheelbase of the vehicle.

[0078] After the dynamic model of the host vehicle is constructed, the obstacle constraint is modeled. It can be understood that one of the cores of the CBF theory system is to construct an obstacle function. After the obstacle function is constructed, the corresponding optimization problem inequality constraint can be obtained according to its mathematical definition. The obstacle function can be understood as the most primitive safety condition, which is the definition of "safe driving". When the function value is greater than or equal to 0, the current driving state is necessarily safe. The formula of the obstacle function adopted in this application is as follows:

[0079]

[0080] where d lon is the lateral distance of the host vehicle from the risk source; d lat is the longitudinal distance of the host vehicle from the risk source; l lon is the lateral coefficient; l lat is the longitudinal coefficient; c safe is the safety distance constant.

[0081] After obtaining the definition of the obstacle function, the obstacle constraint inequality corresponding to the optimization problem can be obtained by following the mathematical formula definition of D-CBF. The formula is as follows:

[0082]

[0083] where L g is the partial derivative of the obstacle function h with respect to f(x); L f is the partial derivative of the obstacle function h with respect to g(x); x obs is the state vector obtained by the perception result conversion layer; α 1 and α 2 are constants.

[0084] Furthermore, in addition to moving obstacles, road boundaries are also one of the traffic risk sources. Existing CBF collision avoidance systems often do not model road boundaries, resulting in the ego vehicle driving into an impassable area instead of bypassing the obstacle during collision avoidance, generating secondary risks. Therefore, in this application, the same theoretical system is also used to model and constrain road boundaries. Considering Frenet coordinate system modeling, the formula is as follows:

[0085]

[0086] where s is the lateral coordinate of a certain position in the Frenet coordinate system; d is the longitudinal coordinate of a certain position in the Frenet coordinate system; μ is the angle between the ego vehicle's heading angle and the road orientation; τ is the curvature of the road; δ f is the steering angle of the ego vehicle's front wheels.

[0087] Under the above road modeling, the road boundary constraint can be expressed as:

[0088]

[0089] where d 0 and d 1 are the directed distances from the road center to the two side boundaries of the road (taking the road orientation as the positive direction, right is positive and left is negative). Generally, the sum of the two distances is 0.

[0090] Optionally, in an embodiment of this application, the final control quantity output after considering safety correction is obtained by using the optimization problem paradigm of control correction for safety control correction, including: taking the final control quantity output after considering safety correction as the acceleration and the steering angle to generate a control instruction; sending the control instruction to the controller terminal of the ego vehicle.

[0091] After completing the modeling of obstacles and road boundaries, the optimization problem paradigm of control correction can be constructed. The optimization problem paradigm is:

[0092]

[0093] where u o is the received initial decision control result; Q is a constant matrix; h Fi is the constraint corresponding to obstacle i, defined as shown in (4); h rj is the constraint corresponding to road boundary j, defined as shown in (6). The final control quantity output after considering safety correction can be obtained from formula (7).

[0094] Furthermore, the calculated corrected control quantity is used as the acceleration and the steering angle and sent to the controller terminal of the ego vehicle to complete the safety control correction.

[0095] The embodiments of the present application can construct various models based on the uniformly processed perception results and initial decision control results, comprehensively considering various safety constraints in the traffic scenario. By constructing the ego-vehicle dynamics model, the relationship between the vehicle's own motion state and control input is clarified, providing a basis for subsequent constraint modeling and control correction. The obstacles and road boundaries are modeled separately, especially making up for the lack of modeling of road boundaries in the traditional CBF collision avoidance system, avoiding the vehicle from driving into dangerous areas during the collision avoidance process, and effectively reducing the secondary risk. The obstacle function and the inequality constraints of the optimization problem are constructed using the CBF theory, strictly defining the safe driving conditions from a mathematical level, ensuring that as long as the constraints are met, the driving state is safe, and enhancing the theoretical basis and reliability of safety. The constructed control correction optimization problem paradigm, while ensuring safety, makes the corrected control quantity as close as possible to the initial decision control result, taking into account both driving stability and comfort, and avoiding negative impacts on vehicle performance caused by excessive correction. This technical solution has strong generalization ability, has no specific requirements for the decision control module of the ego-vehicle, and can adapt to the control quantity output by any decision and control algorithm, broadening its application scope. Finally, the way of converting the corrected control quantity into acceleration and steering angle and sending it to the ego-vehicle controller terminal is simple, direct, and efficient.

[0096] To verify the effectiveness of the present application, the CBF-based autonomous driving generalization safety decision control correction method proposed in the present application can be tested in CARLA (Car Learning to Act) and the physical platform, and the application example effects are as follows.

[0097] Specifically, in the CARLA simulation platform, in the face of the sudden dangerous cut-in of a lateral vehicle, when no safety control correction is applied, the result is as Figure 8 shown. The ego-vehicle (Vehicle A) cannot respond in time according to the real-time state, resulting in a collision accident; when only the obstacle constraint is considered, the effect is as Figure 9 shown. Although it can avoid danger to the left, since the road boundary constraint is not considered, it will drive onto the curb, thus triggering a new safety risk; when the safety decision correction method proposed in the present application is applied, the effect is as Figure 10 shown. While being able to safely avoid the lateral vehicle, it can effectively control the lateral displacement to avoid driving into dangerous areas.

[0098] The present application can also be verified in the physical sand table platform, such as Figure 11 、 12, as shown in Fig. 13, they respectively represent not using safety correction, only considering obstacle constraints, and the method proposed in this application. CAV represents the host vehicle, and HDV represents the interfering vehicle. Through the application example of the physical platform, it can be found that the CBF-based generalized safety decision control correction method for autonomous driving proposed in this application can effectively perform safety control correction on the sudden risks in the traffic scenario and avoid accidents.

[0099] According to the CBF-based generalized safety decision control correction method proposed in the embodiments of this application, by obtaining various types of perception information of the traffic scenario and the host vehicle, the driving environment of the vehicle and its own state can be comprehensively grasped, providing rich basis for decision-making. Unified transformation processing is performed on this information to solve the problem of incompatible heterogeneous data formats, improve the generalization of the decision correction module, and make subsequent processing more efficient and accurate. Based on the processing results and the initial decision control results, a dynamic model of the host vehicle, an obstacle constraint model, and a road boundary constraint model are constructed, comprehensively quantifying and considering various traffic risk sources, and comprehensively improving the ability to ensure driving safety. Generate an optimization problem paradigm for control correction and obtain the final control quantity output after safety correction, effectively avoiding dangerous behaviors, enhancing the safety and reliability of autonomous driving decision control, ensuring the vehicle to drive safely in complex and changeable traffic scenarios, and promoting the development of autonomous driving technology.

[0100] Next, refer to the drawings to describe the CBF-based generalized safety decision control correction device proposed in the embodiments of this application.

[0101] Figure 14 It is a block diagram of the CBF-based generalized safety decision control correction device according to the embodiments of this application.

[0102] As Figure 14 shown, the CBF-based generalized safety decision control correction device 10 includes: an acquisition module 100, a conversion module 200, and a correction module 300.

[0103] Specifically, the acquisition module 100 is used to acquire various types of perception information related to the traffic scenario and the host vehicle.

[0104] The conversion module 200 is used to perform unified conversion processing on various types of perception information to obtain a processing result.

[0105] The correction module 300 is used to construct a dynamic model of the host vehicle, an obstacle constraint model, and a road boundary constraint model based on the processing result and the initial decision control result, to generate an optimization problem paradigm for control correction, and use the optimization problem paradigm for control correction to obtain the final control quantity output after considering safety correction, so as to perform safety control correction.

[0106] Optionally, in an embodiment of the present application, the multi-class perception information includes a grid map and a vectorized bounding box. Among them, the conversion module 200 includes a filtering clustering unit and a solving unit.

[0107] Among them, the filtering clustering unit is used to filter the grid map and cluster the remaining occupied grids to obtain a clustering result.

[0108] The solving unit is used to solve the minimum covering convex polygon for each cluster of the clustering result, and solve the minimum covering rectangle according to the minimum covering convex polygon and calculate its vectorized representation to obtain a processing result.

[0109] Optionally, in an embodiment of the present application, the correction module 300 includes a generation unit and a sending unit.

[0110] The generation unit is used to output the control amount after finally considering safety correction as acceleration and rotation angle, and generate a control instruction.

[0111] The sending unit is used to send the control instruction to the controller terminal of the host vehicle.

[0112] Optionally, in an embodiment of the present application, the correction module 300 includes an optimization problem paradigm as follows:

[0113]

[0114] s.t.L f h Fi +L g h Fi u+β i h Fi ≥0

[0115] L f h rj +L g h rj u+γ i h rj ≥0

[0116] Among them, u is the control input; u o is the received initial decision control result; L f is the partial derivative of the obstacle function h with respect to g(x); L g is the partial derivative of the obstacle function h with respect to f(x); Q is a constant matrix; h Fi is the constraint corresponding to the obstacle i; h rj is the constraint corresponding to the road boundary j; β i and γ i are constants.

[0117] It should be noted that the foregoing explanatory description of the embodiments of the CBF-based autonomous driving generalization safety decision control correction method is also applicable to the CBF-based autonomous driving generalization safety decision control correction device of this embodiment, and will not be elaborated here.

[0118] The CBF-based autonomous driving generalization safety decision control correction device proposed according to the embodiments of the present application can comprehensively master the vehicle driving environment and its own state by acquiring traffic scenarios and various types of vehicle perception information, providing rich bases for decision-making. These information are uniformly transformed and processed to solve the problem of incompatible heterogeneous data formats, improve the generalization of the decision correction module, and make subsequent processing more efficient and accurate. Based on the processing results and the initial decision control results, a vehicle dynamics model, an obstacle constraint model, and a road boundary constraint model are constructed, comprehensively and quantitatively considering various traffic risk sources, and comprehensively improving the ability to ensure driving safety. An optimization problem paradigm for control correction is generated and the final safely corrected control quantity is output, effectively avoiding dangerous behaviors, enhancing the safety and reliability of autonomous driving decision control, ensuring the vehicle to drive safely in complex and changeable traffic scenarios, and promoting the development of autonomous driving technology.

[0119] Figure 15 The structure diagram of the vehicle provided by the embodiments of the present application. The vehicle may include:

[0120] A memory 1501, a processor 1502, and a computer program stored on the memory 1501 and executable on the processor 1502.

[0121] When the processor 1502 executes the program, it implements the CBF-based autonomous driving generalization safety decision control correction method provided in the above embodiments.

[0122] Further, the vehicle further includes:

[0123] A communication interface 1503 for communication between the memory 1501 and the processor 1502.

[0124] The memory 1501 is used to store a computer program executable on the processor 1502.

[0125] The memory 1501 may include a high-speed RAM memory, and may also include a non-volatile memory, such as at least one disk memory.

[0126] If the memory 1501, the processor 1502, and the communication interface 1503 are implemented independently, the communication interface 1503, the memory 1501, and the processor 1502 can be interconnected through a bus and communicate with each other. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, an Extended Industry Standard Architecture (EISA) bus, or the like. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 15 only a thick line is used to represent it in Figure 15 , but it does not mean that there is only one bus or one type of bus.

[0127] Optionally, in a specific implementation, if the memory 1501, the processor 1502, and the communication interface 1503 are integrated on a single chip, the memory 1501, the processor 1502, and the communication interface 1503 can communicate with each other through an internal interface.

[0128] The processor 1502 may be a Central Processing Unit (CPU), or an Application Specific Integrated Circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present application.

[0129] The embodiments of the present application also provide a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, the above-mentioned CBF-based autonomous driving generalization safety decision control correction method is implemented.

[0130] The embodiments of the present application also provide a computer program product, including a computer program, characterized in that the computer program is executed to implement the above-mentioned CBF-based autonomous driving generalization safety decision control correction method.

[0131] In the description of this specification, the descriptions referring to terms such as "one embodiment", "some embodiments", "examples", "specific examples", or "some examples", etc. mean that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of this application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in any one or N embodiments or examples in a suitable manner. In addition, without contradiction, those skilled in the art can combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples.

[0132] In addition, the terms "first" and "second" are used only for descriptive purposes and cannot be construed as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include at least one of such features. In the description of this application, the meaning of "N" is at least two, such as two, three, etc., unless otherwise specifically defined.

[0133] Any process or method description shown in the flowchart or described in other ways herein can be understood as representing a module, segment, or portion of code including one or N executable instructions for implementing a customized logic function or process, and the scope of the preferred embodiments of this application includes additional implementations, where the functions can be executed in a substantially simultaneous manner or in the reverse order according to the functions involved, rather than in the order shown or discussed, which should be understood by those skilled in the art to which the embodiments of this application pertain.

[0134] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a definite sequence list of executable instructions for implementing logical functions, and can be specifically implemented in any computer-readable medium for use by an instruction execution system, apparatus, or device (such as a computer-based system, a system including a processor, or other systems that can fetch and execute instructions from the instruction execution system, apparatus, or device), or used in combination with these instruction execution systems, apparatus, or devices. For the purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device. More specific examples (non-exhaustive list) of computer-readable media include the following: an electrical connection part (electronic device) having one or N wirings, a portable computer disk cartridge (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). Additionally, the computer-readable medium can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically by optically scanning the paper or other media, followed by editing, interpretation, or other suitable processing as necessary, and then stored in a computer memory.

[0135] It should be understood that various parts of the present application can be implemented by hardware, software, firmware, or a combination thereof. In the above-described embodiments, the N steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. If implemented in hardware, as in another embodiment, any one or a combination of the following techniques well known in the art can be used: discrete logic circuits having logic gate circuits for implementing logical functions on data signals, application specific integrated circuits having appropriate combinational logic gate circuits, programmable gate arrays (PGAs), field programmable gate arrays (FPGAs), etc.

[0136] Those of ordinary skill in the art of this technology can understand that all or part of the steps carried by the method of the above-described embodiments can be completed by instructing relevant hardware through a program, and the program can be stored in a computer-readable storage medium. When the program is executed, it includes one or a combination of the steps of the method embodiments.

[0137] In addition, each functional unit in various embodiments of the present application may be integrated into one processing module, or each unit may exist physically alone, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. When the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0138] The above-mentioned storage medium may be a read-only memory, a magnetic disk or an optical disc, etc. Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and should not be construed as limiting the present application. Those of ordinary skill in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present application.

Claims

1. A generalized safety decision control correction method for autonomous driving based on CBF, characterized in that: The following steps are involved: Obtain multiple types of perception information related to traffic scenes and the vehicle itself; Performing unified transformation processing on the multiple types of perception information to obtain processing results; Based on the processing results and the initial decision control results, a dynamic model of the vehicle, an obstacle constraint model, and a road boundary constraint model are constructed to generate an optimization problem paradigm for control correction, and the optimization problem paradigm for control correction is used to obtain the final control quantity output after considering the safety correction to perform safety control correction.

2. The method according to claim 1, characterized in that The multiple types of perception information include a raster map and a vectorized bounding box, wherein the multiple types of perception information are uniformly converted to obtain a processing result, including: Filtering the grid map and clustering the remaining occupied grids to obtain a clustering result; For each cluster of the clustering result, a minimum covering convex polygon is solved, and based on the minimum covering convex polygon, a minimum covering rectangle is solved and its vectorized representation is calculated to obtain the processing result.

3. The method according to claim 1, characterized in that The method of using the optimization problem paradigm of the control correction to obtain the final control quantity output after considering the safety correction to perform the safety control correction includes: The control quantity output after the final safety correction is taken into consideration is used as the acceleration and the rotation angle to generate a control instruction; The control instruction is sent to the controller terminal of the vehicle.

4. The method according to any one of claims 1 to 3, characterized in that: The optimization problem paradigm is: s.t.L f h Fi +L g h Fi u+β i h Fi ≥0 L f h rj +L g h rj u+γ i h rj ≥0 Where u is the control input; u o is the received initial decision control result; L f is the partial derivative of the barrier function h with respect to g(x); L g is the partial derivative of the barrier function h with respect to f(x); Q is a constant matrix; h Fi is the constraint corresponding to obstacle i; h rj is the constraint corresponding to the road boundary j; β i With γ i is a constant.

5. A CBF-based generalized safety decision control correction device for autonomous driving, characterized in that: include: The acquisition module is used to obtain multiple types of perception information related to traffic scenes and the vehicle itself; A conversion module, used for uniformly converting the multiple types of perception information to obtain a processing result; A correction module is used to construct a dynamic model, an obstacle constraint model, and a road boundary constraint model of the vehicle based on the processing results and the initial decision control results to generate an optimization problem paradigm for control correction, and use the optimization problem paradigm for control correction to obtain the final control quantity output after considering the safety correction to perform safety control correction.

6. The device according to claim 5, characterized in that The multiple types of perception information include a raster map and a vectorized bounding box, wherein the conversion module includes: A filtering and clustering unit, used for filtering the grid map and clustering the remaining occupied grids to obtain a clustering result; The solving unit is used to solve the minimum covering convex polygon for each cluster of the clustering result, and solve the minimum covering rectangle based on the minimum covering convex polygon and calculate its vectorized representation to obtain the processing result.

7. The device according to claim 5, characterized in that The correction module comprises: A generating unit, configured to use the control quantity output after the final safety correction as acceleration and angle to generate a control instruction; A sending unit is used to send the control instruction to the controller terminal of the vehicle.

8. A vehicle, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the CBF-based generalized safety decision control correction method for autonomous driving as described in any one of claims 1 to 4.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: The program is executed by a processor to implement the CBF-based generalized safety decision control correction method for autonomous driving as described in any one of claims 1 to 4.

10. A computer program product, comprising a computer program, characterized in that The computer program is executed to implement the CBF-based generalized safety decision control correction method for autonomous driving as described in any one of claims 1 to 4.

Images