An intelligent automatic driving vehicle lane-changing trajectory generation method and system

Abstract

This invention discloses a method and system for generating lane-changing trajectories for intelligent autonomous vehicles, relating to the field of motion planning technology for intelligent driving vehicles. The method includes: acquiring the current state variables of the vehicle and the vehicles in front and behind it in the target lane, as well as the target lane-changing state variable; calculating a gap tension index characterizing the degree of lane-changing gap tension, and dynamically updating the elastic soft constraints accordingly to obtain a dynamic constraint set; dividing the planning time domain and allocating solution precision accordingly, and constructing a hierarchical solution structure based on the solution results of the previous cycle; determining the search range for lane-changing initiation delay, generating feasible candidate lane-changing initiation times and corresponding trajectory models; solving the optimization sub-problems for each candidate time; selecting the optimal path based on a joint cost function with dynamically adjusted weights combined with cross-cycle robustness evaluation, and outputting the target lane-changing trajectory after smooth transition. This invention achieves dynamic adaptive adjustment of the lane-changing planning strategy throughout the entire process according to the gap risk state, balancing efficiency, comfort, and real-time performance while ensuring safety.

Description

Technical Field

[0001] This invention relates to the field of motion planning technology for intelligent driving vehicles, and in particular to a method and system for generating lane-changing trajectories for intelligent autonomous driving vehicles. Background Technology

[0002] Lane changing is a typical maneuver frequently performed by autonomous vehicles in complex traffic scenarios. Lane changing trajectory planning needs to balance comfort and efficiency while ensuring safety. Existing lane changing trajectory planning methods typically optimize the lane changing trajectory based on a fixed set of constraints and a static cost function. The planning parameters remain unchanged throughout the lane change, making it difficult to dynamically adjust the planning strategy according to the real-time risk status of the target lane gap. In scenarios with tight gaps, there is a risk of insufficient safety margin, while in scenarios with ample gaps, there are problems with lane changing efficiency and comfort.

[0003] Furthermore, existing methods often employ fixed-time or uniform discrete search strategies when selecting lane change initiation times, failing to fully utilize the dynamic evolution trend information of target lane gaps to optimize the generation and screening of candidate times. This results in low efficiency in the allocation of computing resources, making it difficult to balance the solution quality and real-time planning of multiple candidate time evaluations under the condition of limited computing power on the vehicle computing platform. Summary of the Invention

[0004] The main objective of this invention is to provide a method and system for generating lane-changing trajectories for intelligent autonomous vehicles, aiming to solve the technical problems in existing lane-changing trajectory planning methods, such as the inability to dynamically adjust planning parameters according to the risk state of the target lane gap, the lack of effective combination of lane-changing initiation timing generation and gap evolution trend, and the difficulty in balancing solution quality and planning real-time performance under limited computing power.

[0005] In a first aspect, the present invention provides a method for generating lane-changing trajectories for intelligent autonomous vehicles, comprising: Obtain the current state of the vehicle, the current state of the vehicle in front of the target lane, the current state of the vehicle behind the target lane, and the target state of the lane change. Based on the current state of the vehicle, the current state of the vehicle in front of the target lane, and the current state of the vehicle behind the target lane, calculate the gap tension index, which characterizes the current lane change gap tension. The lane-changing trajectory optimization constraints are divided into hard constraints and flexible soft constraints. The boundary values ​​of the flexible soft constraints are dynamically updated based on the gap tension index to obtain a dynamic constraint set. The planning time domain is divided into multiple solution segments to construct a hierarchical solution structure, and the hierarchical solution structure is hot-started and initialized based on the solution results of the previous control cycle. The search range for lane change start delay is determined based on the gap tension index. The delay is used as an explicit optimization variable to generate multiple candidate lane change start times, and a corresponding candidate lane change trajectory model is established for each candidate lane change start time. Under the dynamic constraint set and hierarchical solution structure, the optimization sub-problems corresponding to each candidate lane change start time are solved one by one to obtain the corresponding trajectory parameters; Based on the preset joint cost function, the candidate lane-changing start time with the minimum cost is selected from the candidate lane-changing start times with feasible solutions in the solution results, and the corresponding target lane-changing trajectory is output.

[0006] Optionally, the clearance tension index used to calculate the current lane change clearance tension includes: Based on the current state of the vehicle, the current state of the vehicle in front of the target lane and the current state of the vehicle behind the target lane, predict the motion trajectory of the vehicle, the vehicle in front of the target lane and the vehicle behind the target lane in a preset prediction time domain, and calculate the relative motion state between the vehicle and the vehicle in front of the target lane and the relative motion state between the vehicle and the vehicle behind the target lane based on the motion trajectory. The time series sequence for calculating the safety margin of the clearance between the preceding and following vehicles in the time domain is calculated based on the relative motion state of the preceding vehicle, and the time series sequence for calculating the safety margin of the clearance between the following and following vehicles in the time domain is calculated based on the relative motion state of the following vehicle. Based on the time series of the safety margin of the front vehicle clearance and the time series of the safety margin of the rear vehicle clearance, the worst-case conditions on both the front and rear sides in the prediction time domain are comprehensively evaluated and predicted to obtain the clearance tension index and the clearance tension evolution sequence.

[0007] Optionally, the process of determining the dynamic constraint set includes: Based on the correlation between lane change trajectory optimization constraints and lane change safety, lane change trajectory optimization constraints are divided into hard constraints and initial elastic soft constraints. The gap tension index is compared with the preset tension threshold, and the hard constraint and elastic soft constraint are dynamically converted based on the comparison results and the gap tension evolution sequence. For elastic soft constraints that have not undergone type conversion, their boundary values ​​are calculated and updated based on the gap tension index. A dynamic constraint set is constructed based on the hard constraints after type conversion, the flexible soft constraints, and the updated boundary values.

[0008] Optionally, the construction and warm-start initialization process of the hierarchical solution structure includes: The ratio of the length of the near-end segment to the far-end segment in the planning time domain is determined based on the gap tension index; The difference between the clearance tension index of the current control cycle and the clearance tension index stored in the previous control cycle is calculated to assess the degree of change in clearance state. Based on the degree of change in clearance state, the solution results of the previous control cycle are evaluated to determine the initial value for hot start. The solution accuracy is allocated between the proximal and distal segments based on the gap tension index; The planning time domain is divided into a near segment and a far segment according to the length ratio of the time domain. The near segment and the far segment are initialized with the initial values ​​of hot start, and a hierarchical solution structure is constructed according to the solution accuracy.

[0009] Optionally, establishing a corresponding candidate lane change trajectory model for each candidate lane change initiation time includes: The search range for lane change start delay is determined based on the gap tension index and the gap tension evolution sequence; Based on the gap tension evolution sequence, the lane change start delay is non-uniformly discretized within the search range to generate candidate lane change start times; Based on the dynamic constraint set, the feasibility of the trajectory boundary conditions corresponding to each candidate lane change start time is pre-screened to obtain a set of feasible candidate lane change start times. For each candidate lane change start time in the set of feasible candidate lane change start times, a corresponding candidate lane change trajectory model is established based on the current state of the vehicle and the target state of the lane change.

[0010] Optionally, the process of determining the trajectory parameters includes: According to the chronological order of the candidate lane change start times, the solution result of the previous candidate lane change start time is used as the initial value of the optimization subproblem corresponding to the next candidate lane change start time. Under the hierarchical solution structure, the optimization subproblems corresponding to each candidate lane change start time are solved sequentially. Based on the evolution state corresponding to the start time of each candidate lane change in the gap tension evolution sequence, solution resources are allocated to each optimization subproblem; During the solution process, graded penalty weights are applied to violations of soft constraints according to the type hierarchy of constraints in the dynamic constraint set; Output the trajectory parameters and solution status corresponding to the start time of each candidate lane change.

[0011] Optionally, the candidate lane-change start time with the lowest cost is selected from the candidate lane-change start times with feasible solutions in the solution results, and the corresponding target lane-change trajectory is output, including: The weighting coefficients of the smoothness cost term and the gap utilization efficiency cost term in the preset joint cost function are dynamically adjusted based on the gap tension index. Based on the dynamically adjusted preset joint cost function, calculate the joint cost value corresponding to each feasible candidate lane-switching start time; Based on the joint cost value and solution status of each feasible candidate lane change start time, the cross-cycle robustness of each feasible candidate lane change start time is comprehensively evaluated, and the candidate lane change start time with the best comprehensive evaluation result and its corresponding trajectory parameters are selected. Based on the trajectory parameters and the output trajectory of the previous control cycle, a smooth connection is performed to output the target lane-changing trajectory for the current control cycle.

[0012] Secondly, the present invention provides an intelligent autonomous driving vehicle lane-changing trajectory generation system, comprising: The state quantity acquisition module is used to acquire the current state quantity of the vehicle, the current state quantity of the vehicle in front of the target lane, the current state quantity of the vehicle behind the target lane, and the state quantity of the lane change target. The gap tension calculation module is used to calculate the gap tension index, which represents the current lane change gap tension, based on the current state of the vehicle, the current state of the vehicle in front in the target lane, and the current state of the vehicle behind in the target lane. The dynamic constraint construction module is used to divide the lane change trajectory optimization constraints into hard constraints and elastic soft constraints, and dynamically update the boundary values ​​of the elastic soft constraints based on the gap tension index to obtain the dynamic constraint set. The hierarchical solution structure construction module is used to divide the planning time domain into multiple solution segments to construct a hierarchical solution structure, and to perform hot-start initialization of the hierarchical solution structure based on the solution results of the previous control cycle. The candidate timing generation module is used to determine the search range of lane change start delay based on the gap tension index, generate multiple candidate lane change start times by using the delay as an explicit optimization variable, and establish a corresponding candidate lane change trajectory model for each candidate lane change start time. The trajectory parameter solving module is used to solve the optimization sub-problems corresponding to each candidate lane change start time one by one under the dynamic constraint set and hierarchical solution structure to obtain the corresponding trajectory parameters. The target trajectory output module is used to select the candidate lane-change start time with the lowest cost from the candidate lane-change start times with feasible solutions in the solution results based on a preset joint cost function, and output the corresponding target lane-change trajectory.

[0013] Thirdly, the present invention provides an electronic device, comprising: Memory, used to store computer programs; A processor is used to execute computer programs to implement the aforementioned method for generating lane-changing trajectories for intelligent autonomous vehicles.

[0014] Fourthly, the present invention provides a computer-readable storage medium for storing a computer program; wherein, when the computer program is executed by a processor, it implements the aforementioned method for generating lane-changing trajectories for intelligent autonomous vehicles.

[0015] The beneficial effects of this invention are as follows: This invention uses the lane change gap tension index and its evolution sequence as a dynamic risk-driven signal throughout the entire planning process. This synchronously drives adaptive management of lane change constraints, adaptive construction of the solution framework, prospective generation of candidate lane change times, efficient and accurate solution of multiple candidate times, and adaptive adjustment of the selection evaluation criteria, forming a complete closed-loop planning architecture from scene perception to trajectory output. This invention can dynamically adjust the planning strategy based on the real-time risk status and evolution trend of the target lane gap. In scenarios with tight gaps, it prioritizes lane change safety; in scenarios with ample gaps, it balances lane change efficiency and comfort. Simultaneously, through cross-cycle robustness assessment and smooth output transitions, it ensures the continuity of the lane change trajectory and the stability of control execution. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0017] Figure 1 This is a flowchart illustrating a method for generating lane-changing trajectories for intelligent autonomous vehicles, provided in an embodiment of the present invention.

[0018] Figure 2 This is a flowchart illustrating the hierarchical solution structure construction of a method for generating lane-changing trajectories for intelligent autonomous vehicles, as provided in an embodiment of the present invention.

[0019] Figure 3 This is a flowchart illustrating the feasibility pre-screening of candidate lane-changing start times and the establishment of a trajectory model for a lane-changing trajectory generation method for intelligent autonomous vehicles provided in an embodiment of the present invention.

[0020] Figure 4 This is a structural block diagram of an intelligent autonomous driving vehicle lane-changing trajectory generation system provided in an embodiment of the present invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] Reference Figure 1 This invention discloses a method for generating lane-changing trajectories for intelligent autonomous vehicles, which may include: S1: Obtain the current state of the vehicle, the current state of the vehicle in front of the target lane, the current state of the vehicle behind the target lane, and the target state of the lane change.

[0023] In step S1, the current state parameters of the vehicle include its current position, longitudinal velocity, longitudinal acceleration, and heading angle, which are provided in real time by the onboard positioning module and the inertial measurement unit. The current state parameters of the vehicle in front and the vehicle behind in the target lane include their respective current positions, longitudinal velocities, and longitudinal accelerations, which are obtained by the onboard perception system through sensor fusion using LiDAR, millimeter-wave radar, or cameras.

[0024] The target state variables for lane changing are determined by the upstream behavior decision layer when the lane changing decision is triggered. These variables include the target lateral position, target longitudinal velocity, and target heading angle of the vehicle at the moment the lane change is completed. The target state variables for lane changing remain unchanged within the current planning cycle and serve as the boundary condition input for subsequent trajectory parameterization solutions.

[0025] Within each control cycle, all the above state variables are updated based on the latest perception results at the current moment to ensure that the planning input remains synchronized with the actual traffic scenario.

[0026] S2: Based on the current state of the vehicle, the current state of the vehicle in front in the target lane, and the current state of the vehicle behind in the target lane, calculate the gap tension index, which characterizes the current lane change gap tension.

[0027] S2 includes the following steps: S2.1: Based on the current state of the vehicle, the current state of the vehicle in front of the target lane and the current state of the vehicle behind the target lane, predict the motion trajectory of the vehicle, the vehicle in front of the target lane and the vehicle behind the target lane in a preset prediction time domain, and calculate the relative motion state between the vehicle and the vehicle in front of the target lane and the relative motion state between the vehicle and the vehicle behind the target lane based on the motion trajectory.

[0028] Specifically, based on the current state of the vehicle, the current state of the vehicle in front of the target lane, and the current state of the vehicle behind the target lane, the trajectories of the vehicle, the vehicle in front of the target lane, and the vehicle behind the target lane are predicted within a preset prediction time domain, using the current speed and acceleration of each vehicle as initial conditions, so as to obtain the predicted position and predicted speed of the three vehicles at each time in the prediction time domain.

[0029] Based on this, the relative distance and relative speed between the vehicle and the vehicle in front in the target lane are calculated at each time point in the prediction time domain, based on the predicted position and speed of the vehicle and the vehicle in front in the target lane, to obtain the relative motion state of the vehicle in front; based on the predicted position and speed of the vehicle and the vehicle behind in the target lane, the relative distance and relative speed between the vehicle and the vehicle behind in the target lane are calculated at each time point in the prediction time domain, to obtain the relative motion state of the vehicle behind.

[0030] The length of the preset prediction time domain is determined based on the time required for a typical lane-changing scenario, with the principle of covering the complete lane-changing process.

[0031] S2.2: Calculate the time series sequence of the safety margin of the clearance between the preceding vehicles in the time domain based on the relative motion state of the preceding vehicles, and calculate the time series sequence of the safety margin of the clearance between the following vehicles in the time domain based on the relative motion state of the following vehicles.

[0032] Specifically, based on the relative distance and relative speed of the preceding vehicle at each moment in the relative motion state, the clearance safety margin between the vehicle and the preceding vehicle in the target lane at each moment in the prediction time domain is calculated, and the time series of the clearance safety margin of the preceding vehicle is obtained; based on the relative distance and relative speed of the following vehicle at each moment in the relative motion state, the clearance safety margin between the vehicle and the following vehicle in the target lane at each moment in the prediction time domain is calculated, and the time series of the clearance safety margin of the following vehicle is obtained.

[0033] S2.3: Based on the time series of the safety margin of the front vehicle clearance and the time series of the safety margin of the rear vehicle clearance, the worst-case conditions on both sides in the prediction time domain are comprehensively evaluated to obtain the clearance tension index and the clearance tension evolution sequence.

[0034] Specifically, for each discrete time point in the prediction time domain, the smaller value between the safety margin of the front vehicle clearance and the safety margin of the rear vehicle clearance at the corresponding time is extracted to form the clearance tension evolution sequence in the prediction time domain; the global minimum value is extracted from the clearance tension evolution sequence as the clearance tension index characterizing the most unfavorable overall state in the prediction time domain.

[0035] By dynamically predicting and evaluating the safety margin of clearance throughout the lane change process, this invention reflects both the current level of risk and the future trend of risk changes, effectively avoiding misjudgments of lane change risk caused by static clearance assessment based solely on the current moment.

[0036] S3: Divide the lane change trajectory optimization constraints into hard constraints and flexible soft constraints, and dynamically update the boundary values ​​of the flexible soft constraints based on the gap tension index to obtain a dynamic constraint set.

[0037] S3 includes the following steps: S3.1: Based on the correlation between lane change trajectory optimization constraints and lane change safety, lane change trajectory optimization constraints are divided into hard constraints and initial elastic soft constraints.

[0038] In one embodiment, constraints related to lane-changing safety that are inviolable are defined as hard constraints, and constraints related to lane-changing comfort and efficiency whose boundary values ​​are adjustable are defined as initial flexible soft constraints. Hard constraints include, but are not limited to, collision avoidance constraints, road boundary constraints, and vehicle dynamics physical limit constraints; initial flexible soft constraints include, but are not limited to, lane-changing comfort constraints, safety distance margin constraints with vehicles in front and behind, and lane-changing efficiency constraints.

[0039] S3.2: Compare the gap tension index with the preset tension threshold, and perform dynamic type conversion between hard constraints and elastic soft constraints based on the comparison results and the gap tension evolution sequence.

[0040] Specifically, when the gap tension index is lower than the preset tension threshold or the gap tension evolution sequence indicates that the gap is continuously narrowing, the safety-related constraints in the initial elastic soft constraints are upgraded to hard constraints; when the gap tension index is not lower than the preset tension threshold and the gap tension evolution sequence indicates that the gap is continuously expanding, the constraints converted to hard constraints in the above upgrade operation are downgraded back to elastic soft constraints, and the original hard constraints do not participate in the downgrade operation.

[0041] In this step, the preset tension threshold is a pre-calibrated constant used to define the tension boundary of the current lane-changing scenario. The preset tension threshold is determined based on the safety margin distribution under typical lane-changing scenarios, and its value is consistent with the dimension of the clearance safety margin.

[0042] S3.3: For elastic soft constraints that have not undergone type conversion, calculate and update their boundary values ​​based on the gap tension index.

[0043] Furthermore, based on the ratio of the gap tension index to the preset tension threshold, the boundary adjustment coefficient of the elastic soft constraint is calculated. The larger the gap tension index, the larger the boundary adjustment coefficient. The boundary adjustment coefficient is applied to the initial boundary value of each elastic soft constraint to obtain the updated elastic soft constraint boundary value. The larger the boundary adjustment coefficient, the looser the elastic soft constraint boundary value, and the smaller the boundary adjustment coefficient, the tighter the elastic soft constraint boundary value.

[0044] It should be noted that the calculation method of the boundary adjustment coefficient satisfies a monotonic mapping relationship with the gap tension index, ensuring that the boundary value of the elastic soft constraint is continuously and smoothly adjusted with the change of gap tension, and avoiding instability in trajectory solution caused by boundary value jumps.

[0045] S3.4: Construct a dynamic constraint set based on the hard constraints, flexible soft constraints, and updated boundary values ​​after type conversion.

[0046] By dynamically upgrading and downgrading the lane-changing trajectory optimization constraints and adaptively adjusting the boundary values, this invention ensures the execution of safety-critical constraints when the gap is tight, releases constraint redundancy to improve lane-changing efficiency when the gap is loose, and ensures that the constraint boundary values ​​change continuously and smoothly to maintain the stability of trajectory solution.

[0047] S4: Divide the planning time domain into multiple solution segments to construct a hierarchical solution structure, and perform a hot start initialization of the hierarchical solution structure based on the solution results of the previous control cycle.

[0048] The flowchart for constructing the hierarchical solution structure is as follows: Figure 2 As shown, S4 includes the following steps: S4.1: Determine the ratio of the time domain length of the near-end segment to the far-end segment in the planning time domain based on the gap tension index.

[0049] Among them, the smaller the gap tension index, the larger the proportion of the time domain length of the near-end segment. This is because the tighter the gap, the more concentrated the safety-critical time domain of the lane-changing decision is in the near term. Allocating more planning time domain to the near-end segment helps improve the solution quality of the near-term trajectory and ensures the safety of lane changing.

[0050] S4.2: Calculate the difference between the clearance tension index of the current control cycle and the clearance tension index stored in the previous control cycle to assess the degree of change in clearance state, and evaluate the quality of the solution results of the previous control cycle based on the degree of change in clearance state to determine the initial value for hot start.

[0051] Specifically, the absolute value of the difference between the gap tension index of the current control cycle and the gap tension index stored in the previous control cycle is calculated as a quantitative representation of the degree of gap state change. Confidence weights are assigned to the solution results of the previous control cycle based on the degree of gap state change. The greater the degree of gap state change, the lower the confidence weight of the solution results of the previous cycle, and the closer the initial value for hot start is to the preset default initial value. Conversely, the smaller the degree of gap state change, the higher the confidence weight, and the closer the initial value for hot start is to directly use the solution results of the previous cycle. The solution results of the previous control cycle and the preset default initial value are weighted and fused based on the confidence weights, and the fused result is used as the initial value for hot start.

[0052] In the first control cycle of system operation, if there is no solution result from the previous control cycle, a cold start is performed using the preset default initial value.

[0053] S4.3: Assigning solution accuracy to the proximal and distal segments based on the gap tension index.

[0054] Among these, the smaller the gap tension index, the higher the solution accuracy of the near-end segment and the lower the solution accuracy of the far-end segment. This is because when the gap is tight, the trajectory quality of the near-end segment directly determines the safety of lane changing. Improving the solution accuracy of the near-end segment can prioritize the planning quality in the safety-critical time domain within a limited computing power budget.

[0055] S4.4: Divide the planning time domain into a near segment and a far segment according to the length ratio of the time domain. Initialize the near segment and the far segment with the initial value of hot start, and construct a hierarchical solution structure according to the solution accuracy.

[0056] Specifically, the optimization variables of the near-end segment and the far-end segment are initialized with the initial values ​​of the hot start. When the time domain length ratio of adjacent control cycles changes, the solution results of the previous control cycle are interpolated and aligned according to the time domain division of the current cycle, and then used to initialize the optimization variables of the near-end segment and the far-end segment respectively. The corresponding solver parameters are configured for the near-end segment and the far-end segment according to the solution accuracy to build a hierarchical solution structure.

[0057] By dynamically allocating the time domain length and solution accuracy of the near-end segment and the far-end segment, and adaptively adjusting the initial value of the hot start, this invention prioritizes the solution quality of the safety-critical time domain when the gap is tight, while avoiding the initial error of the hot start caused by sudden changes in the scene state, thus balancing solution quality and planning real-time performance within a limited computing power budget.

[0058] S5: Determine the search range of lane change start delay based on the gap tension index, generate multiple candidate lane change start times by using the delay as an explicit optimization variable, and establish a corresponding candidate lane change trajectory model for each candidate lane change start time.

[0059] The flowchart for feasibility pre-screening and trajectory model establishment of candidate lane change start times is as follows: Figure 3 As shown, S5 includes the following steps: S5.1: Determine the search range for lane change start delay based on the gap tension index and gap tension evolution sequence.

[0060] Specifically, the smaller the gap tension index or the smaller the gap tension evolution sequence indicating a continuous narrowing of the gap, the smaller the search range. A smaller gap tension index indicates a higher risk of a lane change gap, requiring a lane change decision to be completed as soon as possible; when the gap tension evolution sequence indicates a continuous narrowing of the gap, the feasibility of late-starting candidate times decreases in actual execution, and both require narrowing the search range to eliminate invalid candidates.

[0061] S5.2: Based on the gap tension evolution sequence, the lane change start delay is non-uniformly discretized within the search range to generate candidate lane change start times.

[0062] In one embodiment, based on the size and trend of the gap safety margin at each moment in the gap tension evolution sequence, the time period with a relatively large gap safety margin and a stable trend is marked as a favorable time period, and the time period with a relatively small gap safety margin or a continuously narrowing gap is marked as an unfavorable time period; within the favorable time period, candidate lane change start times are generated with a smaller step size, and within the unfavorable time period, candidate lane change start times are generated with a larger step size.

[0063] S5.3: Based on the dynamic constraint set, the feasibility of the trajectory boundary conditions corresponding to each candidate lane change start time is pre-screened to obtain a set of feasible candidate lane change start times.

[0064] Specifically, for each candidate lane change start time, based on the current state of the vehicle, the target state of the lane change, and the hard constraints in the dynamic constraint set, the necessary conditions of the corresponding trajectory boundary conditions are quickly checked to determine whether they have the necessary conditions for the existence of a feasible solution under the hard constraints. Candidate lane change start times whose boundary conditions are necessarily infeasible under the hard constraints are eliminated to ensure the real-time performance of the pre-screening process.

[0065] S5.4: For each candidate lane change start time in the set of feasible candidate lane change start times, establish a corresponding candidate lane change trajectory model based on the current state of the vehicle and the target state of the lane change.

[0066] Furthermore, using the current state variable of the vehicle as the boundary condition for the starting point of the trajectory and the target state variable for lane changing as the boundary condition for the ending point of the trajectory, a parameterized trajectory expression describing the lane changing process is established; using the total lane changing time as the optimization variable to be solved in the parameterized trajectory expression, the corresponding candidate lane changing trajectory model is obtained.

[0067] By employing a triple mechanism of adaptive narrowing of the search range driven by two factors, non-uniform discretization, and pre-screening of boundary condition feasibility, this invention achieves adaptive generation of the candidate lane-changing start time set, effectively reducing the solution burden of invalid candidate times and concentrating computing power on the most favorable time period of the gap state, thereby improving the capture probability of the optimal lane-changing opportunity.

[0068] S6: Under the dynamic constraint set and hierarchical solution structure, solve the optimization sub-problems corresponding to each candidate lane change start time one by one to obtain the corresponding trajectory parameters.

[0069] S6 includes the following steps: S6.1: According to the time sequence of the candidate lane change start times, the solution result of the previous candidate lane change start time is used as the initial value of the optimization subproblem corresponding to the next candidate lane change start time. Under the hierarchical solution structure, the optimization subproblems corresponding to each candidate lane change start time are solved sequentially.

[0070] Specifically, for the first candidate lane-change start time in the time sequence, the initial values ​​determined in S4.2 are used for initialization and solution. For each subsequent candidate lane-change start time, the solution result of the previous candidate lane-change start time is used as the initial value for hot start. The correlation between the optimization subproblems in the solution space of adjacent candidate lane-change start times is utilized to reduce the number of solution iterations and improve the overall solution efficiency. It should be noted that when the interval between adjacent candidate lane-change start times is large, the correlation between the corresponding optimization subproblems in the solution space is reduced, and the acceleration effect of hot start is correspondingly weakened, but it can still effectively reduce the number of solution iterations compared to cold start.

[0071] S6.2: Allocate solution resources for each optimization subproblem based on the evolution state corresponding to the start time of each candidate lane change in the gap tension evolution sequence.

[0072] The more favorable the gap evolution state of a candidate lane-change start time, the more solution resources are allocated to it. Solution resources include the upper limit of the number of solver iterations and the convergence accuracy requirement. The more favorable the gap evolution state of a candidate lane-change start time, the higher the upper limit of the number of iterations and the stricter the convergence accuracy requirement, so as to ensure the solution quality of trajectory parameters corresponding to the optimal lane-change timing.

[0073] It should be noted that the resource allocation for solving each candidate lane-switching start time is carried out under the constraint of the total computing power budget to ensure that the overall solution time of all candidate times meets the real-time requirements.

[0074] S6.3: During the solution process, apply graded penalty weights to soft constraint violations according to the type hierarchy of constraints in the dynamic constraint set.

[0075] Specifically, during the optimization process, the solver sums the soft constraint violations by weighted summation and incorporates them into the objective function. The penalty weights for flexible soft constraints related to safety are higher than those for flexible soft constraints related to comfort and efficiency. This causes the solver to naturally prioritize satisfying safety-related constraints when weighing the priority of satisfying soft constraints.

[0076] S6.4: Output the trajectory parameters and solution status corresponding to the start time of each candidate lane change.

[0077] Specifically, after solving the optimization subproblems corresponding to each candidate lane change start time, the trajectory parameters corresponding to each candidate lane change start time are output; at the same time, the solution status corresponding to each candidate lane change start time is recorded, including whether there is a feasible solution that satisfies the set of dynamic constraints and the corresponding objective function value, for S7 to use in optimization.

[0078] By leveraging the synergistic effects of a series-based hot start, adaptive allocation of solution resources, and hierarchical penalty weights, this invention achieves efficient and accurate solutions within a multi-candidate time evaluation framework. Under the constraint of total computing power budget, it prioritizes the quality of trajectory parameter solutions for favorable moments in the gap state, while ensuring the priority satisfaction of safety-related constraints. It balances solution efficiency and trajectory safety within a limited computing power budget.

[0079] S7: Based on the preset joint cost function, select the candidate lane-changing start time with the minimum cost from the candidate lane-changing start times with feasible solutions in the solution results, and output the corresponding target lane-changing trajectory.

[0080] S7 includes the following steps: S7.1: Based on the gap tension index, dynamically adjust the weight coefficients of the smoothness cost term and the gap utilization efficiency cost term in the preset joint cost function.

[0081] Among them, the smaller the gap tension index, the larger the weight coefficient of the gap utilization efficiency cost term and the smaller the weight coefficient of the smoothness cost term.

[0082] Specifically, the pre-defined joint cost function is composed of a weighted sum of a smoothness cost term and a gap utilization efficiency cost term. The smoothness cost term represents the lateral acceleration and lateral jerk of the lane-changing trajectory. The gap utilization efficiency cost term is constructed as the reciprocal of the gap margin between the vehicle and the vehicles in front and behind in the target lane at the moment of lane-changing completion. The smaller the gap margin, the larger the value of the cost term, indicating that the gap utilization efficiency after lane-changing is lower. The smaller the gap tension index, the higher the risk of lane-changing gap. The weight coefficient of the gap utilization efficiency cost term is increased to prioritize the safe passage of the trajectory, while the weight coefficient of the smoothness cost term is decreased. The larger the gap tension index, the more sufficient the safety margin. The weight coefficient of the smoothness cost term is increased to prioritize the comfort of lane-changing.

[0083] S7.2: Based on the dynamically adjusted preset joint cost function, calculate the joint cost value corresponding to the start time of each feasible candidate lane change.

[0084] For each feasible candidate lane change start time output by S6, the corresponding trajectory parameters are substituted into the preset joint cost function to calculate the smoothness cost term and the gap utilization efficiency cost term, and then summed according to the corresponding weight coefficients to obtain the joint cost value corresponding to each candidate lane change start time.

[0085] S7.3: Based on the joint cost value and solution status of each feasible candidate lane change start time, comprehensively evaluate the cross-cycle robustness of each feasible candidate lane change start time, and select the candidate lane change start time with the best comprehensive evaluation result and its corresponding trajectory parameters.

[0086] Furthermore, cross-cycle robustness is evaluated based on the gap evolution trend corresponding to each candidate lane change start time in the gap tension evolution sequence output by S2.3. Candidate lane change start times with a continuously favorable gap evolution trend can remain feasible in the next control cycle, and their cross-cycle robustness is relatively high. In the comprehensive evaluation, the joint cost value is the main selection criterion, and cross-cycle robustness is the auxiliary screening condition. When the joint cost difference between two candidate lane change start times is within the preset cost difference threshold range, the one with higher cross-cycle robustness is selected first, and the candidate lane change start time and its trajectory parameters with the best comprehensive evaluation result are selected.

[0087] S7.4: Based on the trajectory parameters and the output trajectory of the previous control cycle, perform smooth connection processing and output the target lane-changing trajectory of the current control cycle.

[0088] It should be noted that the smooth transition is carried out within the time overlap period between the trajectory of the previous control cycle and the trajectory of the current control cycle. By using boundary constraints, the state of the starting segment of the current cycle trajectory and the trajectory of the previous cycle is kept in a continuous transition during the overlapping period, eliminating the trajectory discontinuity caused by the jump in the cycle optimization result and ensuring a smooth transition of the control execution layer.

[0089] By dynamically adjusting the cost function weights, performing robust joint optimization across cycles, and smoothly connecting the output, this invention enables the evaluation criteria to adaptively switch according to scenario risks. When gaps are tight, it prioritizes ensuring safe passage and when gaps are loose, it prioritizes ensuring lane-changing comfort. Furthermore, by using robustness assessment based on gap evolution trends, it avoids control jitter caused by jumps in optimization results, thereby improving the overall safety and continuity of lane-changing planning.

[0090] Accordingly, refer to Figure 4 This invention also provides an intelligent autonomous driving vehicle lane-changing trajectory generation system, which may include: The state quantity acquisition module is used to acquire the current state quantity of the vehicle, the current state quantity of the vehicle in front of the target lane, the current state quantity of the vehicle behind the target lane, and the state quantity of the lane change target. The gap tension calculation module is used to calculate the gap tension index, which represents the current lane change gap tension, based on the current state of the vehicle, the current state of the vehicle in front in the target lane, and the current state of the vehicle behind in the target lane. The dynamic constraint construction module is used to divide the lane change trajectory optimization constraints into hard constraints and elastic soft constraints, and dynamically update the boundary values ​​of the elastic soft constraints based on the gap tension index to obtain the dynamic constraint set. The hierarchical solution structure construction module is used to divide the planning time domain into multiple solution segments to construct a hierarchical solution structure, and to perform hot-start initialization of the hierarchical solution structure based on the solution results of the previous control cycle. The candidate timing generation module is used to determine the search range of lane change start delay based on the gap tension index, generate multiple candidate lane change start times by using the delay as an explicit optimization variable, and establish a corresponding candidate lane change trajectory model for each candidate lane change start time. The trajectory parameter solving module is used to solve the optimization sub-problems corresponding to each candidate lane change start time one by one under the dynamic constraint set and hierarchical solution structure to obtain the corresponding trajectory parameters. The target trajectory output module is used to select the candidate lane-change start time with the lowest cost from the candidate lane-change start times with feasible solutions in the solution results based on a preset joint cost function, and output the corresponding target lane-change trajectory.

[0091] This invention also provides a computer device, including a memory and a processor, wherein the memory stores a computer program; when the computer program is executed by the processor, the processor performs the steps of the intelligent autonomous driving vehicle lane-changing trajectory generation method as described in any of the above embodiments.

[0092] This invention also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the intelligent autonomous driving vehicle lane-changing trajectory generation method as described in any of the above embodiments.

[0093] The computer-readable storage medium may include various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0094] For a description of the computer-readable storage medium provided by the present invention, please refer to the above method embodiments; the present invention will not be described in detail here.

[0095] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to in the method section.

[0096] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0097] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for generating lane-changing trajectories for intelligent autonomous vehicles, characterized in that, include: Obtain the current state of the vehicle, the current state of the vehicle in front of the target lane, the current state of the vehicle behind the target lane, and the target state of the lane change. Based on the current state of the vehicle, the current state of the vehicle in front of the target lane, and the current state of the vehicle behind the target lane, calculate the gap tension index, which characterizes the current lane change gap tension. The lane-changing trajectory optimization constraints are divided into hard constraints and flexible soft constraints. The boundary values ​​of the flexible soft constraints are dynamically updated based on the gap tension index to obtain a dynamic constraint set. The planning time domain is divided into multiple solution segments to construct a hierarchical solution structure, and the hierarchical solution structure is hot-started and initialized based on the solution results of the previous control cycle. The search range for lane change start delay is determined based on the gap tension index. The delay is used as an explicit optimization variable to generate multiple candidate lane change start times, and a corresponding candidate lane change trajectory model is established for each candidate lane change start time. Under the dynamic constraint set and hierarchical solution structure, the optimization sub-problems corresponding to each candidate lane change start time are solved one by one to obtain the corresponding trajectory parameters; Based on the preset joint cost function, the candidate lane-changing start time with the minimum cost is selected from the candidate lane-changing start times with feasible solutions in the solution results, and the corresponding target lane-changing trajectory is output.

2. The method for generating lane-changing trajectories for intelligent autonomous vehicles according to claim 1, characterized in that, The gap tension index, which characterizes the current lane change gap tension, includes: Based on the current state of the vehicle, the current state of the vehicle in front of the target lane and the current state of the vehicle behind the target lane, predict the motion trajectory of the vehicle, the vehicle in front of the target lane and the vehicle behind the target lane in a preset prediction time domain, and calculate the relative motion state between the vehicle and the vehicle in front of the target lane and the relative motion state between the vehicle and the vehicle behind the target lane based on the motion trajectory. The time series sequence for calculating the safety margin of the clearance between the preceding and following vehicles in the time domain is calculated based on the relative motion state of the preceding vehicle, and the time series sequence for calculating the safety margin of the clearance between the following and following vehicles in the time domain is calculated based on the relative motion state of the following vehicle. Based on the time series of the front vehicle clearance safety margin and the rear vehicle clearance safety margin, the worst-case conditions on both sides in the prediction time domain are comprehensively evaluated to obtain the clearance tension index and the clearance tension evolution sequence.

3. The method for generating lane-changing trajectories for intelligent autonomous vehicles according to claim 1, characterized in that, The process of determining the dynamic constraint set includes: Based on the correlation between lane change trajectory optimization constraints and lane change safety, lane change trajectory optimization constraints are divided into hard constraints and initial elastic soft constraints. The gap tension index is compared with the preset tension threshold, and the hard constraint and elastic soft constraint are dynamically converted based on the comparison results and the gap tension evolution sequence. For elastic soft constraints that have not undergone type conversion, their boundary values ​​are calculated and updated based on the gap tension index. A dynamic constraint set is constructed based on the hard constraints after type conversion, the flexible soft constraints, and the updated boundary values.

4. The method for generating lane-changing trajectories for intelligent autonomous vehicles according to claim 1, characterized in that, The construction and hot-start initialization process of the hierarchical solution structure includes: The ratio of the length of the near-end segment to the far-end segment in the planning time domain is determined based on the gap tension index; The difference between the gap tension index of the current control cycle and the gap tension index stored in the previous control cycle is calculated to assess the degree of change in gap state. Based on the degree of change in gap state, the solution results of the previous control cycle are evaluated to determine the initial value for hot start. The solution accuracy is allocated between the proximal and distal segments based on the gap tension index; The planning time domain is divided into a near segment and a far segment according to the time domain length ratio. The near segment and the far segment are initialized with the hot start initial value, and a hierarchical solution structure is constructed according to the solution accuracy.

5. The method for generating lane-changing trajectories for intelligent autonomous vehicles according to claim 1, characterized in that, The step of establishing a corresponding candidate lane change trajectory model for each candidate lane change initiation time includes: The search range for lane change start delay is determined based on the gap tension index and the gap tension evolution sequence; Based on the gap tension evolution sequence, the lane change start delay is non-uniformly discretized within the search range to generate candidate lane change start times. Based on the dynamic constraint set, the feasibility of the trajectory boundary conditions corresponding to each candidate lane change start time is pre-screened to obtain a set of feasible candidate lane change start times. For each candidate lane change start time in the set of feasible candidate lane change start times, a corresponding candidate lane change trajectory model is established based on the current state of the vehicle and the target state of the lane change.

6. The method for generating lane-changing trajectories for intelligent autonomous vehicles according to claim 1, characterized in that, The process of determining the trajectory parameters includes: According to the chronological order of the candidate lane change start times, the solution result of the previous candidate lane change start time is used as the initial value of the optimization subproblem corresponding to the next candidate lane change start time. Under the hierarchical solution structure, the optimization subproblems corresponding to each candidate lane change start time are solved sequentially. Based on the evolution state corresponding to the start time of each candidate lane change in the gap tension evolution sequence, solution resources are allocated to each optimization subproblem; During the solution process, graded penalty weights are applied to violations of soft constraints according to the type hierarchy of constraints in the dynamic constraint set; Output the trajectory parameters and solution status corresponding to the start time of each candidate lane change.

7. The method for generating lane-changing trajectories for intelligent autonomous vehicles according to claim 1, characterized in that, The step of selecting the candidate lane-changing start time with the lowest cost from the candidate lane-changing start times with feasible solutions in the solution results and outputting the corresponding target lane-changing trajectory includes: The weighting coefficients of the smoothness cost term and the gap utilization efficiency cost term in the preset joint cost function are dynamically adjusted based on the gap tension index. Based on the dynamically adjusted preset joint cost function, calculate the joint cost value corresponding to each feasible candidate lane-switching start time; Based on the joint cost value and solution status of each feasible candidate lane change start time, the cross-cycle robustness of each feasible candidate lane change start time is comprehensively evaluated, and the candidate lane change start time with the best comprehensive evaluation result and its corresponding trajectory parameters are selected. Based on the trajectory parameters, a smooth connection is performed with the output trajectory of the previous control cycle to output the target lane-changing trajectory for the current control cycle.

8. A system for generating lane-changing trajectories for intelligent autonomous vehicles, characterized in that, include: The state quantity acquisition module is used to acquire the current state quantity of the vehicle, the current state quantity of the vehicle in front of the target lane, the current state quantity of the vehicle behind the target lane, and the state quantity of the lane change target. The gap tension calculation module is used to calculate the gap tension index, which represents the current lane change gap tension, based on the current state of the vehicle, the current state of the vehicle in front in the target lane, and the current state of the vehicle behind in the target lane. The dynamic constraint construction module is used to divide the lane change trajectory optimization constraints into hard constraints and elastic soft constraints, and dynamically update the boundary values ​​of the elastic soft constraints based on the gap tension index to obtain the dynamic constraint set. The hierarchical solution structure construction module is used to divide the planning time domain into multiple solution segments to construct a hierarchical solution structure, and to perform hot-start initialization of the hierarchical solution structure based on the solution results of the previous control cycle. The candidate timing generation module is used to determine the search range of lane change start delay based on the gap tension index, generate multiple candidate lane change start times by using the delay as an explicit optimization variable, and establish a corresponding candidate lane change trajectory model for each candidate lane change start time. The trajectory parameter solving module is used to solve the optimization sub-problems corresponding to each candidate lane change start time one by one under the dynamic constraint set and hierarchical solution structure to obtain the corresponding trajectory parameters. The target trajectory output module is used to select the candidate lane-change start time with the lowest cost from the candidate lane-change start times with feasible solutions in the solution results based on a preset joint cost function, and output the corresponding target lane-change trajectory.

9. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor is configured to execute the computer program to implement the intelligent autonomous driving vehicle lane-changing trajectory generation method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, Used to store a computer program, wherein the computer program, when executed by a processor, implements the intelligent autonomous driving vehicle lane-changing trajectory generation method as described in any one of claims 1 to 7.

Images