An automatic driving space-time trajectory repairing method and system based on perception confidence

Abstract

The application provides a kind of automatic driving space-time trajectory repair method and system based on perception confidence, method includes detecting space-time conflict point in original planning trajectory and locking failure point;Based on vehicle-mounted multi-sensor data fusion calculation includes environment robustness and space-time consistency comprehensive perception confidence;Strong coupling time dimension constructs four-dimensional space-time state diagram;According to the comprehensive perception confidence dynamic adjustment safety boundary, and to the high-risk candidate connection edge that crosses safety boundary is topologically pruned to greatly reduce search space;Based on the four-dimensional state diagram after pruning, efficient full space-time parallel trajectory search is executed;Finally, the discrete node sequence is dynamically smoothed and physically constrained, and the functional safety degradation control strategy is output to the safe and reliable chassis execution trajectory.The application breaks the open loop barrier of perception and planning, significantly improves the safety, real-time performance and system-level functional safety level of trajectory repair in complex dynamic and severe scenarios.

Description

Technical Field

[0001] This invention belongs to the technical field of motion planning for intelligent connected vehicles, and specifically relates to a method and system for spatiotemporal trajectory repair of autonomous driving based on perception confidence. Background Technology

[0002] With the development of autonomous driving technology, vehicles need to continuously output motion trajectories that meet safety, comfort, and real-time requirements in complex and dynamic traffic environments. However, in complex scenarios such as high-speed cut-ins, sudden pedestrian appearances, abnormal obstacle movements, and severe weather, the original planned trajectory is prone to spatiotemporal conflicts, leading to trajectory failure.

[0003] Existing trajectory repair solutions typically employ two-dimensional local replanning methods, such as traditional SL / ST decoupled planning. Due to the lack of tightly coupled time dimension constraints, it is difficult to effectively avoid spatiotemporal collisions in scenarios where dynamic obstacles suddenly intersect.

[0004] Meanwhile, existing solutions generally rely on the single target classification confidence output by the perception algorithm, lacking a closed-loop constraint mechanism for perception uncertainties, such as interference from multiple sensors in harsh environments and inconsistent spatiotemporal observations, and cannot dynamically adjust the safety boundary based on the overall perception reliability.

[0005] Furthermore, traditional trajectory repair algorithms are prone to getting stuck in local optima or timeouts in complex, high-dimensional search spaces, making it difficult to simultaneously ensure both safety and real-time performance.

[0006] Therefore, there is an urgent need for an autonomous driving trajectory repair method and system that can integrate multi-sensory confidence assessment, four-dimensional spatiotemporal topological constraints, functional safety closed-loop degradation, and efficient dynamic trajectory search. Summary of the Invention

[0007] To address the aforementioned technical issues, this invention proposes an autonomous driving spatiotemporal trajectory repair method and system based on perception confidence. By constructing a hierarchical system architecture, a comprehensive perception confidence, including environmental robustness and spatiotemporal consistency, is introduced into the topology construction and trimming process of the four-dimensional spatiotemporal state diagram, thereby achieving safe closed-loop trajectory repair in complex dynamic traffic environments.

[0008] In a first aspect, the present invention provides the following technical solution: a method for spatiotemporal trajectory repair of autonomous driving based on perception confidence, comprising: Step S1: Real-time detection of the spatiotemporal relationship between the vehicle's original planned trajectory and the future multi-probability predicted trajectory of the dynamic obstacle. When a conflict is detected between the original planned trajectory and the future movement trajectory of the dynamic obstacle, the trajectory failure point is locked and the trajectory repair program is activated. Step S2: Based on the perception data collected by the vehicle's multi-sensor system, the evaluation module calculates the comprehensive perception confidence level in real time under the current environment. ; Step S3: Starting from the current actual state node of the vehicle, construct a four-dimensional spatiotemporal state graph that includes longitudinal position, lateral position, heading angle and time dimension, and perform high-dimensional non-uniform spatiotemporal sampling in the look-ahead time to generate candidate connection edges. Step S4: Based on the comprehensive perceived confidence level The safety boundary is dynamically adjusted, and candidate connection edges that cross the safety boundary are safely pruned and eliminated to obtain the pruned four-dimensional spatiotemporal state diagram. Step S5: Based on the clipped four-dimensional spatiotemporal state diagram, perform a full spatiotemporal parallel trajectory search with the goal of minimizing the total trajectory cost function to find the optimal spatiotemporal discrete node sequence; Step S6: Use a high-order curve to perform full-time-space joint smoothing fitting on the searched optimal spatiotemporal discrete node sequence, and output the vehicle control trajectory after passing the vehicle physical constraint verification.

[0009] Compared to existing technologies, the beneficial effects of this invention are as follows: This invention simultaneously establishes a closed-loop linkage mechanism between trajectory repair and functional safety degradation control. When the overall perception confidence decreases due to environmental interference or poor observation consistency of multiple sensors, the system not only dynamically expands the topological boundary to increase safety redundancy, but also automatically triggers multi-level functional safety control logic to execute active speed limiting, increasing following distance, or minimum risk maneuver (MRM), thereby improving the functional safety level of autonomous driving and the real-time performance of trajectory repair.

[0010] Preferably, the comprehensive perceived confidence level in step S2 Calculate using the following formula: ; In the formula, The target classification confidence score output by multi-sensor fusion; For multi-sensor environment robustness factor; It is a spatiotemporal consistency factor; , , These are the weight coefficients corresponding to the target classification confidence, the multi-sensor environmental robustness factor, and the spatiotemporal consistency factor, respectively.

[0011] Preferably, the multi-sensor environmental robustness factor Calculate using the following formula: ; In the formula, This represents the real-time measurement variance of current multi-sensor data under environmental interference. This represents the maximum perceptual tolerance variance allowed by the system.

[0012] Preferably, the spatiotemporal consistency factor Calculate using the following formula: ; in, The sliding window Euclidean distance deviation is the difference between the predicted historical trajectory curve of the target and the actual observed trajectory in the current frame under multi-frame continuous observation.

[0013] Preferably, the dynamic safety boundary in step S4 Calculate using the following formula: ; In the formula, The basic safe topology distance is determined by the current vehicle speed; The preset safety penalty coefficient; To comprehensively perceive confidence level.

[0014] Preferably, the topological nodes of the four-dimensional spatiotemporal state diagram in step S3 Represented as: ; In the formula, Indicates the cumulative distance in the vertical direction; Indicates the lateral offset; Indicates the vehicle's heading angle; Indicates an absolute time step.

[0015] Preferably, the total trajectory cost function in step S5 Calculate using the following formula: ; In the formula, The edge risk cost is the cost of a single spatiotemporal candidate connection edge on the search path; The cumulative cost of all spatiotemporal candidate connection edges on the search path; This is a geometric constraint penalty term for the curve smoothness of the entire candidate trajectory; The preset smoothness weighting coefficient; Risk cost Calculate using the following formula:

[0016] In the formula, Spatiotemporal candidate connection edges and the dynamic safety boundary The collision risk and cost between the expanded spatiotemporal envelopes of obstacles; The trade-off for the vehicle's expected acceleration and jerk. Cost of deviation from target speed and time efficiency; , , These are the corresponding collision risk weight coefficient, comfort weight coefficient, and traffic efficiency weight coefficient.

[0017] Preferably, the collision risk cost The actual spatiotemporal distance from the candidate spatiotemporal connection edge to the obstacle exhibits a nonlinear inverse proportional mapping, and its specific quantitative calculation formula is as follows: ; In the formula, The minimum Euclidean distance between the candidate connecting edge and the spatiotemporal envelope of the obstacle; For dynamic safety boundaries.

[0018] Preferably, between step S2 and step S4, the method further includes a functional safety closed-loop degradation linkage step: Set a level 1 security threshold With Level 2 security threshold ,and ; When detected When this occurs, a Level 1 functional safety degradation control is triggered, and the system expands the dynamic safety boundary. While performing trajectory repair, the chassis is simultaneously applied to actively limit speed and increase following distance. To comprehensively perceive confidence; When detected When the system triggers the Level 2 Functional Safety Degradation Control, it refuses to perform routine trajectory repair and directly triggers the Minimum Risk Maneuvering (MRM) control strategy to control the vehicle to safely decelerate and pull over.

[0019] Secondly, the present invention provides the following technical solution: an autonomous driving spatiotemporal trajectory repair system based on perception confidence, the system comprising: The perception module is used to detect the spatiotemporal relationship between the vehicle's original planned trajectory and the future multi-probability predicted trajectory of the dynamic obstacle in real time. When a conflict is detected between the original planned trajectory and the future movement trajectory of the dynamic obstacle, the trajectory failure point is locked and the trajectory repair program is activated. The evaluation module is used to calculate the comprehensive perception confidence level in the current environment in real time based on the perception data collected by the vehicle's multiple sensors. ; The repair module is used to construct a four-dimensional spatiotemporal state diagram, including longitudinal position, lateral position, heading angle and time dimension, starting from the current actual state node of the vehicle, and to perform high-dimensional non-uniform spatiotemporal sampling in the look-ahead time to generate candidate connection edges. The adjustment module is used to adjust the overall perceived confidence level. The safety boundary is dynamically adjusted, and candidate connection edges that cross the safety boundary are safely pruned and eliminated to obtain the pruned four-dimensional spatiotemporal state diagram. The search module is used to perform a full-space-time parallel trajectory search based on the clipped four-dimensional spatiotemporal state diagram, with the goal of minimizing the total trajectory cost function, and to search for the optimal spatiotemporal discrete node sequence. The fitting module is used to perform full-time-space joint smoothing fitting on the searched optimal spatiotemporal discrete node sequence using high-order curves, and outputs the vehicle control trajectory after passing the vehicle physical constraint verification.

[0020] Thirdly, the present invention provides the following technical solution: a computer, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, it implements the above-described method for spatiotemporal trajectory repair of autonomous driving based on perception confidence.

[0021] Fourthly, the present invention provides the following technical solution: a storage medium storing a computer program, wherein when the computer program is executed by a processor, it implements the above-described method for repairing the spatiotemporal trajectory of autonomous driving based on perception confidence. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of the present invention, 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 some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 A flowchart of the spatiotemporal trajectory repair method for autonomous driving based on perception confidence provided in Embodiment 1 of the present invention; Figure 2 This is a structural block diagram of the autonomous driving spatiotemporal trajectory repair system based on perception confidence provided in Embodiment 2 of the present invention; Figure 3 This is a schematic diagram of the hardware structure of a computer provided for another embodiment of the present invention.

[0024] The embodiments of the present invention will be further described below with reference to the accompanying drawings. Detailed Implementation

[0025] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain embodiments of the present invention, and should not be construed as limiting the present invention.

[0026] Example 1 In Embodiment 1 of the present invention, as Figure 1 As shown, an autonomous driving spatiotemporal trajectory repair method based on perception confidence includes: Step S1: Real-time detection of the spatiotemporal relationship between the vehicle's original planned trajectory and the future multi-probability predicted trajectory of the dynamic obstacle. When a conflict is detected between the original planned trajectory and the future movement trajectory of the dynamic obstacle, the trajectory failure point is locked and the trajectory repair program is activated. Specifically, within each planning cycle, the system extracts the vehicle's original planned trajectory in real time and performs cross-collision verification with the future multi-probability spatiotemporal trajectories of dynamic obstacles extrapolated by the prediction module. When a conflict is detected between the vehicle's planned trajectory and the future motion trajectory of an obstacle at a future time step (i.e., overlap of the envelope), a spatiotemporal conflict is determined, the trajectory failure point is identified, and the trajectory repair process is activated.

[0027] Step S2: Based on the perception data collected by the vehicle's multi-sensor system, the evaluation module calculates the comprehensive perception confidence level in real time under the current environment. ; Specifically, the confidence assessment layer acquires heterogeneous sensing data from cameras, millimeter-wave radar, and lidar in real time. First, it uses fusion algorithms such as extended Kalman filtering to spatially register and temporally synchronize the target data from each source, outputting the target classification confidence score. Next, the real-time signal characteristics of the current sensor are extracted, and the multi-sensor environmental robustness factor is calculated. : ; In the formula, This represents the real-time measurement variance of current multi-sensor data under environmental interference. This represents the maximum perceptual tolerance variance allowed by the system. It should be noted that the real-time measurement variance of the current multi-sensor data under environmental interference... The extraction and calculation method is as follows: The signal-to-noise ratio and grayscale entropy of the current image from the vehicle-mounted camera are extracted as mapping indicators of light and rain / snow interference. Simultaneously, the noise point ratio and signal echo attenuation coefficient of the current point cloud from the lidar are extracted. The camera features and point cloud features are then weighted and their intersection calculated using a Gaussian mixture model (GMM) to quantitatively characterize the real-time measurement variance of the sensor due to environmental interference such as noise, light, rain, and snow. When the environment is harsh and noise increases... rise, decline.

[0028] Simultaneously, by comparing the historical prediction curves of the target across multiple consecutive frames with the observation overlap of the current frame, the spatiotemporal consistency factor is calculated. :

[0029] in, This represents the Euclidean distance deviation between the predicted historical trajectory of the same dynamic obstacle within the sliding window and the actual observed trajectory in the current frame. A larger deviation indicates a sudden change in the target's motion state or false alarm jitter from the sensor, affecting spatiotemporal consistency. The lower.

[0030] Finally, the overall perceived confidence level is calculated according to the following weighted formula. : ; In the formula, The target classification confidence score output by multi-sensor fusion; For multi-sensor environment robustness factor; It is a spatiotemporal consistency factor; , , These are the weight coefficients corresponding to the target classification confidence, the multi-sensor environmental robustness factor, and the spatiotemporal consistency factor, respectively. in, , , The corresponding weight relationship satisfies Simultaneously, the system establishes a closed-loop linkage mechanism between trajectory repair and functional safety degradation control: setting a first-level safety threshold. With Level 2 security threshold ,and .

[0031] Step S3: Starting from the current actual state node of the vehicle, construct a four-dimensional spatiotemporal state graph that includes longitudinal position, lateral position, heading angle and time dimension, and perform high-dimensional non-uniform spatiotemporal sampling in the look-ahead time to generate candidate connection edges. Specifically, in order to comprehensively solve dynamic intersection conflict scenarios, this invention addresses the issue in a three-dimensional topological space. Based on strong coupling time dimension Construct a four-dimensional spatiotemporal state diagram. The nodes of the topology diagram are represented as follows: ; in, Indicates the cumulative distance in the vertical direction; Indicates the lateral offset; Indicates the vehicle's heading angle; This represents the absolute time step. Starting from the current actual vehicle state node, the system performs high-dimensional non-uniform spatiotemporal sampling in the look-ahead time, that is, increasing the sampling density in dense time step regions or regions with potential conflicts. Nodes are interconnected through control edges that satisfy vehicle dynamics feasibility.

[0032] Step S4: Based on the comprehensive perceived confidence level The safety boundary is dynamically adjusted, and candidate connection edges that cross the safety boundary are safely pruned and eliminated to obtain the pruned four-dimensional spatiotemporal state diagram. The system calculates based on step S2 Real-time updates of dynamic security boundaries : ; in, The basic safe topology distance is determined by the current vehicle speed; The preset safety penalty coefficient; The overall perceived confidence level. The dynamic security boundary. With the aforementioned comprehensive perceived confidence The system dynamically expands as the threshold decreases. It traverses all candidate edges in the four-dimensional spatiotemporal state graph, and once it finds an edge traversing a completely new... If the expanded obstacle's spatiotemporal envelope is then identified as a high-risk connection edge, topology pruning will be performed immediately, and the edge will be removed from the graph.

[0033] At the same time, the functional safety layer monitors the overall perception confidence level in real time. The changing trend, i.e., the steps to achieve functional safety closed-loop degradation linkage: Set a level 1 security threshold With Level 2 security threshold ,and ; When detected When this occurs, a Level 1 functional safety degradation control is triggered, and the system expands the dynamic safety boundary. While performing trajectory repair, the chassis is simultaneously applied to actively limit speed and increase following distance. To comprehensively perceive confidence; When detected When the system triggers the Level 2 Functional Safety Degradation Control, it refuses to perform routine trajectory repair and directly triggers the Minimum Risk Maneuvering (MRM) control strategy to control the vehicle to safely decelerate and pull over.

[0034] Step S5: Based on the clipped four-dimensional spatiotemporal state diagram, perform a full spatiotemporal parallel trajectory search with the goal of minimizing the total trajectory cost function to find the optimal spatiotemporal discrete node sequence; Specifically, in the clipped four-dimensional spatiotemporal state diagram, a graph search algorithm is used to perform a full-spatiotemporal parallel trajectory search. The search process is based on the constructed trajectory total cost function. Minimization-oriented: ; In the formula, The edge risk cost is the cost of a single spatiotemporal candidate connection edge on the search path; The cumulative cost of all spatiotemporal candidate connection edges on the search path; This is a geometric constraint penalty term for the curve smoothness of the entire candidate trajectory; The preset smoothness weighting coefficient; Risk cost Calculate using the following formula:

[0035] In the formula, Spatiotemporal candidate connection edges and the dynamic safety boundary The collision risk and cost between the expanded spatiotemporal envelopes of obstacles; The trade-off for the vehicle's expected acceleration and jerk. Cost of deviation from target speed and time efficiency; , , These are the corresponding collision risk weight coefficient, comfort weight coefficient, and traffic efficiency weight coefficient.

[0036] Among them, the collision risk cost The actual spatiotemporal distance from the candidate spatiotemporal connection edge to the obstacle exhibits a nonlinear inverse proportional mapping, and its specific quantitative calculation formula is as follows: ; In the formula, The minimum Euclidean distance between the candidate connecting edge and the spatiotemporal envelope of the obstacle; This is a dynamic safety boundary. The formula guarantees that once a candidate connection edge approaches or falls below this dynamic safety boundary... When the collision cost increases exponentially, the graph search algorithm will quickly avoid the region and find the globally optimal spatiotemporally collision-free discrete node sequence.

[0037] Step S6: Use a high-order curve to perform full-time-space joint smoothing fitting on the searched optimal spatiotemporal discrete node sequence, and output the vehicle control trajectory after passing the vehicle physical constraint verification.

[0038] The system uses the searched discrete nodes as boundary conditions and employs a high-order polynomial curve or constrained quadratic programming algorithm for joint temporal and spatial smoothing fitting. After smoothing, the velocity of the entire trajectory is... acceleration accelerometer and curvature Perform rigorous physical constraint checks on the vehicle to verify whether it meets the following kinematic and dynamic limits: ; In the formula, This is the maximum permissible acceleration for the vehicle. The maximum allowable trajectory curvature for the vehicle is determined, and after verification, the smoothed control trajectory is officially output to the vehicle chassis actuator.

[0039] Meanwhile, in order to verify the practical application and effect of the present invention, the present invention provides the following scenarios: 1. Sudden Cut-in Scenario on Highway A vehicle is traveling at 100 km / h on a highway when a neighboring vehicle suddenly cuts in. Step S1 immediately detects the spatiotemporal conflict, causing trajectory determination to fail. The confidence assessment layer calculates a high classification confidence for the vehicle, but the sudden change in its motion leads to a decrease in the spatiotemporal consistency factor S. TC A brief decrease, overall perceived confidence level C final A slight decrease. Step S4 is based on the decrease in C. final The safety boundary is dynamically expanded, and the four-dimensional spatiotemporal map automatically trims and removes high-risk edges such as those that force a detour. Step S5 searches for an optimal smooth trajectory in the trimmed spatiotemporal map, which involves longitudinal deceleration and speed control to avoid the obstacle, followed by a gentle lane change, allowing the vehicle to avoid the obstacle smoothly.

[0040] 2. A sudden "ghost peek" incident in the city center: a pedestrian suddenly appears and crosses the street. A car is traveling at 40 km / h on a city road with blind spots on both sides. A pedestrian suddenly runs across the road at high speed. Due to the blind spots and the suddenness of the event, multiple sensors experience instantaneous observation time differences, which reduces the multi-sensor environmental robustness factor R. env Spatiotemporal consistency factor S TC A sharp drop, with a comprehensive confidence level of C. finalThe vehicle rapidly degrades to below the first-level degradation threshold T1. The functional safety layer immediately intervenes in a closed-loop manner to execute the first-level degradation, forcibly issuing the maximum safe braking and deceleration command, and widening the forward topology boundary. In the four-dimensional graph space that has been significantly topologically cropped, the spatiotemporal trajectory repair module quickly converges to a smooth trajectory of "emergency braking and deceleration avoidance," and the vehicle finally comes to a stop within a safe range.

[0041] 3. Low visibility scenarios during extreme weather events such as heavy rain. The vehicle encountered a severe rainstorm on the highway. The onboard camera's image became extremely blurry due to water stains on the lens, and the lidar point cloud suffered extensive raindrop noise scattering. Multiple sensors were also involved in real-time variance measurement. Approaching the maximum tolerance limit leads to an environmental robustness factor R. env Reduced to extremely low, overall perceived confidence level C final The vehicle's speed dropped below the highest safety threshold (T2) for Level 2. The Functional Safety Layer was instantly activated and took over vehicle control. Determining that the current environment could not guarantee the safety of a conventional repair trajectory, the Minimum Risk Maneuver (MRM) control strategy was directly triggered. The system activated its hazard lights and, in the full-time map, retained only the safe topology branch for slowing down and moving to the rightmost emergency lane, controlling the vehicle to safely pull over and stop.

[0042] 4. Complex detour scenarios for construction sections As the vehicle approached the section of road undergoing sudden construction, it encountered a dense but irregularly arranged array of reflective cones and crash barriers. Due to the irregular characteristics of the obstacles, the camera's target classification confidence fluctuated frequently, and the radar's multipath reflections caused inconsistencies in spatial observation, resulting in an overall perception confidence level of C. final It is in a critical state with moderate fluctuations. The dynamic safety boundary D calculated by the system is... safety This results in a dynamic, slight expansion. Based on the expanded safety boundary, the spatiotemporal graph construction module removes all high-risk candidate connecting edges attempting to accurately weave through the narrow gaps between reflector cones. The graph search algorithm then searches for a smooth, safe detour trajectory within the remaining, safe far-side topological space after the removal, ensuring the vehicle does not scratch any construction facilities.

[0043] 5. Scenario of the vehicle in front braking unexpectedly. The vehicle is following another vehicle at 80 km / h on an urban expressway when the other vehicle suddenly brakes sharply, decelerating drastically. Step S1 detects that if the vehicle continues along its original trajectory, a severe rear-end collision will occur with the other vehicle in the near future. At this point, the multi-sensor tracking of the other vehicle's features is clear, with extremely high classification confidence. However, the extremely high jerk of the other vehicle causes a significant increase in the spatiotemporal consistency factor S of the forward observation. TCA brief drop caused a moderate increase in the safety margin. In the four-dimensional spatiotemporal state graph, the system instantly pruned and removed all lateral blind lane-changing connecting edges and forward topological edges maintaining the original speed. The graph search algorithm instantly located the optimal solution in the entire spatiotemporal graph online: a longitudinal collision avoidance trajectory that maximizes the vehicle's dynamic deceleration limits, controlling the vehicle to closely follow the vehicle in front to complete the maximum smooth deceleration, successfully avoiding a rear-end collision.

[0044] The first embodiment of this invention provides a spatiotemporal trajectory repair method for autonomous driving based on perception confidence. This invention simultaneously establishes a closed-loop linkage mechanism between trajectory repair and functional safety degradation control. When the overall perception confidence decreases due to environmental interference or deteriorating observation consistency of multiple sensors, the system not only dynamically expands the topological boundary to increase safety redundancy but also automatically triggers multi-level functional safety control logic, executing active speed limiting, increasing following distance, or minimum risk maneuver (MRM), thereby improving the functional safety level of autonomous driving and the real-time performance of trajectory repair.

[0045] Example 2 like Figure 2 As shown, in Embodiment 2 of the present invention, an autonomous driving spatiotemporal trajectory repair system based on perception confidence is provided, the system comprising: The perception module 1 is used to detect the spatiotemporal relationship between the vehicle's original planned trajectory and the future multi-probability predicted trajectory of the dynamic obstacle in real time. When a conflict is detected between the original planned trajectory and the future movement trajectory of the dynamic obstacle, the trajectory failure point is locked and the trajectory repair program is activated. Evaluation module 2 is used to calculate the comprehensive perception confidence level in real time under the current environment based on the perception data collected by the vehicle's multi-sensor system. ; Repair module 3 is used to construct a four-dimensional spatiotemporal state diagram that includes longitudinal position, lateral position, heading angle and time dimension, starting from the current actual state node of the vehicle, and to perform high-dimensional non-uniform spatiotemporal sampling in the look-ahead time to generate candidate connection edges. Adjustment module 4 is used to adjust the overall perceived confidence level. The safety boundary is dynamically adjusted, and candidate connection edges that cross the safety boundary are safely pruned and eliminated to obtain the pruned four-dimensional spatiotemporal state diagram. Search module 5 is used to perform full-space-time parallel trajectory search based on the clipped four-dimensional spatio-temporal state diagram with the goal of minimizing the total trajectory cost function, and to search for the optimal spatio-temporal discrete node sequence. Fitting module 6 is used to perform full-time-space joint smoothing fitting on the searched optimal spatiotemporal discrete node sequence using high-order curves, and output the vehicle control trajectory after passing the vehicle physical constraint verification.

[0046] Among them, the comprehensive perceived confidence level Calculate using the following formula: ; In the formula, The target classification confidence score output by multi-sensor fusion; For multi-sensor environment robustness factor; It is a spatiotemporal consistency factor; , , These are the weight coefficients corresponding to the target classification confidence, the multi-sensor environmental robustness factor, and the spatiotemporal consistency factor, respectively.

[0047] Among them, the multi-sensor environmental robustness factor Calculate using the following formula: ; In the formula, This represents the real-time measurement variance of current multi-sensor data under environmental interference. This represents the maximum perceptual tolerance variance allowed by the system.

[0048] Among them, the spatiotemporal consistency factor Calculate using the following formula: ; in, The sliding window Euclidean distance deviation is the difference between the predicted historical trajectory curve of the target and the actual observed trajectory in the current frame under multi-frame continuous observation.

[0049] Wherein, the dynamic security boundary Calculate using the following formula: ; In the formula, The basic safe topology distance is determined by the current vehicle speed; The preset safety penalty coefficient; To comprehensively perceive confidence level.

[0050] Among them, the topological nodes of the four-dimensional spatiotemporal state diagram Represented as: ; In the formula, Indicates the cumulative distance in the vertical direction; Indicates the lateral offset; Indicates the vehicle's heading angle; Indicates an absolute time step.

[0051] Wherein, the total cost function of the trajectory Calculate using the following formula: ; In the formula, The edge risk cost is the cost of a single spatiotemporal candidate connection edge on the search path; The cumulative cost of all spatiotemporal candidate connection edges on the search path; This is a geometric constraint penalty term for the curve smoothness of the entire candidate trajectory; The preset smoothness weighting coefficient; Risk cost Calculate using the following formula:

[0052] In the formula, Spatiotemporal candidate connection edges and the dynamic safety boundary The collision risk and cost between the expanded spatiotemporal envelopes of obstacles; The trade-off for the vehicle's expected acceleration and jerk. Cost of deviation from target speed and time efficiency; , , These are the corresponding collision risk weight coefficient, comfort weight coefficient, and traffic efficiency weight coefficient.

[0053] Among them, the collision risk cost The actual spatiotemporal distance from the candidate spatiotemporal connection edge to the obstacle exhibits a nonlinear inverse proportional mapping, and its specific quantitative calculation formula is as follows: ; In the formula, The minimum Euclidean distance between the candidate connecting edge and the spatiotemporal envelope of the obstacle; For dynamic safety boundaries.

[0054] The system also includes a closed-loop control module, which is used for: Set a level 1 security threshold With Level 2 security threshold ,and ; When detected When this occurs, a Level 1 functional safety degradation control is triggered, and the system expands the dynamic safety boundary. While performing trajectory repair, the chassis is simultaneously applied to actively limit speed and increase following distance. To comprehensively perceive confidence; When detected When the system triggers the Level 2 Functional Safety Degradation Control, it refuses to perform routine trajectory repair and directly triggers the Minimum Risk Maneuvering (MRM) control strategy to control the vehicle to safely decelerate and pull over.

[0055] In other embodiments of the present invention, the present invention provides the following technical solution: a computer, including a memory 102, a processor 101, and a computer program stored in the memory 102 and executable on the processor 101, wherein the processor 101 executes the computer program to implement the above-described method for repairing the spatiotemporal trajectory of autonomous driving based on perception confidence.

[0056] Specifically, the processor 101 may include a central processing unit (CPU), or an application specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement the embodiments of the present invention.

[0057] The memory 102 may include a large-capacity memory for data or instructions. For example, and not limitingly, the memory 102 may include a hard disk drive (HDD), a floppy disk drive, a solid-state drive (SSD), flash memory, an optical disk drive, a magneto-optical disk drive, magnetic tape, or a Universal Serial Bus (USB) drive, or a combination of two or more of these. Where appropriate, the memory 102 may include removable or non-removable (or fixed) media. Where appropriate, the memory 102 may be internal or external to a data processing device. In a particular embodiment, the memory 102 is non-volatile memory. In a particular embodiment, the memory 102 includes read-only memory (ROM) and random access memory (RAM). Where appropriate, the ROM may be a mask-programmed ROM, a programmable read-only memory (PROM), an erasable read-only memory (EPROM), an electrically erasable read-only memory (EEPROM), an electrically alterable read-only memory (EAROM), or flash memory, or a combination of two or more of these. Where appropriate, the RAM can be Static Random-Access Memory (SRAM) or Dynamic Random-Access Memory (DRAM). DRAM can be Fast Page Mode Dynamic Random Access Memory (FPMDRAM), Extended Data Out Dynamic Random Access Memory (EDODRAM), Synchronous Dynamic Random-Access Memory (SDRAM), etc.

[0058] The memory 102 can be used to store or cache various data files that need to be processed and / or used for communication, as well as possible computer program instructions executed by the processor 101.

[0059] The processor 101 reads and executes the computer program instructions stored in the memory 102 to implement the above-mentioned method for repairing the spatiotemporal trajectory of autonomous driving based on perception confidence.

[0060] In some embodiments, the computer may further include a communication interface 103 and a bus 100. For example, Figure 3 As shown, the processor 101, memory 102, and communication interface 103 are connected through bus 100 and communicate with each other.

[0061] The communication interface 103 is used to enable communication between the various modules, devices, units, and / or equipment in the embodiments of the present invention. The communication interface 103 can also enable data communication with other components such as external devices, image / data acquisition devices, databases, external storage, and image / data processing workstations.

[0062] Bus 100 includes hardware, software, or both, that couples components of a computer device together. Bus 100 includes, but is not limited to, at least one of the following: data bus, address bus, control bus, expansion bus, and local bus. For example, and not as a limitation, bus 100 may include an Accelerated Graphics Port (AGP) or other graphics bus, an Extended Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), a Hyper Transport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an InfiniBand interconnect, a Low Pin Count (LPC) bus, a memory bus, a Micro Channel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local Bus (VLB) bus, or other suitable buses, or a combination of two or more of these. Where appropriate, bus 100 may include one or more buses. Although specific buses are described and illustrated in the embodiments of the present invention, the present invention is contemplated by any suitable bus or interconnect.

[0063] The computer can execute the autonomous driving spatiotemporal trajectory repair method based on perception confidence of the present invention based on perception confidence, thereby realizing autonomous driving spatiotemporal trajectory repair based on perception confidence.

[0064] In some further embodiments of the present invention, in conjunction with the above-described method for repairing the spatiotemporal trajectory of autonomous driving based on perception confidence, the present invention provides the following technical solution: a storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the above-described method for repairing the spatiotemporal trajectory of autonomous driving based on perception confidence.

[0065] Those skilled in the art will understand that the logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as a ordered 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 mean any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0066] More specific examples of readable media (a non-exhaustive list) include: electrical connections (electronic devices) with 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, since 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.

[0067] 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.

[0068] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0069] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this invention patent should be determined by the appended claims.

Claims

1. A method for spatiotemporal trajectory repair of autonomous driving based on perception confidence, characterized in that, include: Step S1: Real-time detection of the spatiotemporal relationship between the vehicle's original planned trajectory and the future multi-probability predicted trajectory of the dynamic obstacle. When a conflict is detected between the original planned trajectory and the future movement trajectory of the dynamic obstacle, the trajectory failure point is locked and the trajectory repair program is activated. Step S2: Based on the perception data collected by the vehicle's multi-sensor system, the comprehensive perception confidence level in the current environment is calculated in real time by the evaluation module. ; Step S3: Starting from the current actual state node of the vehicle, construct a four-dimensional spatiotemporal state graph that includes longitudinal position, lateral position, heading angle and time dimension, and perform high-dimensional non-uniform spatiotemporal sampling in the look-ahead time to generate candidate connection edges. Step S4: Based on the comprehensive perceived confidence level The safety boundary is dynamically adjusted, and candidate connection edges that cross the safety boundary are safely pruned and eliminated to obtain the pruned four-dimensional spatiotemporal state diagram. Step S5: Based on the clipped four-dimensional spatiotemporal state diagram, perform a full spatiotemporal parallel trajectory search with the goal of minimizing the total trajectory cost function to find the optimal spatiotemporal discrete node sequence; Step S6: Use a high-order curve to perform full-time-space joint smoothing fitting on the searched optimal spatiotemporal discrete node sequence, and output the vehicle control trajectory after passing the vehicle physical constraint verification.

2. The method for spatiotemporal trajectory repair of autonomous driving based on perception confidence according to claim 1, characterized in that, The comprehensive perceived confidence level in step S2 Calculate using the following formula: ； In the formula, The target classification confidence score output by multi-sensor fusion; For multi-sensor environment robustness factor; It is a spatiotemporal consistency factor; , , These are the weight coefficients corresponding to the target classification confidence, the multi-sensor environmental robustness factor, and the spatiotemporal consistency factor, respectively.

3. The method for spatiotemporal trajectory repair of autonomous driving based on perception confidence according to claim 2, characterized in that, The multi-sensor environmental robustness factor Calculate using the following formula: ； In the formula, This represents the real-time measurement variance of current multi-sensor data under environmental interference. This represents the maximum perceptual tolerance variance allowed by the system.

4. The method for spatiotemporal trajectory repair of autonomous driving based on perception confidence according to claim 2, characterized in that, The spatiotemporal consistency factor Calculate using the following formula: ； in, The sliding window Euclidean distance deviation is the difference between the predicted historical trajectory curve of the target and the actual observed trajectory in the current frame under multi-frame continuous observation.

5. The method for spatiotemporal trajectory repair of autonomous driving based on perception confidence according to claim 1, characterized in that, The dynamic safety boundary in step S4 Calculate using the following formula: ； In the formula, The basic safe topology distance is determined by the current vehicle speed; The preset safety penalty coefficient; To comprehensively perceive confidence level.

6. The method for spatiotemporal trajectory repair of autonomous driving based on perception confidence according to claim 1, characterized in that, The topological nodes of the four-dimensional spatiotemporal state diagram in step S3 Represented as: ； In the formula, Indicates the cumulative distance in the vertical direction; Indicates the lateral offset; Indicates the vehicle's heading angle; Indicates an absolute time step.

7. The method for spatiotemporal trajectory repair of autonomous driving based on perception confidence according to claim 1, characterized in that, The total cost function of the trajectory in step S5 Calculate using the following formula: ； In the formula, The edge risk cost is the cost of a single spatiotemporal candidate connection edge on the search path; The cumulative cost of all spatiotemporal candidate connection edges on the search path; This is a geometric constraint penalty term for the curve smoothness of the entire candidate trajectory; The preset smoothness weighting coefficient; Risk cost Calculate using the following formula: In the formula, Spatiotemporal candidate connection edges and the dynamic safety boundary The collision risk and cost between the expanded spatiotemporal envelopes of obstacles; The trade-off for the vehicle's expected acceleration and jerk. Cost of deviation from target speed and time efficiency; , , These are the corresponding collision risk weight coefficient, comfort weight coefficient, and traffic efficiency weight coefficient.

8. The method for spatiotemporal trajectory repair of autonomous driving based on perception confidence according to claim 7, characterized in that, The cost of collision risk The actual spatiotemporal distance from the candidate spatiotemporal connection edge to the obstacle exhibits a nonlinear inverse proportional mapping, and its specific quantitative calculation formula is as follows: ； In the formula, The minimum Euclidean distance between the candidate connecting edge and the spatiotemporal envelope of the obstacle; For dynamic safety boundaries.

9. The method for spatiotemporal trajectory repair of autonomous driving based on perception confidence according to claim 1, characterized in that, Between step S2 and step S4, the method further includes a functional safety closed-loop degradation linkage step: Set a level 1 security threshold With Level 2 security threshold ,and ; When detected When this occurs, a Level 1 functional safety degradation control is triggered, and the system expands the dynamic safety boundary. While performing trajectory repair, the chassis is simultaneously implementing active speed limiting and increasing following distance. To comprehensively perceive confidence; When detected When the system triggers the Level 2 Functional Safety Degradation Control, it refuses to perform routine trajectory repair and directly triggers the Minimum Risk Maneuvering (MRM) control strategy to control the vehicle to safely decelerate and pull over to the side of the road.

10. An autonomous driving spatiotemporal trajectory repair system based on perception confidence, characterized in that, The system includes: The perception module is used to detect the spatiotemporal relationship between the vehicle's original planned trajectory and the future multi-probability predicted trajectory of the dynamic obstacle in real time. When a conflict is detected between the original planned trajectory and the future movement trajectory of the dynamic obstacle, the trajectory failure point is locked and the trajectory repair program is activated. The evaluation module is used to calculate the comprehensive perception confidence level in the current environment in real time based on the perception data collected by the vehicle's multiple sensors. ; The repair module is used to construct a four-dimensional spatiotemporal state diagram, including longitudinal position, lateral position, heading angle and time dimension, starting from the current actual state node of the vehicle, and to perform high-dimensional non-uniform spatiotemporal sampling in the look-ahead time to generate candidate connection edges. The adjustment module is used to adjust the overall perceived confidence level. The safety boundary is dynamically adjusted, and candidate connection edges that cross the safety boundary are safely pruned and eliminated to obtain the pruned four-dimensional spatiotemporal state diagram. The search module is used to perform a full-space-time parallel trajectory search based on the clipped four-dimensional spatiotemporal state diagram, with the goal of minimizing the total trajectory cost function, and to search for the optimal spatiotemporal discrete node sequence. The fitting module is used to perform full-time-space joint smoothing fitting on the searched optimal spatiotemporal discrete node sequence using high-order curves, and outputs the vehicle control trajectory after passing the vehicle physical constraint verification.

Images