Automatic driving lane changing decision-making method, electronic equipment and vehicle

Abstract

The embodiment of the invention provides an automatic driving lane changing decision-making method, electronic equipment and a vehicle. The method comprises the following steps: acquiring a first path time diagram of a current lane and a second path time diagram of a target lane; an upper bound curve and a lower bound curve of each lane are obtained through acceleration constraint of the automatic driving vehicle; determining a first drivable convex space of the current lane according to the first path time diagram and the upper and lower boundary curves of the current lane, determining a second drivable convex space of the target lane according to the second path time diagram and the upper and lower boundary curves of the target lane, and generating a candidate lane changing longitudinal convex space together with the first drivable convex space and the second drivable convex space; a lane changing speed curve is solved through speed optimization, a lane changing decision result is determined according to the cost value of the lane changing speed curve and comprises at least one of the lane changing opportunity, the lane changing speed, the lane changing mark and the cancel mark, and therefore a safe and real-time lane changing decision conforming to the driving intention is achieved, the lane changing success rate is increased, and the safety of driving operation is ensured.

Description

Technical Field

[0001] This application relates to the field of autonomous driving technology in intelligent transportation, and more particularly to an autonomous driving lane change decision-making method, electronic equipment, and vehicle. Background Technology

[0002] Autonomous driving lane-changing decisions impact driving comfort and efficiency, and can also prevent traffic accidents and congestion in complex traffic environments. However, in complex scenarios, autonomous driving lane-changing decision-making methods first determine the timing of the lane change and then plan the trajectory, leading to a disconnect between decision-making and planning, which cannot guarantee completeness and reduces the success rate and safety of lane changes. Summary of the Invention

[0003] This application provides an autonomous driving lane change decision-making method, electronic device, and vehicle, aiming to improve the problem in related technologies where autonomous driving lane change decision-making methods first determine the timing of lane change and then plan the trajectory, resulting in a disconnect between decision-making and planning that cannot guarantee completeness, and reduces the success rate and safety of lane changes.

[0004] Firstly, an autonomous driving lane-change decision-making method is provided, comprising: acquiring a first path time map of the current lane and a second path time map of the target lane; obtaining an upper bound curve and a lower bound curve of each lane based on the acceleration constraints of the autonomous vehicle; determining a first drivable convex space of the current lane based on the first path time map, the upper bound curve and the lower bound curve of the current lane, and determining a second drivable convex space of the target lane based on the second path time map, the upper bound curve and the lower bound curve of the target lane; generating candidate lane-change longitudinal convex spaces based on the first drivable convex spaces and the second drivable convex spaces; obtaining a lane-change speed curve by speed optimization solution based on the candidate lane-change longitudinal convex spaces; and determining the lane-change decision result of the autonomous vehicle based on the cost of the lane-change speed curve. The lane-change decision result includes at least one of lane-change timing, lane-change speed, lane-change flag, and cancellation flag, wherein the lane-change flag indicates whether the autonomous vehicle is allowed to change lanes, and the cancellation flag indicates that the autonomous vehicle has cancelled the lane-change.

[0005] Through the above technical solutions, this embodiment of the invention achieves accurate modeling of the spatiotemporal distribution of dynamic obstacles in both lanes by obtaining the first path time map of the current lane and the second path time map of the target lane, thereby improving the accuracy of lane change decisions; it obtains the upper and lower bound curves of each lane based on the acceleration constraints of the autonomous vehicle, ensuring that the generated motion boundary conforms to the feasibility of vehicle dynamics and improving driving safety; it determines the first drivable convex space based on the first path time map and the upper and lower bound curves of the current lane, and determines the second drivable convex space based on the second path time map and the upper and lower bound curves of the target lane, transforming the non-convex feasible region into a structured convex set to support... It features efficient search and optimized computational efficiency; it generates candidate longitudinal convex spaces for lane changing based on two drivable convex spaces, effectively filtering out feasible longitudinal trajectories that meet the lane-changing time window, enhancing the flexibility of lane-changing timing selection; it performs speed optimization solution based on the candidate longitudinal convex spaces to obtain the lane-changing speed curve, and determines the lane-changing decision result based on its cost value, including at least one of the following: lane-changing timing, lane-changing speed, lane-changing indicator, and cancellation indicator. The lane-changing indicator represents whether lane changing is allowed, and the cancellation indicator represents canceling lane changing, thereby achieving safe, real-time, and driver-intent-compliant lane-changing decisions, improving the success rate of lane changing, and ensuring the rationality and safety of driving operations.

[0006] In conjunction with the first aspect, in some possible implementations, obtaining a first path time map of the current lane and a second path time map of the target lane includes: obtaining a first reference line of the current lane and a second reference line of the target lane; obtaining the size and lateral safety threshold of the autonomous vehicle, and scanning along the first and second reference lines respectively in combination with the size and lateral safety threshold; generating a first path time map based on the scanning result of the first reference line; and generating a second path time map based on the scanning result of the second reference line.

[0007] Through the above technical solution, the embodiments of the present invention obtain a first reference line of the current lane and a second reference line of the target lane, and scan along the two reference lines in combination with the size of the autonomous vehicle and the lateral safety threshold to generate a first path time map and a second path time map. This ensures that the actual physical size and dynamic behavior characteristics of the vehicle are fully considered during the lane change decision process, and can accurately depict the drivable area of ​​the vehicle in the current lane and the target lane, improving the accuracy and reliability of lane change decisions. The use of the lateral safety threshold further enhances the safety of lane change operations and avoids potential collision risks. Based on these path time maps, the timing of lane changes can be more accurately identified and the optimal speed curve can be planned, thereby improving the success rate of lane changes and driving safety, while ensuring that the entire process complies with vehicle dynamics constraints.

[0008] In conjunction with the first aspect, in some possible implementations, the upper and lower bound curves of each lane are obtained based on the acceleration constraints of the autonomous vehicle, including: identifying the maximum acceleration and maximum deceleration in the acceleration constraints of the autonomous vehicle; using the maximum planning time of the autonomous vehicle as a constraint, generating the upper bound curves of the current lane and the target lane based on the maximum acceleration and the actual planning time of the autonomous vehicle; using the maximum planning time of the autonomous vehicle as a constraint, generating the lower bound curve of the current lane based on the maximum deceleration and the actual planning time of the autonomous vehicle, and obtaining the pre-set lower bound curve of the target lane.

[0009] Through the above technical solution, the embodiments of the present invention can identify the maximum acceleration and maximum deceleration of autonomous vehicles and generate upper and lower bound curves for each lane based on the maximum planning time, ensuring that the dynamic characteristics of the vehicle are fully considered during the lane-changing decision-making process. Accurately determining the vehicle's motion limits in the current lane improves the safety and reliability of path planning. The upper and lower bound curves provide clear speed boundaries for the vehicle, avoiding risks caused by speeding or insufficient deceleration. This enhances the safety of lane-changing operations and ensures consistency between the decision and the vehicle's physical limitations, improving the success rate of lane changes and driving comfort.

[0010] In conjunction with the first aspect, in some possible implementations, determining the first drivable convex space of the current lane based on the first path time map, the upper bound curve and the lower bound curve of the current lane includes: determining the slice range of the first path time map based on the upper bound curve and the lower bound curve of the current lane; calculating the first drivable slice at each moment on the first path time map within the slice range, and establishing directed edges of the first drivable slice at adjacent moments; constructing a first multi-branch tree structure based on the directed edges of the first drivable slice overlapping in the target direction, and obtaining the first drivable convex space over the entire time domain by performing a depth-first search on the first multi-branch tree structure.

[0011] Through the above technical solution, this embodiment of the invention determines the slice range of the first path time map by using the upper and lower bound curves of the current lane. Within the slice range, the first drivable slice at each time moment is calculated, and directed edges are established between slices at adjacent time moments. Based on the overlapping directed edges in the target direction, a first multi-branch tree structure is constructed, and then the first drivable convex space over the entire time domain is obtained through depth-first search. This effectively ensures that all feasible trajectories of the current lane can be completely extracted during the lane change decision process, avoiding the omission of potential solutions, providing a complete and vehicle-dynamically compatible feasible space, thereby improving the reliability and success rate of lane change planning.

[0012] In conjunction with the first aspect, in some possible implementations, determining the second drivable convex space of the target lane based on the second path time map, the upper bound curve, and the lower bound curve of the target lane includes: determining the slice range of the second path time map based on the upper bound curve and the lower bound curve of the target lane; calculating the second drivable slice at each moment on the second path time map within the slice range, and establishing directed edges of the second drivable slice at adjacent moments; constructing a second multi-branch tree structure based on the directed edges of the second drivable slice overlapping in the target direction, and obtaining the second drivable convex space over the entire time domain by performing a depth-first search on the second multi-branch tree structure.

[0013] Through the above technical solution, the embodiments of the present invention determine the slice range of the second path time map based on the upper and lower bound curves of the target lane, calculate the second drivable slice at each time within the slice range, establish directed edges between slices at adjacent time points, construct a second multi-branch tree structure based on the overlapping directed edges in the target direction, and obtain the second drivable convex space in the entire time domain through depth-first search. This completely extracts all trajectory regions in the target lane that satisfy vehicle dynamics constraints and spatiotemporal feasibility, providing a complete foundation for the generation of candidate lane change spaces and improving the reliability and success rate of lane change decisions.

[0014] In conjunction with the first aspect, in some possible implementations, generating candidate lane-changing longitudinal convex spaces based on the first drivable convex space and the second drivable convex space includes: traversing each second drivable convex space of the target lane; matching the second drivable convex space with the first drivable convex space of the current lane; and generating candidate lane-changing longitudinal convex spaces based on the matching results.

[0015] Through the above technical solution, the embodiments of the present invention generate candidate lane-changing longitudinal convex spaces by traversing each second drivable convex space of the target lane and matching it with the first drivable convex space of the current lane. The system covers all possible spatiotemporal combinations of lane changes, ensuring that no feasible lane-changing opportunities are missed, providing a complete and effective candidate solution set for subsequent speed optimization and decision-making, and improving the success rate of lane changes and the completeness of planning.

[0016] In conjunction with the first aspect, in some possible implementations, the second drivable convex space is matched with the first drivable convex space of the current lane, and candidate lane-changing longitudinal convex spaces are generated based on the matching results. This includes: calculating the longest duration of continuous overlap between the first drivable convex space and the second drivable convex space, and estimating the minimum remaining lane-changing time of the autonomous vehicle at the current moment; if the longest duration is greater than the minimum remaining lane-changing time, then the matched first drivable convex space and the second drivable convex space are constructed as candidate lane-changing longitudinal convex spaces.

[0017] Through the above technical solution, the embodiments of the present invention calculate the longest duration of continuous overlap between the first drivable convex space and the second drivable convex space, and estimate the minimum remaining lane-changing time at the current moment. Only when the overlap time is greater than the remaining lane-changing time is a candidate lane-changing longitudinal convex space constructed, ensuring that the generated candidate space is sufficient in time to complete a safe lane change, effectively eliminating infeasible matching combinations, and improving the feasibility and real-time performance of lane-changing decisions.

[0018] In conjunction with the first aspect, in some possible implementations, the lane change speed curve is obtained by solving the speed optimization problem in the longitudinal convex space of the candidate lane change, including: obtaining the optimization variables of the autonomous vehicle and establishing an optimization objective based on the optimization variables; identifying the overlap start time and overlap end time in the longitudinal convex space of the candidate lane change, and determining the environmental constraints of the optimization objective based on the overlap start time and overlap end time; obtaining the acceleration constraints, jerk constraints, and speed limit constraints of the autonomous vehicle and the current area, and generating constraint conditions based on the environmental constraints, acceleration constraints, jerk constraints, and speed limit constraints; establishing a speed optimization problem in the longitudinal convex space of the candidate lane change based on the optimization objective and constraint conditions, solving the speed optimization problem using a pre-set solver, and generating the lane change speed curve based on the solution results.

[0019] Through the above technical solution, the embodiments of the present invention determine environmental constraints based on the start and end times of overlap within the longitudinal convex space of the candidate lane change, and construct a speed optimization problem containing optimization objectives and complete constraint conditions by combining acceleration, jerk and speed limit constraints. The solver is used to generate the lane change speed curve, so that the obtained speed curve strictly meets the vehicle dynamics constraints, ride comfort requirements and road speed limits, while adapting to the actual drivable time and space window, thereby improving the feasibility, safety and comfort of the lane change trajectory.

[0020] In a second aspect, an electronic device is provided, including a processor and a memory, wherein the memory is used to store computer programs; and the processor is used to execute the programs stored in the memory to implement the autonomous driving lane change decision method in the first aspect or any possible implementation thereof.

[0021] Thirdly, a vehicle is provided that includes the electronic equipment described in the second aspect or possible implementations thereof. Attached Figure Description

[0022] Figure 1 This is a flowchart of the autonomous driving lane change decision method provided in the embodiments of this application; Figure 2 This is a schematic diagram of an autonomous driving lane change decision-making scenario provided in an embodiment of this application; Figure 3 This is a schematic diagram of the generation of a drivable convex space provided in an embodiment of this application; Figure 4 This is a flowchart of the autonomous driving lane change decision-making process provided in an embodiment of this application; Figure 5 This is a structural diagram of the electronic device provided in the embodiments of this application. Detailed Implementation

[0023] To make the technical problems, technical solutions, and beneficial effects solved by this application clearer, the following detailed description is provided in conjunction with embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0024] The main autonomous driving lane-changing decision-making methods in related technologies include: rule-based methods, some of which consider the safety of the autonomous vehicle and obstacles in front and behind the target lane at the current moment, but cannot actively accelerate or decelerate in response to future drivable gaps in the target lane, resulting in a low success rate of lane changes; some of which consider the game between the autonomous vehicle and the vehicle in the target lane, but require prediction of the other vehicle's acceleration, which places high demands on perception; some methods only evaluate the benefits of changing lanes in adjacent lanes to generate intent, without considering how to execute it; some solutions rely on V2X (Vehicle-to-Everything) vehicle-road cooperation, which limits the applicable scenarios; other methods select the optimal solution by sampling multiple trajectories and evaluating costs, but collision detection is required for each trajectory, which is computationally time-consuming and cannot guarantee the completeness of the solution, and is prone to missing feasible solutions in congested scenarios; in addition, end-to-end methods based on deep learning are black-box models, requiring a large amount of high-quality data for training, lacking interpretability and traceability, and are difficult to meet safety requirements.

[0025] Figure 1 This is a flowchart of an autonomous driving lane change decision method according to an embodiment of the present invention.

[0026] like Figure 1 As shown, the autonomous driving lane change decision method according to an embodiment of the present invention includes the following steps: Step S110: Obtain the first path time map of the current lane and the second path time map of the target lane.

[0027] Step S120: Obtain the upper and lower bound curves for each lane based on the acceleration constraints of the autonomous vehicle.

[0028] Step S130: Determine the first drivable convex space of the current lane based on the first path time map, the upper boundary curve and the lower boundary curve of the current lane, and determine the second drivable convex space of the target lane based on the second path time map, the upper boundary curve and the lower boundary curve of the target lane.

[0029] Step S140: Generate candidate lane-changing longitudinal convex spaces based on the first drivable convex space and the second drivable convex space.

[0030] Step S150: Obtain the lane change speed curve by optimizing the speed of the candidate lane change longitudinal convex space. Determine the lane change decision result of the autonomous vehicle based on the cost of the lane change speed curve. The lane change decision result includes at least one of the following: lane change timing, lane change speed, lane change flag, and cancellation flag. The lane change flag indicates whether the autonomous vehicle is allowed to change lanes, and the cancellation flag indicates that the autonomous vehicle has cancelled the lane change.

[0031] It should be noted that the first path time map refers to the region map showing the change in the feasible position of a vehicle in the current lane over time; the second path time map refers to the region map showing the change in the feasible position of a vehicle in the target lane over time; the upper bound curve refers to the maximum longitudinal position limit calculated based on the initial speed and maximum acceleration; the lower bound curve refers to the minimum longitudinal position limit calculated based on the initial speed and maximum deceleration; the first drivable convex space refers to the continuous drivable area in the current lane that conforms to the kinematic rules and has no collision risk; the second drivable convex space refers to the continuous drivable area in the target lane that conforms to the kinematic rules and has no collision risk; candidate lane change. The longitudinal convex space refers to the area that coincides with two drivable convex spaces in both time and space, suitable for safe lane changing; the lane change speed curve refers to the optimal speed change curve found within the selected lane change area, ensuring the safety and efficiency of the lane change process; the current lane refers to the lane in which the autonomous vehicle is currently traveling; the target lane refers to the adjacent lane into which the autonomous vehicle plans to change; the lane change timing refers to the specific moment when the lane change begins; the lane change speed refers to the longitudinal speed of the entire lane change process obtained through speed optimization; the lane change sign is used to determine whether the autonomous vehicle is allowed to change lanes; the cancellation sign is used to determine whether the autonomous vehicle's lane change is cancelled.

[0032] This application embodiment can achieve accurate modeling of the spatiotemporal distribution of dynamic obstacles in both lanes by obtaining the first path time map of the current lane and the second path time map of the target lane, thereby improving the accuracy of lane change decisions; it obtains the upper and lower bound curves of each lane based on the acceleration constraints of the autonomous vehicle, ensuring that the generated motion boundary conforms to the feasibility of vehicle dynamics and improving driving safety; it determines the first drivable convex space based on the first path time map and the upper and lower bound curves of the current lane, and determines the second drivable convex space based on the second path time map and the upper and lower bound curves of the target lane, transforming the non-convex feasible region into a structured convex set to support efficient The system searches and optimizes computational efficiency; it generates candidate longitudinal convex spaces for lane changes based on two drivable convex spaces, effectively filtering out feasible longitudinal trajectories that meet the lane change time window, thus enhancing the flexibility of lane change timing selection; it performs speed optimization solutions based on the candidate longitudinal convex spaces to obtain lane change speed curves, and determines lane change decision results based on their cost value, including at least one of lane change timing, lane change speed, lane change sign, and cancellation sign. The lane change sign indicates whether a lane change is allowed, and the cancellation sign indicates cancellation of a lane change, thereby achieving safe, real-time, and driver-intent-compliant lane change decisions, improving the success rate of lane changes, and ensuring the rationality and safety of driving operations.

[0033] In step S110, obtaining the first path time map of the current lane and the second path time map of the target lane includes: obtaining the first reference line of the current lane and the second reference line of the target lane; obtaining the size and lateral safety threshold of the autonomous vehicle, and scanning along the first reference line and the second reference line respectively in combination with the size and lateral safety threshold; generating the first path time map based on the scanning result of the first reference line; and generating the second path time map based on the scanning result of the second reference line.

[0034] The first reference line refers to the current lane centerline or the longitudinal reference trajectory used in the planning; the second reference line refers to the target lane centerline or the longitudinal reference trajectory used in the planning; the lateral safety threshold refers to the additional lateral distance based on the vehicle width, used to ensure lane changing safety.

[0035] It is understood that this application embodiment obtains a first reference line for the current lane and a second reference line for the target lane, and scans along the two reference lines in conjunction with the size of the autonomous vehicle and a lateral safety threshold to generate a first path time map and a second path time map. This ensures that the actual physical size and dynamic behavior characteristics of the vehicle are fully considered during the lane-changing decision-making process. It accurately depicts the drivable area of ​​the vehicle in the current lane and the target lane, improving the accuracy and reliability of lane-changing decisions. The use of lateral safety thresholds further enhances the safety of lane-changing operations, avoiding potential collision risks. Based on these path time maps, the timing of lane changes can be more accurately identified and the optimal speed curve can be planned, thereby improving the success rate of lane changes and driving safety, while ensuring that the entire process complies with vehicle dynamics constraints.

[0036] It should be noted that the lateral safety threshold is related to the vehicle's current speed; the higher the speed, the higher the lateral safety threshold.

[0037] Specifically, Figure 2 A lane change scenario includes the current lane, the target lane, and the adjacent lane, such as... Figure 2 As shown, the autonomous vehicle is in the current lane, obstacle A is in the adjacent lane and intrudes diagonally, and obstacles B and C are in the target lane. Each lane has reference lines for subsequent path time map generation and drivable space calculation.

[0038] In step S120, the upper and lower bound curves of each lane are obtained based on the acceleration constraints of the autonomous vehicle, including: identifying the maximum acceleration and maximum deceleration in the acceleration constraints of the autonomous vehicle; using the maximum planned duration of the autonomous vehicle as a constraint, generating the upper bound curves of the current lane and the target lane based on the maximum acceleration and the actual planned duration of the autonomous vehicle; using the maximum planned duration of the autonomous vehicle as a constraint, generating the lower bound curve of the current lane based on the maximum deceleration and the actual planned duration of the autonomous vehicle, and obtaining the pre-set lower bound curve of the target lane.

[0039] Among them, maximum acceleration refers to the maximum longitudinal acceleration that autonomous vehicles are allowed to use; maximum deceleration refers to the maximum longitudinal deceleration that autonomous vehicles are allowed to use; and maximum planning duration refers to the longest planning time window considered in lane change decisions.

[0040] For example, based on information about obstacles within 50 meters behind the target lane, the lower boundary curve of the target lane is pre-set as a low-speed curve that meets the safe following distance requirement.

[0041] It is understood that the embodiments of this application, by identifying the maximum acceleration and maximum deceleration of the autonomous vehicle and generating upper and lower bound curves for each lane based on the maximum planning time, ensure that the vehicle's dynamic characteristics are fully considered during lane-changing decisions. Accurately determining the vehicle's motion limits in the current lane improves the safety and reliability of path planning. The upper and lower bound curves provide clear speed boundaries for the vehicle, avoiding risks caused by speeding or insufficient deceleration. This enhances the safety of lane-changing operations and ensures consistency between the decision and the vehicle's physical limitations, improving lane-changing success rate and driving comfort.

[0042] Specifically, the formulas for calculating the upper and lower bound curves of the current lane are as follows:

[0043]

[0044] in, This is the upper bound curve; This is the lower bound curve; for =The initial velocity of the autonomous vehicle at time 0; This represents the maximum acceleration of an autonomous vehicle. This represents the maximum deceleration of the autonomous vehicle. To plan for the maximum duration.

[0045] It should be noted that the calculation method for the upper limit curve of the target lane is the same as that for the upper limit curve of the current lane. The lower limit curve of the target lane is a pre-set curve that can be pre-calibrated or set in conjunction with requirements such as the safe following distance of vehicles.

[0046] In step S130, determining the first drivable convex space of the current lane based on the first path time map, the upper boundary curve and the lower boundary curve of the current lane includes: determining the slice range of the first path time map based on the upper boundary curve and the lower boundary curve of the current lane; calculating the first drivable slice at each moment on the first path time map within the slice range, and establishing directed edges of the first drivable slice at adjacent moments; constructing a first multi-branch tree structure based on the directed edges of the first drivable slice overlapping in the target direction, and obtaining the first drivable convex space over the entire time domain by performing a depth-first search on the first multi-branch tree structure.

[0047] Here, the slice range refers to the longitudinal position interval defined by the upper and lower bound curves of the current lane in the first path time graph; the first drivable slice refers to the set of longitudinal drivable positions of the current lane at a certain moment, defined by the upper bound curve, the lower bound curve, and obstacle constraints; the directed edge refers to the directed association connecting drivable slices at adjacent moments; the first multi-branch tree structure refers to the tree structure composed of overlapping directed edges between the first drivable slices, used to represent the temporal connection relationship of the drivable region of the current lane; depth-first exploration refers to the search method of traversing the first multi-branch tree structure to extract continuous and complete drivable convex space.

[0048] It is understood that, in this embodiment of the application, the slice range of the first path time map is determined by the upper and lower bound curves of the current lane. Within the slice range, the first drivable slice at each time moment is calculated, and directed edges are established between slices at adjacent time moments. Based on the overlapping directed edges in the target direction, a first multi-branch tree structure is constructed, and then the first drivable convex space over the entire time domain is obtained through depth-first search. This effectively ensures that all feasible trajectories of the current lane can be completely extracted during the lane change decision process, avoiding the omission of potential solutions, providing a complete and vehicle-dynamically compatible feasible space, thereby improving the reliability and success rate of lane change planning.

[0049] In this embodiment of the invention, determining the second drivable convex space of the target lane based on the second path time map, the upper boundary curve and the lower boundary curve of the target lane includes: determining the slice range of the second path time map based on the upper boundary curve and the lower boundary curve of the target lane; calculating the second drivable slice at each moment on the second path time map within the slice range, and establishing directed edges of the second drivable slice at adjacent moments; constructing a second multi-branch tree structure based on the directed edges of the second drivable slice overlapping in the target direction, and obtaining the second drivable convex space over the entire time domain by performing a depth-first search on the second multi-branch tree structure.

[0050] The second drivable slice refers to the set of longitudinal drivable positions of the target lane at a certain moment, defined by the upper boundary curve, the lower boundary curve, and the obstacle constraint; the second multi-branch tree structure refers to the tree structure formed by the overlapping directed edges between the second drivable slices, which is used to represent the temporal connection relationship of the drivable area of ​​the target lane.

[0051] It is understood that, in this embodiment of the application, the slice range of the second path time map is determined based on the upper and lower bound curves of the target lane. Within the slice range, the second drivable slice at each time moment is calculated, directed edges are established between slices at adjacent time moments, a second multi-branch tree structure is constructed based on the overlapping directed edges in the target direction, and the second drivable convex space in the entire time domain is obtained through depth-first search. This process completely extracts all trajectory regions in the target lane that satisfy vehicle dynamics constraints and spatiotemporal feasibility, providing a complete foundation for the generation of candidate lane change spaces and improving the reliability and success rate of lane change decisions.

[0052] Specifically, the process of generating a drivable convex space includes: calculating the drivable slice at each time step, and establishing a drivable convex space for the slice. Time's up If the directed edges at time s overlap in the s direction, a multi-branch tree structure can be obtained. Finally, the drivable convex space over the entire time domain can be obtained through depth-first search.

[0053] Furthermore, such as Figure 3 As shown, a first drivable convex space and a second drivable convex space are generated based on the first path time map and the second path time map. The first drivable convex space includes first drivable convex space C1 and first drivable convex space C2. Regions 1 and 2 form the first drivable convex interval C1, and regions 1 and 3 form the first drivable convex interval C2. The second drivable convex space includes second drivable convex spaces T1, T2, and T3. The upper and lower bound curves of the motion range are determined by the maximum acceleration and maximum deceleration constraints. Obstacles A, B, and C are projected as non-drivable areas in the path time map. The feasible space for each time period is obtained through slicing and overlap analysis.

[0054] In step S140, generating candidate lane-changing longitudinal convex spaces based on the first drivable convex space and the second drivable convex space includes: traversing each second drivable convex space of the target lane; matching the second drivable convex space with the first drivable convex space of the current lane; and generating candidate lane-changing longitudinal convex spaces based on the matching results.

[0055] It is understood that the embodiments of this application generate candidate lane change longitudinal convex spaces by traversing each second drivable convex space of the target lane and matching it with the first drivable convex space of the current lane. The system covers all possible spatiotemporal combinations of lane changes, ensuring that no feasible lane change opportunities are missed, providing a complete and effective candidate solution set for subsequent speed optimization and decision-making, and improving the success rate of lane changes and the completeness of planning.

[0056] In this embodiment of the invention, matching the second drivable convex space with the first drivable convex space of the current lane and generating candidate lane-changing longitudinal convex spaces based on the matching results includes: calculating the longest duration of continuous overlap between the first drivable convex space and the second drivable convex space, and estimating the minimum remaining lane-changing time of the autonomous vehicle at the current moment; if the longest duration is greater than the minimum remaining lane-changing time, then the matched first drivable convex space and the second drivable convex space are constructed as candidate lane-changing longitudinal convex spaces.

[0057] Among them, the longest duration refers to the longest time period during which the first drivable convex space and the second drivable convex space continuously overlap in time; the minimum remaining lane change time refers to the shortest time required to change lanes, obtained by dividing the lateral remaining distance by the lateral average lane change speed.

[0058] It is understood that, in this embodiment of the application, the longest duration of continuous overlap between the first drivable convex space and the second drivable convex space is calculated, and the minimum remaining lane-changing time at the current moment is estimated. Only when the overlap time is greater than the remaining lane-changing time is a candidate lane-changing longitudinal convex space constructed, so as to ensure that the generated candidate space is sufficient in time to complete a safe lane change, effectively eliminate infeasible matching combinations, and improve the feasibility and real-time performance of lane-changing decisions.

[0059] It should be noted that the minimum remaining lane change time is the lateral distance between the current position of the autonomous vehicle and the reference line of the target lane. The average lateral speed of lane change is related to the current speed of the autonomous vehicle; the higher the speed, the lower the lateral speed.

[0060] Specifically, each drivable convex space of the target lane is traversed and matched with the drivable convex space of the current lane, and the longest duration of continuous overlap between the first and second drivable convex spaces is calculated. Here, overlap refers to the spatial overlap between the first and second drivable slices at a certain moment. The formula for calculating the minimum remaining lane-changing time is as follows:

[0061] in, This represents the remaining lateral lane-changing distance. This represents the average lane-changing speed laterally. This represents the minimum remaining lane-changing time.

[0062] In step S150, the lane change speed curve is obtained by solving the speed optimization problem of the candidate lane change longitudinal convex space, including: obtaining the optimization variables of the autonomous vehicle and establishing the optimization objective based on the optimization variables; identifying the overlap start time and overlap end time in the candidate lane change longitudinal convex space and determining the environmental constraints of the optimization objective based on the overlap start time and overlap end time; obtaining the acceleration constraints, jerk constraints and speed limit constraints of the autonomous vehicle and the current area, and generating constraint conditions based on the environmental constraints, acceleration constraints, jerk constraints and speed limit constraints; establishing the speed optimization problem of the candidate lane change longitudinal convex space based on the optimization objective and constraint conditions, solving the speed optimization problem using a pre-set solver, and generating the lane change speed curve based on the solution results.

[0063] Among them, the optimization variable refers to the velocity value at each discrete moment in the planning time domain, which is used to derive acceleration and jerk; the environmental constraint refers to the allowable range of longitudinal position at each moment determined by the longitudinal convex space of the candidate lane change; the acceleration constraint refers to the maximum acceleration and maximum deceleration limit allowed by vehicle dynamics; and the speed limit constraint refers to the maximum driving speed limit specified by the current road or area.

[0064] It is understood that the embodiments of this application determine environmental constraints based on the start and end times of overlap within the longitudinal convex space of the candidate lane change, and construct a speed optimization problem containing optimization objectives and complete constraints by combining acceleration, jerk and speed limit constraints. The solver is used to generate the lane change speed curve, so that the obtained speed curve strictly meets the vehicle dynamics constraints, ride comfort requirements and road speed limits, while adapting to the actual drivable time and space window, thereby improving the feasibility, safety and comfort of the lane change trajectory. It should be noted that when solving speed optimization problems, multiple MPC (Model Predictive Control) speed optimization problems need to be solved. Therefore, it is necessary to control the time consumption of a single MPC optimization. This goal can be achieved by setting the maximum number of iterations to 300 in the solver.

[0065] Specifically, to construct an MPC speed optimization problem for each candidate lane change longitudinal convex space, the environmental constraints first need to be preprocessed: let the start time of the maximum overlap between the first and second drivable convex spaces be... The end time is For obstacles above and below the second drivable convex space Previously, these were not considered environmental constraints; obstacles above and below the first drivable convex space were... In the moments that follow, it will not be considered an environmental constraint.

[0066] The optimization variables after discretization are... Composition, in which express The total number of discretized segments. The optimization objective is as follows:

[0067] in, Weights for reference speed; Weights for acceleration; Weights for accelerometers; For the desired speed; Let be the velocity at time i; Let be the acceleration at time i; Let be the jerk at time i.

[0068] The following difference formula is used for calculation:

[0069] in, Let be the acceleration at time i; Let be the acceleration at time i+1; For discrete time intervals, , To plan for the maximum duration.

[0070] The system equations are constrained by the following two formulas:

[0071]

[0072] in, Let i be the position at time i. It is the velocity at time i; Let be the acceleration at time i; For time step; To predict the total number of segments.

[0073] Specifically, constraint conditions are generated based on environmental constraints, acceleration constraints, jerk constraints, and speed limit constraints, and the specific constraint conditions are as follows: (1) Environmental constraints: , .in, This is the lower bound of the convex space obstacle at the i-th time step; Let be the upper bound of the convex space obstacle at the i-th time.

[0074] (2) Speed ​​limit constraints: .in, This is the upper bound of the velocity at time i.

[0075] (3) Acceleration constraints: .in, This is the maximum deceleration; This is the maximum acceleration.

[0076] (4) Jerk constraint: .in, Minimum jerk; This is the maximum jerk.

[0077] Furthermore, such as Figure 4 As shown, the specific process of lane-changing decision-making in autonomous driving includes: In step 401, a first path time map and a second path time map are generated.

[0078] In step 402, the first drivable convex space and the second drivable convex space are determined.

[0079] In step 403, candidate lane change longitudinal convex spaces are generated.

[0080] In step 404, MPC speed optimization is constructed for each candidate lane change longitudinal convex space. In step 405, it is determined whether to change lanes based on the cost value.

[0081] In step 406, if the minimum cost is greater than a preset threshold, then lane keeping is enabled.

[0082] In step 407, if the minimum cost is less than a preset threshold, the optimal lane change speed curve is selected.

[0083] Specifically, the design of the cost function needs to comprehensively consider safety, efficiency, and comfort, and the weights of the three can be adjusted according to different scenarios to ultimately achieve the goal of conforming to human driving habits.

[0084] In step 408, the decision result is output.

[0085] Specifically, when the upstream "lane change request" module sends a lane change request and lane change direction, the autonomous driving lane change decision method is activated. Based on the longitudinal convex space of the candidate lane change, speed optimization is performed to obtain the speed curves of each lane change and their cost values. The decision is output based on the comparison between the minimum cost value and a preset threshold: if the minimum cost value is greater than the threshold, "lane keeping" is output, and the current lane driving trajectory is generated by the subsequent trajectory planning module; if the minimum cost value is less than or equal to the threshold, "execute lane change" and the corresponding optimal lane change speed curve are output for the trajectory planning module to generate the lane change trajectory.

[0086] According to the proposed autonomous driving lane change decision-making method, by acquiring the first path time map of the current lane and the second path time map of the target lane, the upper and lower bound curves of each lane are obtained by the acceleration constraints of the autonomous vehicle. Based on the first path time map and the upper and lower bound curves of the current lane, the first drivable convex space of the current lane is determined, and based on the second path time map and the upper and lower bound curves of the target lane, the second drivable convex space of the target lane is determined. These two methods together generate candidate lane change longitudinal convex spaces. The lane change speed curve is obtained through speed optimization, and the lane change decision result is determined by the cost of the lane change speed curve, including at least one of the following: lane change timing, lane change speed, lane change flag, and cancellation flag. The lane change flag indicates whether a lane change is allowed, and the cancellation flag indicates that the lane change is cancelled. This achieves safe, real-time lane change decisions that conform to driving intentions, improves the success rate of lane changes, and ensures the safety of driving operations.

[0087] This application also provides an electronic device, please refer to... Figure 5 It includes a memory 510 and a processor 520, wherein the memory 510 is used to store computer programs; and the processor 520 is used to execute the programs stored in the memory to implement the autonomous driving lane change decision method described in any embodiment of this application.

[0088] This application also provides a vehicle that includes the electronic equipment described in the above embodiments.

[0089] In this application, "multiple" refers to two or more.

[0090] In this application, unless otherwise expressly defined, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection between two components. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.

[0091] The terms “first,” “second,” “third,” “fourth,” etc., in this application (if present) are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence.

[0092] In this application, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, in this application, the character " / " generally indicates that the preceding and following related objects have an "or" relationship.

[0093] Unless otherwise specified, all steps in this application may be performed sequentially or randomly. For example, if a method includes steps A and B, it means that the method may include steps A and B performed sequentially, or it may include steps B and A performed sequentially. For example, if a method may also include step C, it means that step C may be added to the method in any order. For example, the method may include steps A, B, and C, or it may include steps A, C, and B, or it may include steps C, A, and B, etc.

[0094] The above are merely preferred embodiments of this application and are not intended to limit this application. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. An autonomous driving lane change decision-making method, characterized in that, Includes the following steps: Obtain the first path time map of the current lane and the second path time map of the target lane; The upper and lower bound curves of each lane are obtained based on the acceleration constraints of the autonomous vehicle. The first drivable convex space of the current lane is determined based on the first path time map, the upper boundary curve and the lower boundary curve of the current lane, and the second drivable convex space of the target lane is determined based on the second path time map, the upper boundary curve and the lower boundary curve of the target lane. Candidate lane-changing longitudinal convex spaces are generated based on the first drivable convex space and the second drivable convex space. The lane change speed curve is obtained by speed optimization solution of the longitudinal convex space of the candidate lane change, and the lane change decision result of the autonomous vehicle is determined according to the cost of the lane change speed curve. The lane change decision result includes at least one of lane change timing, lane change speed, lane change flag and cancellation flag. The lane change flag indicates whether the autonomous vehicle is allowed to change lanes, and the cancellation flag indicates that the lane change of the autonomous vehicle is cancelled.

2. The autonomous driving lane change decision-making method according to claim 1, characterized in that, The process of obtaining the first path time map of the current lane and the second path time map of the target lane includes: Obtain the first reference line of the current lane and the second reference line of the target lane; The dimensions and lateral safety threshold of the autonomous vehicle are obtained, and the vehicle is scanned along the first reference line and the second reference line in combination with the dimensions and the lateral safety threshold. The first path time map is generated based on the scanning results of the first reference line; The second path time map is generated based on the scanning results of the second reference line.

3. The autonomous driving lane change decision-making method according to claim 1, characterized in that, The process of obtaining the upper and lower bound curves for each lane based on the vehicle's acceleration constraints includes: Identify the maximum acceleration and maximum deceleration in the acceleration constraints of the autonomous vehicle; Using the maximum planning time of the autonomous vehicle as a constraint, the upper bound curves of the current lane and the target lane are generated based on the maximum acceleration and the actual planning time of the autonomous vehicle. Using the maximum planned duration of the autonomous vehicle as a constraint, the lower bound curve of the current lane is generated based on the maximum deceleration and the actual planned duration of the autonomous vehicle, and the lower bound curve of the target lane is obtained in advance.

4. The autonomous driving lane change decision-making method according to claim 1, characterized in that, The step of determining the first drivable convex space of the current lane based on the first path time map, the upper boundary curve and the lower boundary curve of the current lane includes: The slice range of the first path time map is determined based on the upper and lower bound curves of the current lane. Within the slice range, calculate the first drivable slice at each moment on the first path time map, and establish directed edges of the first drivable slice at adjacent moments; A first multi-branch tree structure is constructed based on the directed edges that overlap in the target direction of the first drivable slice. The first drivable convex space in the entire time domain is obtained by depth-first search of the first multi-branch tree structure.

5. The autonomous driving lane change decision-making method according to claim 1, characterized in that, Determining the second drivable convex space of the target lane based on the second path time map, the upper boundary curve, and the lower boundary curve of the target lane includes: The slice range of the second path time map is determined based on the upper and lower bound curves of the current lane. Within the slice range, calculate the second drivable slice at each time point on the second path time graph, and establish directed edges between adjacent time points for the second drivable slice; A second multi-branch tree structure is constructed based on the directed edges that overlap in the target direction of the second drivable slice. A second drivable convex space over the entire time domain is obtained by depth-first search of the second multi-branch tree structure.

6. The autonomous driving lane change decision method according to claim 1, characterized in that, The step of generating candidate lane-changing longitudinal convex spaces based on the first drivable convex space and the second drivable convex space includes: Traverse each second drivable convex space of the target lane; The second drivable convex space is matched with the first drivable convex space of the current lane, and the candidate lane change longitudinal convex space is generated based on the matching result.

7. The autonomous driving lane change decision method according to claim 6, characterized in that, The step of matching the second drivable convex space with the first drivable convex space of the current lane, and generating the candidate lane-changing longitudinal convex space based on the matching result, includes: Calculate the longest duration of continuous overlap between the first drivable convex space and the second drivable convex space, and estimate the minimum remaining lane-changing time of the autonomous vehicle at the current moment; If the longest duration is greater than the minimum remaining lane change time, then the matched first drivable convex space and the second drivable convex space are constructed as the candidate lane change longitudinal convex space.

8. The autonomous driving lane change decision method according to claim 1, characterized in that, The process of obtaining the lane change speed curve based on the speed optimization solution of the candidate lane change longitudinal convex space includes: Obtain the optimization variables of the autonomous vehicle, and establish an optimization objective based on the optimization variables; Identify the overlap start time and overlap end time in the longitudinal convex space of the candidate lane change, and determine the environmental constraints of the optimization target based on the overlap start time and overlap end time; Obtain the acceleration constraints, jerk constraints, and speed limit constraints of the current area for the autonomous vehicle, and generate constraint conditions based on the environmental constraints, acceleration constraints, jerk constraints, and speed limit constraints; Based on the optimization objective and the constraints, a speed optimization problem is established for the candidate lane change longitudinal convex space. The speed optimization problem is solved using a pre-set solver, and the lane change speed curve is generated based on the solution results.

9. An electronic device, characterized in that, Including processor and memory, among which Memory, used to store computer programs; A processor is used to execute a program stored in memory to implement the autonomous driving lane change decision method according to any one of claims 1-8.

10. A vehicle, characterized in that, It includes the electronic device as described in claim 9.

Images