Multi-objective dynamic weight path planning method based on confidence feedback

Abstract

This invention belongs to the field of autonomous driving technology. It provides a multi-objective dynamic weighted path planning method based on confidence feedback, comprising: acquiring an environmental information set of targets; calculating the comprehensive confidence of each detected target in the environmental information set, and then determining a key object set using a target importance scoring function; determining the scene comprehensive confidence based on the key object set; determining the weight parameters, safety cost term, efficiency cost term, and comfort cost term of the comprehensive cost function based on the scene comprehensive confidence; determining the optimal control sequence of the autonomous vehicle in future time steps using an OSQP solver based on the current state of the autonomous vehicle and the comprehensive cost function, and minimizing the multi-objective weighted cost function; and performing periodic control processing on the autonomous vehicle based on the optimal control sequence in future time steps. The beneficial effects of this invention are: improved safety, efficiency, and comfort of autonomous vehicles.

Description

Technical Field

[0001] This invention relates to the field of autonomous driving technology, and in particular to a multi-objective dynamic weighted path planning method based on confidence feedback. Background Technology

[0002] With the continuous development of autonomous driving technology, the application of autonomous vehicles in various traffic scenarios is gradually increasing. Path planning, as one of the core technologies in autonomous driving systems, aims to calculate the optimal driving path based on real-time environmental information to ensure vehicle safety, efficiency, and comfort. Path planning systems not only need to consider static factors, such as road topology and obstacle locations, but also dynamic factors, such as the impact of other road users, traffic signal changes, and unexpected situations. Therefore, how to perform efficient and safe path planning in complex and dynamic traffic environments has become an important issue in the development of autonomous driving technology.

[0003] With the rapid development of sensing technology, vehicle-to-everything (V2X) communication, and intelligent optimization algorithms, research on path planning for autonomous driving systems is gradually evolving from traditional static optimization models to more complex dynamic multi-objective optimization models. Traditional static path planning typically optimizes only a single objective (such as the shortest distance or shortest time), neglecting the multi-dimensional needs during driving. Multi-objective optimization models, however, can simultaneously consider multiple factors such as safety, driving efficiency, and passenger comfort, and combine dynamic sensing data with real-time environmental information to achieve adaptive adjustments to the planning strategy, thus better reflecting real-world driving needs. For example, in their paper "Ant Colony Hybrid Algorithm Optimization for Path Planning of Autonomous Sweeping Vehicles [J]. Equipment Management and Maintenance, 2025(7):42-46," Li et al. proposed a multi-objective optimization ant colony hybrid algorithm to address the problems of the rigid pheromone mechanism and singular heuristic function in traditional ant colony algorithms for path planning of autonomous sweeping vehicles in industrial parks, making them difficult to adapt to complex outdoor environments. In their paper "Road Multi-Target Detection Algorithm for Unmanned Driving Scenarios [J]. Computer Applications and Software, 2024, 41(8): 283–288", Niu Wenjie et al. proposed an improved YOLOv3 road multi-target detection algorithm to solve the problems of high false detection rate and poor detection effect of small and occluded targets in the traditional YOLOv3 in unmanned driving scenarios. In their paper "Multi-Target Detection of Underground Unmanned Electric Locomotives Based on DYCS-YOLOv8n [J]. Industrial and Mining Automation, 2025, 51(4): 86-92, 130", Xu Jinhui et al. proposed a multi-target detection model of underground unmanned electric locomotives based on DYCS-YOLOv8n to solve the problems of difficult image feature extraction and difficult identification of small targets caused by low light, high noise and motion blur in underground environments.

[0004] While the aforementioned research has driven the rapid development of autonomous driving path planning technology, significant shortcomings remain in real-world, complex traffic environments. First, existing path planning methods generally assume completely reliable perception information, failing to adequately consider the uncertainties of the perception system. When sensors are affected by rain, fog, strong light, or occlusion, their outputs may deviate. If the planning module still makes decisions based on such low-confidence data, it can easily lead to unsafe paths or control anomalies. Second, existing multi-objective optimization models typically employ fixed weights, failing to dynamically adjust optimization preferences based on real-time environment and perception confidence. For example, under low-confidence conditions, the system should prioritize safety; while in high-confidence environments, efficiency and comfort weights can be appropriately increased. Fixed-weight strategies lack adaptability, making it difficult for vehicles to maintain optimal driving performance under different confidence scenarios. Summary of the Invention

[0005] In order to at least solve one of the technical problems existing in the prior art, the present invention provides a multi-objective dynamic weighted path planning method based on confidence feedback.

[0006] This invention provides a multi-objective dynamic weighted path planning method based on confidence feedback, comprising: The system acquires environmental perception data from autonomous vehicles, determines the state data of multiple detection targets based on the environmental perception data, and fuses the state data with timestamps to obtain an environmental information set of multiple detection targets. Calculate the overall confidence level of each detected target in the environmental information set; Based on the current state of the autonomous vehicle, the relative state of each detected target with respect to the autonomous vehicle is calculated. Based on the relative state, the set of key objects is determined using a target importance scoring function. The overall confidence level of a scenario is determined based on a set of key objects. This overall confidence level is used to characterize the global trust that autonomous vehicles have in environmental perception data. Based on the overall confidence level of the scenario, the weight parameters of the overall cost function are determined, whereby the overall cost function is used to characterize the optimal trajectory under the constraints of safety cost, efficiency cost, and comfort cost. Based on the current state of the autonomous vehicle, the safety cost term, efficiency cost term, and comfort cost term of the comprehensive cost function are determined using a discrete bicycle model and corresponding constraint conditions. Based on the current state and comprehensive cost function of the autonomous vehicle, the minimum multi-objective weighted cost function is determined, and the OSQP solver is used to determine the optimal control sequence of the autonomous vehicle at future time steps based on the minimum multi-objective weighted cost function. Based on the linearized minimization of the multi-objective weighted cost function and the current state of the autonomous vehicle, the OSQP solver is used to determine the optimal control sequence of the autonomous vehicle at future time steps. Based on the optimal control sequence for future moments, periodic control processes are performed on the autonomous vehicle.

[0007] According to the aforementioned multi-objective dynamic weighted path planning method based on confidence feedback, environmental perception data of an autonomous vehicle is acquired, state data of multiple detection targets are determined based on the environmental perception data, and the state data is fused with timestamps to obtain an environmental information set of multiple detection targets, including: Environmental perception data from multiple sensors installed on autonomous vehicles is acquired, and the data is fused with timestamps to obtain an environmental information set for multiple detected targets:

[0008]

[0009] in, For the set of environmental objectives, The total number of environmental targets. This serves as the serial number identifier for the target being detected. Indicates time; To detect the target's state data, including velocity vectors and position vector in global coordinate system , For transpose, The covariance matrix for detecting the position of a target is used to represent the uncertainty of geometric measurements. This represents the target detection confidence level determined by the perception algorithm.

[0010] According to the aforementioned multi-objective dynamic weighted path planning method based on confidence feedback, the method further includes: The environmental information set of multiple detection targets is preprocessed and anomaly corrected, including anomaly detection and filtering of the environmental information set of multiple detection targets, and removal of distorted or missing frame information; if an abnormal measurement point is detected, the missing state of the abnormal measurement point is reconstructed by neighborhood interpolation or Kalman filtering.

[0011] According to the aforementioned multi-objective dynamic weighted path planning method based on confidence feedback, the calculation of the comprehensive confidence of each detected object in the environmental information set includes: Based on the environmental information set of multiple detection targets, a weighted fusion model of detection probability and geometric variance is used to determine the overall confidence level of the targets. for:

[0012] in, The trace operation represents the matrix operation, used to characterize the overall measurement variance; This is a weighting coefficient used to represent the relative importance of target detection confidence and geometric stability; This is the attenuation coefficient, used to map the geometric variance to the confidence space; A time-series smoothing strategy is used to exponentially weight the target confidence level for updating, resulting in the time step. Target overall confidence level for:

[0013] in, For smoothing coefficients; And, it also includes the time. Target overall confidence level Constraints are applied using a single-step maximum descent limit; the time after constraint... Target overall confidence level for:

[0014] in, This represents the maximum confidence decrease in a single step, and .

[0015] According to the aforementioned multi-objective dynamic weighted path planning method based on confidence feedback, the relative state of each detected target to the autonomous vehicle is calculated based on the current state of the autonomous vehicle. Based on the relative state, an importance function is used to determine the set of key objects, including: The target importance scoring function is:

[0016] in, The scoring result of the target importance scoring function; Location of autonomous vehicles With the detection target location European distance, For the speed of autonomous vehicles, For detecting the speed of the target; position of the autonomous vehicle. The speed of autonomous vehicles The heading angle of the autonomous vehicle constitutes the current state of the autonomous vehicle; As distance difference weight, As the speed difference weight, For the heading angle difference weight, To prevent small constants from being divided by zero, To detect the difference in heading angle between the target and the autonomous vehicle; Based on the scoring results of the target importance scoring function Select Each detection target yields a set of key objects. .

[0017] According to the aforementioned multi-objective dynamic weighted path planning method based on confidence feedback, determining the overall confidence level of the scenario based on the set of key objects includes: Based on the key object set Calculate the average confidence level of the scenario and minimum confidence :

[0018] Based on the average confidence level of the scenario and minimum confidence Calculate the overall confidence level of the scenario for:

[0019] in, These are the weighting coefficients.

[0020] According to the aforementioned multi-objective dynamic weighted path planning method based on confidence feedback, the weight parameters of the comprehensive cost function are determined based on the comprehensive confidence of the scenario. The comprehensive cost function characterizes the optimal trajectory under constraints of safety cost, efficiency cost, and comfort cost, and includes: The comprehensive cost function is:

[0021] in, This is a safety cost term, used to characterize the risk of collision between the vehicle and an obstacle. For security weight parameters; This is an efficiency cost term, used to characterize the path length, time, or energy consumption of a vehicle's journey. For efficiency weighting parameters; This is a comfort cost term, used to characterize the smoothness of vehicle acceleration and steering changes. For comfort weighting parameters; , and All are greater than 0. ,and , and These are dynamic weight parameters; A monotonic mapping model using the Sigmoid function is used to map the overall scene confidence to... Get the time Security weights for:

[0022] in, The steepness coefficient of the function determines the sensitivity of the overall confidence level of the scene to the weight adjustment. This is the balance threshold; Based on security weight Determine the time Efficiency weight With comfort weight for:

[0023] in, This is the ratio coefficient between efficiency and comfort.

[0024] According to the aforementioned multi-objective dynamic weighted path planning method based on confidence feedback, the safety cost term, efficiency cost term, and comfort cost term of the comprehensive cost function are determined based on the current state of the autonomous vehicle using a discrete bicycle model and corresponding constraint conditions, including: The safety cost term, efficiency cost term, and comfort cost term of the comprehensive cost function are determined separately, including determining the safety cost term based on the obstacle potential field model and collision time. for:

[0025] in, For autonomous vehicles and the first Each detection target at time The Euclidean distance; This is the safe distance threshold; To prevent division by zero of small constants; This is the trade-off factor between the safety distance and the TTC term; The collision time prediction between the vehicle and the target is expressed as:

[0026] Determine the efficiency cost item based on the overall performance of path length, time, and energy consumption. for:

[0027] in, For autonomous vehicles at all times Location; It is longitudinal acceleration; For a long walk; , and These are the path length weight, time weight, and energy consumption weight, respectively. The comfort penalty is determined based on the vehicle's rapid acceleration, sharp turns, or frequent changes in handling. for:

[0028] in, It is longitudinal acceleration; For path curvature; , and These are acceleration weight, jerk weight, and curvature weight, respectively.

[0029] According to the aforementioned multi-objective dynamic weighted path planning method based on confidence feedback, a multi-objective weighted cost function is minimized based on the current state and comprehensive cost function of the autonomous vehicle. Then, an OSQP solver is used to determine the optimal control sequence for the autonomous vehicle at future time steps based on the minimized multi-objective weighted cost function, including: Obtain the real-time information of autonomous vehicles state vector and control input vector for:

[0030]

[0031] in, These are the position coordinates of the autonomous vehicle in the global coordinate system. The heading angle of the autonomous vehicle; The longitudinal speed of the autonomous vehicle; For autonomous vehicles, the longitudinal acceleration represents the speed change control quantity; The front wheel steering angle represents the steering control amount. Based on the state vector and control input vector The state transition relationship of an autonomous vehicle between adjacent time points is determined as follows:

[0032] in, The discretized nonlinear dynamic equations for the state of an autonomous vehicle changing over time are expressed as:

[0033] The control and state constraints are as follows:

[0034] in, To minimize the speed limit, Maximum speed limit; To limit the acceleration to the maximum; For the maximum turning angle, This is the upper limit of the rate of change of the rotation angle; With time state vector As an initial condition, determine the future. Step state sequence and control input sequence for:

[0035]

[0036] Determining the future based on the comprehensive cost function The minimization of the multi-objective weighted cost function for the step is:

[0037] Obtain the state points of the current state of the autonomous vehicle. Based on the state points of autonomous vehicles Certainty and the Future The minimization of the multi-objective weighted cost function determines the multi-objective weighted cost function that includes only the control quantity as follows:

[0038] in, To control the quantity, [ ; For a weighted quadratic matrix, by , and Dynamic adjustment; It is a linear term vector, including state bias and cost gradient information; The optimal control sequence is calculated by applying the OSQP solver to the multi-objective weighted cost function that includes only the control variables. for:

[0039] According to the aforementioned multi-objective dynamic weighted path planning method based on confidence feedback, the method further includes performing periodic control processing on the autonomous vehicle based on the optimal control sequence at future time moments, and also includes: The first control input is obtained from the optimal control sequence, and the first control input is executed at each time step using a rolling optimization control strategy:

[0040] in, Indicates the current moment of autonomous vehicles The actual execution control quantity; The optimal control decision at the current moment is determined through model predictive control; After executing the first control variable, the vehicle state of the autonomous vehicle is updated, and the result is obtained. :

[0041] Based on the updated vehicle status, the overall confidence level of the scenario is recalculated as follows: and the weight parameters of the comprehensive cost function Then, a new optimal control sequence is calculated and the process continues to the next time step; The above single-cycle control and multi-objective path collaborative planning are executed repeatedly.

[0042] The beneficial effects of this invention are as follows: By establishing a confidence feedback mechanism, an information loop is achieved between perception and path planning, enabling the planning strategy to dynamically adjust based on the reliability of the perceived data. When the perception confidence is low, the system automatically increases the safety weight and adopts conservative path planning; when the confidence is high, the system correspondingly increases the efficiency and comfort weights, thereby achieving an adaptive dynamic balance of the planning objective under different environments, significantly improving the safety and robustness of path planning. Furthermore, a confidence-driven multi-objective dynamic weight update model and a path optimization solution mechanism based on model predictive control (MPC) are constructed. This allows for real-time calculation of the optimal control sequence while satisfying vehicle dynamics constraints, achieving a closed-loop mapping from perception reliability to path decision strategy. Through these technical means, this invention can achieve highly safe, highly reliable, and highly adaptive path planning in complex and ever-changing traffic environments, providing strong support for the stable operation and intelligent decision-making of autonomous driving systems. Attached Figure Description

[0043] Figure 1 This is a schematic diagram of the multi-objective dynamic weighted path planning process based on confidence feedback according to an embodiment of the present invention.

[0044] Figure 2 This is a flowchart of another multi-objective path collaborative planning method based on confidence feedback according to an embodiment of the present invention.

[0045] Figure 3 This is a schematic diagram of a multi-objective dynamic weighted path planning device based on confidence feedback according to an embodiment of the present invention. Detailed Implementation

[0046] The embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings. Throughout the description, the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions. In the following description, suffixes such as "module," "part," or "unit" used to denote elements are used only for the purpose of illustrative purposes and have no specific meaning in themselves. Therefore, "module," "part," or "unit" can be used interchangeably. Terms such as "first," "second," etc., are used only to distinguish technical features and should not be construed as indicating or implying relative importance, or implicitly indicating the number of indicated technical features, or implicitly indicating the sequential relationship of the indicated technical features. In the following description, the consecutive reference numerals for method steps are for ease of review and understanding. Adjusting the implementation order of steps, in conjunction with the overall technical solution of the present invention and the logical relationship between the various steps, will not affect the technical effect achieved by the technical solution of the present invention. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0047] refer to Figure 1 , Figure 1 This is a schematic diagram of a multi-objective dynamic weighted path planning process based on confidence feedback, which includes, but is not limited to, steps S100~S900: S100: Acquire environmental perception data of autonomous vehicles, determine the state data of multiple detection targets based on the environmental perception data, and fuse the state data with timestamps to obtain an environmental information set of multiple detection targets.

[0048] In some embodiments, by acquiring environmental perception data from multiple sensors installed on an autonomous vehicle, and fusing the data with timestamps, an environmental information set for multiple detected targets is obtained:

[0049]

[0050] in, For the set of environmental objectives, The total number of environmental targets. This serves as the serial number identifier for the target being detected. Indicates time; To detect the target's state data, including velocity vectors and position vector in global coordinate system , For transpose, The covariance matrix for detecting the position of a target is used to represent the uncertainty of geometric measurements. This represents the target detection confidence level determined by the perception algorithm.

[0051] In some embodiments, the environmental information set of multiple detected targets represents the environmental information set of the current frame.

[0052] It should be noted that the autonomous vehicle is the self-driving vehicle, that is, the vehicle that needs to perform path planning in this embodiment of the invention, while the detection target is other vehicles that are driving in the same scene as the autonomous vehicle.

[0053] In some embodiments, multi-source sensor excitation includes photoradar, camera, millimeter-wave radar, IMU, etc., installed in the autonomous vehicle.

[0054] This invention also includes preprocessing and anomaly correction of the environmental information sets of multiple detection targets. This includes anomaly detection and filtering of the environmental information sets of multiple detection targets, and removing distorted or missing frame information. If an abnormal measurement point is detected, neighborhood interpolation or Kalman filtering is used to reconstruct the missing state of the abnormal measurement point, and finally a smooth and stable target state estimation result is obtained. .

[0055] S200 calculates the overall confidence level of each detected target in the environmental information set.

[0056] The probability confidence level of the perception algorithm in this embodiment of the invention With geometric measurement variance The physical uncertainties represented come from different sources, so they need to be integrated to obtain a comprehensive index that better reflects the credibility of the target.

[0057] In some embodiments, the steps for calculating the target comprehensive confidence level are as follows: For a single target The overall confidence level of the target is determined by introducing a weighted fusion model of detection probability and geometric variance. for:

[0058] in, The trace operation represents the matrix operation, used to characterize the overall measurement variance; This is a weighting coefficient used to represent the relative importance of target detection confidence and geometric stability; is the attenuation coefficient, used to map the geometric variance to the confidence space.

[0059] It should be noted that when the variance of geometric measurements is large, Rapidly decrease, making A decrease indicates a reduction in the reliability of the detected target's information; conversely, when the geometric error is small and the detection probability is high, A value close to 1 indicates that the identification result of the detected target is reliable.

[0060] To suppress instantaneous fluctuations, this invention employs a time-series smoothing strategy to exponentially weight the target confidence level, obtaining the time-series smoothing update. Target overall confidence level for:

[0061] in, This is a smoothing coefficient; a larger value indicates a more stable confidence update.

[0062] And, it also includes the time. Target overall confidence level Constraints are applied using a single-step maximum descent limit; the time after constraint... Target overall confidence level for:

[0063] in, This represents the maximum confidence decrease in a single step, and .

[0064] S300 calculates the relative state of each detected target to the autonomous vehicle based on the current state of the autonomous vehicle, and determines the set of key objects based on the relative state using a target importance scoring function.

[0065] It is understood that the embodiments of the present invention provide a stable confidence level for each target. For path planning, a scenario-level single scalar is needed to drive weight adaptation, so key objects need to be selected to aggregate into scenario confidence.

[0066] In some embodiments, the present invention calculates the importance of each detected target and performs screening using a target importance scoring function. Specifically, the target importance scoring function is calculated as follows:

[0067] in, The scoring result of the target importance scoring function; Location of autonomous vehicles With the detection target location European distance, For the speed of autonomous vehicles, For detecting the speed of the target; position of the autonomous vehicle. The speed of autonomous vehicles The heading angle of the autonomous vehicle constitutes the current state of the autonomous vehicle; As distance difference weight, As the speed difference weight, For the heading angle difference weight, To prevent small constants from being divided by zero, To detect the difference in heading angle between the target and the autonomous vehicle; Then, based on the scoring results of the target importance scoring function... Select Each detection target yields a set of key objects. This key object collection Used for subsequent calculation of scene-level confidence.

[0068] S400 determines the overall scene confidence level based on the set of key objects. The overall scene confidence level is used to characterize the autonomous vehicle's global trust in environmental perception data.

[0069] The technical solution of this invention is based on a set of key objects. Calculate the average confidence level of the scenario and minimum confidence :

[0070] Based on the average confidence level of the scenario and minimum confidence Calculate the overall confidence level of the scenario for:

[0071] in, These are the weighting coefficients.

[0072] The technical solution of this invention achieves the following effect: when the detection of a certain key target is unreliable ( Low), This will be automatically reduced, thus increasing the safety weight in subsequent planning (e.g., when a pedestrian crossing is obstructed). A rapid decline, thus pulling down Conversely, when key objects are generally trustworthy, promote.

[0073] In another implementation scenario, embodiments of the present invention further include handling autonomous vehicles under special operating conditions through boundary conditions and anomaly handling mechanisms, including: To ensure stable output even under extreme and degrading conditions Therefore, boundary constraints need to be set in extreme scenarios: When there is no effective target, that is At that time, set ,in This is the default confidence level, used to maintain normal driving in empty scenarios; When the covariance matrix When singular or non-positive time, perform correction:

[0074] It is an identity matrix.

[0075] When the sensor health status detection flag When (indicating perceptual degradation), a forced confidence downgrade is performed:

[0076] This is the upper limit threshold for health degradation, ensuring that the autonomous driving system automatically switches to conservative planning upon detecting degradation.

[0077] It should be noted that the confidence level Used to reflect the reliability of environmental information. When When the level is low, the safety weight should be increased to ensure driving safety; when When the weight is high, the safety weight can be reduced and the efficiency and comfort weights can be increased, thereby improving the overall system performance.

[0078] S500 determines the weight parameters of the comprehensive cost function based on the overall confidence level of the scenario. The comprehensive cost function is used to characterize the optimal trajectory under the constraints of safety cost, efficiency cost, and comfort cost.

[0079] In some embodiments, the comprehensive cost function is:

[0080] in, This is a safety cost term, used to characterize the risk of collision between the vehicle and an obstacle. For security weight parameters; This is an efficiency cost term, used to characterize the path length, time, or energy consumption of a vehicle's journey. For efficiency weighting parameters; This is a comfort cost term, used to characterize the smoothness of vehicle acceleration and steering changes. For comfort weighting parameters; , and All are greater than 0. ,and , and The dynamic weighting parameter means that during subsequent path planning, the dynamic weighting parameter changes to achieve the effect of real-time response to changes in confidence and dynamic control.

[0081] A monotonic mapping model using the Sigmoid function is used to map the overall scene confidence to... Get the time Security weights for:

[0082] in, The steepness coefficient of the function determines the sensitivity of the overall confidence level of the scene to the weight adjustment. This is the balance threshold; Based on security weight Determine the time Efficiency weight With comfort weight for:

[0083] in, This is the ratio coefficient between efficiency and comfort.

[0084] S600 determines the safety cost term, efficiency cost term, and comfort cost term of the comprehensive cost function based on the current state of the autonomous vehicle using a discrete bicycle model and corresponding constraint conditions.

[0085] The specific solutions of this invention are as follows: The safety cost term, efficiency cost term, and comfort cost term of the comprehensive cost function are determined separately, including determining the safety cost term based on the obstacle potential field model and time-to-collision (TTC). for:

[0086] In other words, safety cost is used to characterize the spatial safety distance and collision risk between a vehicle and an obstacle.

[0087] in, For autonomous vehicles and the first Each detection target at time The Euclidean distance; This is the safe distance threshold; To prevent division by zero of small constants; This is the trade-off factor between the safety distance and the TTC term; The collision time prediction between the vehicle and the target is expressed as:

[0088] When the distance is too close or the velocity direction tends towards collision. The value increases significantly, thereby guiding the planning process to actively avoid obstacles or slow down.

[0089] Determine the efficiency cost item based on the overall performance of path length, time, and energy consumption. for:

[0090] in, For autonomous vehicles at all times Location; It is longitudinal acceleration; For a long walk; , and These are the path length weight, time weight, and energy consumption weight, respectively. The comfort penalty is determined based on the vehicle's rapid acceleration, sharp turns, or frequent changes in handling. for:

[0091] in, It is longitudinal acceleration; For path curvature; , and These are the acceleration weight, jerk weight, and curvature weight, respectively. The curvature is approximately calculated as follows:

[0092] in, For the front wheel steering angle, This refers to the vehicle's wheelbase.

[0093] S700 determines the minimum multi-objective weighted cost function based on the current state of the autonomous vehicle and the comprehensive cost function, and uses the OSQP solver to determine the optimal control sequence of the autonomous vehicle at future time steps based on the minimum multi-objective weighted cost function.

[0094] It should be noted that, after completing the confidence-driven multi-objective dynamic weight modeling in this embodiment of the invention, the control system of the autonomous vehicle is already able to determine the scene confidence level. The weights of safety, efficiency, and comfort objectives are adaptively adjusted. To transform this multi-objective optimization model into executable trajectories and control commands, this invention introduces a Model Predictive Control (MPC) mechanism in this module to achieve dynamic optimal path planning for autonomous vehicles under constraints.

[0095] The embodiments of the present invention comprehensively consider vehicle dynamics constraints, state feasibility, and real-time requirements, and jointly optimize the control quantities at multiple future moments within a rolling prediction window to ensure that the generated path not only meets physical feasibility but also responds in real time to changes in perception confidence, thus achieving a closed loop from planning model to execution control.

[0096] In some embodiments, autonomous vehicles need to consider longitudinal velocity, lateral position, and attitude changes simultaneously during path planning. Therefore, a simplified bicycle model is used to describe its kinematic constraints, and the kinematic constraints are obtained by acquiring the autonomous vehicle's kinematic parameters at time points. state vector and control input vector for:

[0097]

[0098] in, These are the position coordinates of the autonomous vehicle in the global coordinate system. The heading angle of the autonomous vehicle; The longitudinal speed of the autonomous vehicle; For autonomous vehicles, the longitudinal acceleration represents the speed change control quantity; The front wheel steering angle represents the steering control amount. Based on the state vector and control input vector The state transition relationship of an autonomous vehicle between adjacent time points is determined as follows:

[0099] in, The discretized nonlinear dynamic equations for the state of an autonomous vehicle changing over time are expressed as:

[0100] The control and state constraints are as follows:

[0101] in, To minimize the speed limit, Maximum speed limit; To limit the acceleration to the maximum; For the maximum turning angle, This represents the upper limit of the rate of change of the rotation angle.

[0102] These constraints ensure that autonomous vehicles do not encounter physically unreachable or dynamically unstable states on the optimized trajectory, thereby improving vehicle driving safety.

[0103] In some embodiments, after completing confidence-driven multi-objective modeling, the path planning problem is essentially transformed into a dynamic optimal control problem with nonlinear constraints. To obtain an executable approximate optimal solution within a finite computation time, this embodiment of the invention employs the MPC rolling optimization method for optimization, including: With time state vector As an initial condition, determine the future. Step state sequence and control input sequence for:

[0104]

[0105] Determining the future based on the comprehensive cost function The minimization of the multi-objective weighted cost function for the step is:

[0106] In some embodiments, since vehicle dynamics is a nonlinear function Directly solving nonlinear programming (NLP) problems involves high computational complexity. Therefore, a linearization approximation and quadratic cost expansion are used to transform the problem into a quadratic programming (QP) form, enabling a faster solution.

[0107] At the current state point Perform a first-order Taylor expansion nearby:

[0108] in, , They are respectively At the current state point right and The first-order partial derivative, This is the linearization constant term.

[0109] use For the future Expanding the time-series recursion, we get:

[0110] All future states can be written as linear combinations of control inputs, such that all future states... All become control quantities Linear functions:

[0111] in, ; , , It is by , , The derived constant matrix.

[0112] This embodiment of the invention further obtains the state point of the current state of the autonomous vehicle. Based on the state points of autonomous vehicles Certainty and the Future The minimization of the multi-objective weighted cost function determines the multi-objective weighted cost function that includes only the control quantity as follows:

[0113] in, To control the quantity, [ ; For a weighted quadratic matrix, by , and Dynamic adjustment; It is a linear term vector, including state bias and cost gradient information; The aforementioned multi-objective weighted cost function, which only includes the control variables, is a standard QP problem. Therefore, the OSQP solver is used to calculate the optimal control sequence for the multi-objective weighted cost function that only includes the control variables. for:

[0114] The S800 performs periodic control processing on autonomous vehicles based on the optimal control sequence for future moments.

[0115] In some embodiments, the first control quantity is obtained from the optimal control sequence, and a rolling optimization control strategy is used to execute the first control quantity at each time step:

[0116] in, Indicates the current moment of autonomous vehicles The actual execution control quantity; The optimal control decision at the current moment is determined through model predictive control; After executing the first control variable, the vehicle state of the autonomous vehicle is updated, and the result is obtained. :

[0117] Based on the updated vehicle status, the overall confidence level of the scenario is recalculated as follows: and the weight parameters of the comprehensive cost function Then calculate the new optimal control sequence. And proceed to the next moment; cyclically execute the above single control and multi-objective path collaborative planning.

[0118] refer to Figure 2 Another multi-objective path collaborative planning method based on confidence feedback is shown below: (1) Environmental perception data acquisition: The system collects environmental information in real time through multi-source sensors (LiDAR, camera, millimeter-wave radar, IMU, etc.) and extracts target objects. Location, speed, category and detection confidence The perception module synchronously performs timestamp alignment and multi-sensor fusion to form the environmental information set for the current frame. ; (2) Sensing data preprocessing and anomaly correction: anomaly detection and filtering are performed on the original sensing data to remove distorted or missing frame information; when an abnormal measurement point is detected, the missing state is reconstructed through neighborhood interpolation or Kalman filtering to obtain a smooth and stable target state estimation result. ; (3) Based on the fusion model, use the detection confidence level Covariance of Geometric Measurement Calculate the overall confidence level at the target level Exponential smoothing is performed through a time-sliding window; (4) Key target screening and importance assessment: Based on the vehicle's current speed and heading, calculate the relative distance, relative speed, and heading difference for each target, and use the importance function. Screening out the most potentially risky front A collection of key objects This provides input for scene confidence calculation; (5) Scene-level confidence aggregation, in the key object set Inside, according to The overall scene confidence score is calculated to reflect the system's global confidence in the current perception results. (6) with As input, the Sigmoid mapping function is called to calculate the weights of three objectives—safety, efficiency, and comfort—in real time. This allows the planning weights to change dynamically with perceived credibility.

[0119] (7) Read the current status of the vehicle This includes position, heading angle, and velocity information, providing initial conditions for subsequent MPC modeling.

[0120] (8) Based on the discrete bicycle model (i.e., the nonlinear dynamic equation) and the control and state constraints, the optimization objective is to minimize the multi-objective weighted cost function.

[0121] (9) Optimize the linearization of the model by linearizing the nonlinear dynamic equations at the current state point and discretizing the optimization objective into a standard quadratic programming (QP) form; (10) Solve for the optimal control sequence using the OSQP solver to calculate the future in the prediction time domain. The optimal control sequence for each step; (11) Execute the first control variable and state update, and extract the first control variable from the optimal sequence. This information is used as the vehicle's current command. The vehicle control system then uses this information to perform throttle, braking, and steering operations, and updates its status accordingly. Complete one path advancement; (12) Rolling time-domain update and closed-loop execution: After the vehicle executes, the perception module re-collects environmental data and calculates the new confidence level. The planning module updates the weights synchronously. The MPC optimization model is then rebuilt and the process moves to the next time step.

[0122] Figure 3 This is a schematic diagram of a multi-objective dynamic weighted path planning device based on confidence feedback according to an embodiment of the present invention. The device includes a first module 310, a second module 320, a third module 330, a fourth module 340, a fifth module 350, a sixth module 360, a seventh module 370, and an eighth module 380.

[0123] The system comprises five modules: a first module for acquiring environmental perception data from the autonomous vehicle, determining the state data of multiple detection targets based on the environmental perception data, and fusing the state data with timestamps to obtain an environmental information set for multiple detection targets; a second module for calculating the comprehensive confidence score of each detection target in the environmental information set; a third module for calculating the relative state of each detection target to the autonomous vehicle based on the current state of the autonomous vehicle, and determining a set of key objects based on the relative state using a target importance scoring function; a fourth module for determining the scene comprehensive confidence score based on the set of key objects, whereby the scene comprehensive confidence score characterizes the autonomous vehicle's global trust in the environmental perception data; and a fifth module for determining the scene comprehensive confidence score based on the scene comprehensive confidence score. The system comprises three modules: a reliability module, a sixth module, and a seventh module. The first module determines the weight parameters of the comprehensive cost function, which characterizes the optimal trajectory under constraints of safety, efficiency, and comfort costs. The sixth module determines the safety, efficiency, and comfort costs of the comprehensive cost function based on the current state of the autonomous vehicle, using a discrete bicycle model and corresponding constraints. The seventh module determines the minimum multi-objective weighted cost function based on the current state of the autonomous vehicle and the comprehensive cost function. The second module uses an OSQP solver to determine the optimal control sequence for the autonomous vehicle at future time steps based on the minimum multi-objective weighted cost function. Finally, the third module performs periodic control processing on the autonomous vehicle based on the optimal control sequence at future time steps.

[0124] For example, with the cooperation of the first to eighth modules in the device, the embodiment device can implement any of the aforementioned multi-objective dynamic weighted path planning methods based on confidence feedback, namely, acquiring environmental perception data of the autonomous vehicle, determining the state data of multiple detection targets based on the environmental perception data, fusing the state data with timestamps to obtain an environmental information set of multiple detection targets; calculating the comprehensive confidence of each detection target in the environmental information set; calculating the relative state of each detection target with the autonomous vehicle based on the current state of the autonomous vehicle, determining the key object set using a target importance scoring function based on the relative state; and determining the scene comprehensive confidence based on the key object set, wherein the scene comprehensive confidence is used to characterize the autonomous vehicle's perception of the environment. The invention establishes a global confidence level for scene perception data. Based on the overall scene confidence level, it determines the weight parameters of the comprehensive cost function, which characterizes the optimal trajectory under constraints of safety, efficiency, and comfort costs. According to the current state of the autonomous vehicle, it uses a discrete bicycle model and corresponding constraints to determine the safety, efficiency, and comfort costs of the comprehensive cost function. Based on the current state of the autonomous vehicle and the comprehensive cost function, it determines a minimized multi-objective weighted cost function. Using the OSQP solver, it determines the optimal control sequence for the autonomous vehicle at future time steps. Based on the optimal control sequence at future time steps, it performs periodic control processing on the autonomous vehicle. The beneficial effects of this invention are: through a confidence feedback mechanism, it achieves an information closed loop between perception and path planning, enabling the planning strategy to dynamically adjust based on the reliability of the perception data. When the perceived confidence level is low, the system automatically increases the safety weight and adopts conservative path planning; when the confidence level is high, the system correspondingly increases the efficiency and comfort weights, thereby achieving an adaptive dynamic balance of the planning objective under different environments, significantly improving the safety and robustness of path planning. Furthermore, a confidence-driven multi-objective dynamic weight update model and a path optimization solution mechanism based on model predictive control (MPC) are constructed, which can calculate the optimal control sequence in real time while satisfying vehicle dynamics constraints, realizing a closed-loop mapping from perceived reliability to path decision strategy. Through the above technical means, this invention can achieve high safety, high reliability and high adaptability path planning in complex and ever-changing traffic environments, providing strong support for the stable operation and intelligent decision-making of autonomous driving systems.

[0125] This invention also provides an electronic device, which includes a processor and a memory; The memory stores the program; The processor executes the program to perform the aforementioned multi-objective dynamic weighted path planning method based on confidence feedback; the electronic device has the function of carrying and running the software system of multi-objective dynamic weighted path planning based on confidence feedback provided in the embodiments of the present invention, such as a personal computer, minicomputer, mainframe, workstation, network or distributed computing environment, standalone or integrated computer platform, or communicating with charged particle tools or other imaging devices, etc.

[0126] This invention also provides a computer-readable storage medium storing a program that is executed by a processor to implement the multi-objective dynamic weighted path planning method based on confidence feedback as described above.

[0127] In some alternative embodiments, the functions / operations mentioned in the block diagrams may not occur in the order shown in the operation diagrams. For example, depending on the functions / operations involved, two consecutively shown blocks may actually be executed substantially simultaneously, or the blocks may sometimes be executed in reverse order. Furthermore, the embodiments presented and described in the flowcharts of this invention are provided by way of example to provide a more comprehensive understanding of the technology. The disclosed methods are not limited to the operations and logic flows presented in the embodiments of this invention. Alternative embodiments are contemplated, in which the order of various operations is changed and sub-operations described as part of a larger operation are executed independently.

[0128] This invention also discloses a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device can read the computer instructions from the computer-readable storage medium and execute the computer instructions, causing the computer device to perform the aforementioned multi-objective dynamic weighted path planning method based on confidence feedback.

[0129] Furthermore, although the invention has been described in the context of functional modules, it should be understood that, unless otherwise stated, one or more of the described functions and / or features may be integrated into a single physical device and / or software module, or one or more functions and / or features may be implemented in a separate physical device or software module. It is also understood that a detailed discussion of the actual implementation of each module is unnecessary for understanding the invention. Rather, considering the properties, functions, and internal relationships of the various functional modules in the apparatus disclosed in the embodiments of the invention, the actual implementation of the module will be understood within the scope of conventional skill of an engineer. Therefore, those skilled in the art can implement the invention as set forth in the claims using ordinary techniques without excessive experimentation. It is also understood that the specific concepts disclosed are merely illustrative and are not intended to limit the scope of the invention, which is determined by the full scope of the appended claims and their equivalents.

[0130] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, essentially, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes 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.

[0131] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can include, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0132] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0133] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0134] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0135] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

[0136] The above is a detailed description of the preferred embodiments of the present invention, but the present invention is not limited to the embodiments described. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of the present invention, and these equivalent modifications or substitutions are all included within the scope defined by the claims of this application.

Claims

1. A multi-objective dynamic weight path planning method based on confidence feedback, characterized in that, include: The system acquires environmental perception data from autonomous vehicles, determines the state data of multiple detection targets based on the environmental perception data, and fuses the state data with timestamps to obtain an environmental information set of multiple detection targets. Calculate the overall confidence level of each detected target in the environmental information set; Based on the current state of the autonomous vehicle, the relative state of each detected target with respect to the autonomous vehicle is calculated. Based on the relative state, the set of key objects is determined using a target importance scoring function. The overall confidence level of a scenario is determined based on a set of key objects. This overall confidence level is used to characterize the global trust that autonomous vehicles have in environmental perception data. Based on the overall confidence level of the scenario, the weight parameters of the overall cost function are determined, whereby the overall cost function is used to characterize the optimal trajectory under the constraints of safety cost, efficiency cost, and comfort cost. Based on the current state of the autonomous vehicle, the safety cost term, efficiency cost term, and comfort cost term of the comprehensive cost function are determined using a discrete bicycle model and corresponding constraint conditions. Based on the current state and comprehensive cost function of the autonomous vehicle, the minimum multi-objective weighted cost function is determined, and the OSQP solver is used to determine the optimal control sequence of the autonomous vehicle at future time steps based on the minimum multi-objective weighted cost function. Based on the optimal control sequence for future moments, periodic control processing is performed on the autonomous vehicle; The process involves acquiring environmental perception data from the autonomous vehicle, determining the state data of multiple detection targets based on the environmental perception data, and fusing the state data with timestamps to obtain an environmental information set for multiple detection targets, including: Environmental perception data from multiple sensors installed on autonomous vehicles is acquired, and the data is fused with timestamps to obtain an environmental information set for multiple detected targets: wherein, is a set of environmental targets, is a total number of sets of environmental targets, is a serial number identification of the detection target, denotes a time instant; is state data of the detection target, including a velocity vector and a position vector in a global coordinate system , is a transpose, is a position measurement covariance matrix of the detection target, used to represent the uncertainty of geometric measurement, is a target detection confidence determined by a perception algorithm; The calculation of the overall confidence score of each detected target in the environmental information set includes: According to the environmental information set of multiple detection targets, a weighted fusion model of detection probability and geometric variance is used to determine a target comprehensive confidence To: wherein, denotes the trace operation of a matrix, used to characterize the overall measurement variance; is a weighting coefficient, used to represent the relative importance of target detection confidence and geometric stability; is a decay coefficient, used to map the geometric variance to the confidence space; The target confidence is updated by using a time sequence smoothing strategy and an exponential weighting, and a target comprehensive confidence at a time point t is obtained as follows: ​​ wherein is a smoothing coefficient; And, also includes the target comprehensive confidence of time By single step maximum descent limiting, the target comprehensive confidence of time is:​​ wherein is the single step maximum confidence decrease, and .

2. The method of claim 1, wherein, The method further includes: The environmental information set of multiple detection targets is preprocessed and anomaly corrected, including anomaly detection and filtering of the environmental information set of multiple detection targets, and removal of distorted or missing frame information; if an abnormal measurement point is detected, the missing state of the abnormal measurement point is reconstructed by neighborhood interpolation or Kalman filtering.

3. The method of claim 1, wherein, The process involves calculating the relative state of each detected target to the autonomous vehicle based on its current state, and determining the set of key objects using an importance function based on the relative state. This includes: The target importance scoring function is: wherein, is a score result of the target importance scoring function; is an autonomous vehicle position is an Euclidean distance between the detected target position and the autonomous vehicle position is a speed of the autonomous vehicle, is a speed of the detected target; the autonomous vehicle position , the speed of the autonomous vehicle and a heading angle of the autonomous vehicle constitute a current state of the autonomous vehicle; is a distance difference weight, is a speed difference weight, is a heading angle difference weight, is a small constant to prevent division by zero, is a heading angle difference between the detected target and the autonomous vehicle. Based on the scoring results of the target importance scoring function Select Each detection target yields a set of key objects. .

4. The multi-objective dynamic weighted path planning method based on confidence feedback according to claim 3, characterized in that, The determination of the overall confidence level of the scenario based on the set of key objects includes: Based on the key object set Calculate the average confidence level of the scenario and minimum confidence : Based on the average confidence level of the scenario and minimum confidence Calculate the overall confidence level of the scenario for: in, These are the weighting coefficients.

5. The multi-objective dynamic weighted path planning method based on confidence feedback according to claim 4, characterized in that, The weight parameters of the comprehensive cost function are determined based on the overall confidence level of the scenario. The comprehensive cost function characterizes the optimal trajectory under constraints of safety, efficiency, and comfort costs, and includes: The comprehensive cost function is: in, This is a safety cost term, used to characterize the risk of collision between the vehicle and an obstacle. For security weight parameters; This is an efficiency cost term, used to characterize the path length, time, or energy consumption of a vehicle's journey. For efficiency weighting parameters; This is a comfort cost term, used to characterize the smoothness of vehicle acceleration and steering changes. For comfort weighting parameters; , and All are greater than 0. ,and , and For dynamic weight parameters; A monotonic mapping model using the Sigmoid function is used to map the overall scene confidence to... Get the time Security weights for: in, The steepness coefficient of the function determines the sensitivity of the overall confidence level of the scene to weight adjustments; This is the balance threshold; Based on security weight Determine the time Efficiency weight With comfort weight for: in, This is the ratio coefficient between efficiency and comfort.

6. The multi-objective dynamic weighted path planning method based on confidence feedback according to claim 5, characterized in that, The process of determining the safety cost term, efficiency cost term, and comfort cost term of the comprehensive cost function based on the current state of the autonomous vehicle, using a discrete bicycle model and corresponding constraint conditions, includes: The safety cost term, efficiency cost term, and comfort cost term of the comprehensive cost function are determined separately, including determining the safety cost term based on the obstacle potential field model and collision time. for: in, For autonomous vehicles and the first Each detection target at time The Euclidean distance; This is the safe distance threshold; To prevent division by zero of small constants; This is the trade-off factor between the safety distance and the TTC term; The collision time prediction between the vehicle and the target is expressed as: Determine the efficiency cost item based on the overall performance of path length, time, and energy consumption. for: in, For autonomous vehicles at all times Location; It is longitudinal acceleration; For a long walk; , and These are the path length weight, time weight, and energy consumption weight, respectively. The comfort penalty is determined based on the vehicle's rapid acceleration, sharp turns, or frequent changes in handling. for: in, It is longitudinal acceleration; For path curvature; , and These are acceleration weight, jerk weight, and curvature weight, respectively.

7. The multi-objective dynamic weighted path planning method based on confidence feedback according to claim 6, characterized in that, The process of determining a minimum multi-objective weighted cost function based on the current state and comprehensive cost function of the autonomous vehicle, and then using an OSQP solver to determine the optimal control sequence for the autonomous vehicle at future time steps based on the minimum multi-objective weighted cost function, includes: Obtain the real-time information of autonomous vehicles state vector and control input vector for: in, These are the position coordinates of the autonomous vehicle in the global coordinate system. The heading angle of the autonomous vehicle; The longitudinal speed of the autonomous vehicle; For autonomous vehicles, the longitudinal acceleration represents the speed change control quantity; The front wheel steering angle represents the steering control amount. Based on the state vector and control input vector The state transition relationship of an autonomous vehicle between adjacent time points is determined as follows: in, The discretized nonlinear dynamic equations for the state of an autonomous vehicle changing over time are expressed as: The control and state constraints are as follows: in, To minimize the speed limit, Maximum speed limit; To limit the acceleration to the maximum; For the maximum turning angle, This is the upper limit of the rate of change of the rotation angle; With time state vector As initial conditions, determine the future. Step state sequence and control input sequence for: Determining the future based on the comprehensive cost function The minimization of the multi-objective weighted cost function for the step is: Obtain the state points of the current state of the autonomous vehicle. Based on the state points of autonomous vehicles Certainty and the Future The minimization of the multi-objective weighted cost function determines the multi-objective weighted cost function that includes only the control quantity as follows: in, To control the quantity, [ ; For a weighted quadratic matrix, by , and Dynamic adjustment; It is a linear term vector, including state bias and cost gradient information; The OSQP solver is used to calculate the optimal control sequence for a multi-objective weighted cost function that includes only control variables. for: 。 8. The multi-objective dynamic weighted path planning method based on confidence feedback according to claim 7, characterized in that, The method of performing periodic control processing on the autonomous vehicle based on the optimal control sequence for future moments also includes: The first control input is obtained from the optimal control sequence, and the first control input is executed at each time step using a rolling optimization control strategy: in, Indicates the current moment of autonomous vehicles The actual execution control quantity; The optimal control decision at the current moment is determined through model predictive control; After executing the first control variable, the vehicle state of the autonomous vehicle is updated, and the result is obtained. : Based on the updated vehicle status, the overall confidence level of the scenario is recalculated as follows: and the weight parameters of the comprehensive cost function Then, a new optimal control sequence is calculated and the process continues to the next time step; The above control is executed repeatedly.

Images