An automatic driving system and driving method based on a big data map library

Abstract

The application relates to the technical field of automatic driving, in particular to an automatic driving system and method based on a big data map library, in which map library data of a local road region corresponding to a current position of a vehicle in the big data map library and real-time sensing data of the vehicle on the local road region are acquired, road topology data and corridor verification data are extracted, and a candidate passing region is generated; further, a local passable corridor credibility field is generated through consistency analysis; on this basis, passable corridor reconstruction, target trajectory generation, control continuity adjustment of a horizontal control process and a longitudinal control process corresponding to the target trajectory, generation of a vehicle control instruction, and updating of the map library data in combination with passing feedback data are completed, so that the synergy between passable corridor reconstruction, target trajectory generation, vehicle horizontal and longitudinal control adjustment, and map library data updating processing under the condition that there is a local difference between the map library data and the real-time sensing data is improved.

Description

Technical Field

[0001] This invention relates to the field of autonomous driving technology, specifically to an autonomous driving system and driving method based on a large-scale map database. Background Technology

[0002] During operation, autonomous vehicles typically rely on a combination of information, including vehicle location data, road map data, environmental perception data, trajectory planning results, and vehicle control status, to make autonomous driving decisions. Road map data provides prior information such as lane centerlines, lane boundaries, road boundaries, lane connections, intersection connections, steerable lanes, and road restrictions. Environmental perception data reflects real-time road conditions around the vehicle, including lane lines, road boundaries, obstacles, dynamic targets, traffic facilities, and perceptible drivable areas. Through the collaborative processing of road map data and environmental perception data, autonomous vehicles can determine the drivable space within a local road area and generate trajectories and control the vehicle based on this drivable space.

[0003] In relevant autonomous driving solutions, the road structure near the vehicle's current location is typically determined based on road topology information in a map database, and a reference path or candidate driving area is generated by combining this with the vehicle's current positioning information. Subsequently, based on real-time environmental information collected by onboard sensing devices, obstacles, road boundaries, dynamic targets, or temporary road conditions in the candidate driving area are identified, and the vehicle's planned trajectory and control parameters are adjusted accordingly. Some solutions also collect information on the vehicle's actual trajectory, road perception results, and road condition changes during vehicle operation, and use this information for map data verification or road data updates to improve the consistency between map data and the actual road conditions.

[0004] However, in complex road environments, temporary closures, road construction, changes in road boundaries, sensor occlusion, map data update delays, or multi-branch road topologies, local discrepancies may arise between map database data and real-time vehicle perception data. In such cases, determining drivable space solely based on road topology or local perception results can easily fail to adequately reflect the reliability of different spatial locations within candidate drivable areas in the reconstruction of drivable corridors, target trajectory generation, and lateral and longitudinal control adjustments. This results in insufficient coordination between drivable corridors, target trajectories, and vehicle control commands. When vehicles pass through areas with significant changes in confidence, unconfirmed areas, or areas adjacent to low-confidence areas, trajectory adjustments may be less smooth, control parameter changes may be discontinuous, or safety margins may be insufficient. Furthermore, if the feedback data generated during vehicle passage cannot establish an effective update relationship with the map database data for the corresponding local road area, it will also affect the adaptability of map data to actual road conditions during subsequent cyclical driving. Summary of the Invention

[0005] The purpose of this invention is to provide an autonomous driving system and driving method based on a big data map database, in order to solve the problems mentioned in the background art. Specifically, it addresses the problem that insufficient coordination between reconstructing passable corridors, generating target trajectories, adjusting vehicle lateral and longitudinal control, and updating map database data when there are local differences between map database data and real-time perception data, which affects the continuous and stable passage of vehicles in local road areas.

[0006] To achieve the above objectives, one objective of this invention is to provide an autonomous driving method based on a large-scale map database, comprising the following method steps: S1. Obtain map data of the local road area corresponding to the vehicle's current location in the big data map database, and obtain real-time perception data of the vehicle of the local road area; specifically including: Obtain the vehicle's current location information, which includes the vehicle's current location coordinates, heading angle, speed, and road sign information. Determine the local road area corresponding to the vehicle's current location based on the vehicle's current location information; Send data requests corresponding to local road areas to the big data map library, and receive map library data returned by the big data map library; The vehicle-mounted sensing devices collect environmental data within the sensing coverage area of ​​a local road area; Environmental data is preprocessed, feature extracted, target associated, and multi-source fusion processed to obtain real-time perception data including lane lines, road boundaries, obstacles, obstacle-occupied areas, dynamic targets, dynamic target positions, dynamic target speeds, dynamic target movement directions, traffic facilities, and real-time perception traffic areas. Map database data and real-time perception data are uniformly converted to the vehicle coordinate system for coordinate alignment.

[0007] Step S1 obtains map library data of the local road area corresponding to the vehicle's current location in the big data map library, and simultaneously obtains real-time perception data of the vehicle on the local road area, providing a data foundation for subsequent comparison of map library data and real-time perception data under the same coordinate basis; through coordinate alignment processing, it is beneficial to reduce the consistency analysis deviation caused by the different coordinate references of the two types of data.

[0008] S2. Extract road topology data and corridor verification data from the map database, specifically including: Read the lane centerline, lane boundary line, road boundary, lane width, lane direction, lane connection relationship, intersection connection relationship, turning relationship and road restriction relationship within a local road area from the map database data; Filter road topology elements that are consistent with the vehicle's current direction of travel based on the vehicle's current position and heading angle; A local road topology map is constructed based on lane connection relationships and intersection connection relationships, and the local road topology map, lane geometric boundaries, and road connection relationships are used as road topology data. Read data fields related to the traffic status of local road areas from the map database, and filter valid data fields based on the vehicle's current location, the range of the local road area, and the data timestamp; The filtered historical traffic trajectories, road change records, and vehicle group perception results are mapped to the spatial units and road topology elements corresponding to the local road area to form corridor verification data.

[0009] Candidate traffic areas are generated based on road topology data; specifically including: Based on the lane centerlines, lane boundary lines, and road connection relationships in the road topology data, determine the lane segments and road connection segments that the vehicle can reach from its current location; Boundary correction processing is performed on the reachable lane sections and road connection sections; The road space after boundary correction is matched with vehicle motion constraints, road direction constraints and lane connection constraints, and the road space that meets the conditions of vehicle passage width, minimum turning radius and continuous connectivity is retained as candidate passage area; When there are multiple drivable branches in the road topology data, candidate traffic areas corresponding to each drivable branch are generated, and the topological connection relationship between each candidate traffic area is preserved. When there are prohibited, closed, or inaccessible areas in the road topology data, these areas will be removed from the candidate traffic areas.

[0010] Step S2 extracts road topology data and corridor verification data from the map database, and generates candidate traffic areas based on the road topology data. This ensures that the candidate traffic areas are constrained by road structure factors such as lane connection relationships, intersection connection relationships, and road traffic restriction relationships, while also retaining corridor verification data for subsequent verification of traffic status. This step is beneficial in providing candidate ranges with a road topology basis for subsequent traffic credibility evaluation when there are local differences between the map database data and real-time perception data.

[0011] S3. Perform consistency analysis between the corridor verification data and real-time sensing data and the candidate passage areas to generate a local passable corridor confidence field corresponding to the candidate passage areas; specifically including: The candidate traffic areas are divided into spatial units, and the corridor verification data is mapped to the spatial units corresponding to the candidate traffic areas to obtain the historical traffic characteristics, road change characteristics and vehicle group perception characteristics corresponding to each spatial unit. For each spatial unit, the historical traffic consistency value, road change consistency value, and vehicle group perception consistency value are calculated, and the historical traffic consistency value, road change consistency value, and vehicle group perception consistency value are weighted and fused to obtain the corridor verification consistency result; Map lane lines, road boundaries, obstacles, dynamic targets, traffic facilities, and real-time perceived traffic areas from real-time sensing data to the spatial units corresponding to candidate traffic areas; For each spatial unit, the boundary consistency value, obstacle occupancy impact value, dynamic risk consistency value, and real-time perceived passage area consistency value are calculated. The boundary consistency value, dynamic risk consistency value, and real-time perceived passage area consistency value are then fused, and suppression processing is performed based on the obstacle occupancy impact value to obtain the real-time perception consistency result. The corridor verification weight is determined based on the timeliness, data source reliability, and spatial coverage integrity of the corridor verification data. The real-time perception weight is determined based on the sensor confidence, target recognition confidence, occlusion status, and perception coverage of the real-time perception data. For each spatial unit, the corridor verification consistency result and the real-time perception consistency result are weighted and fused according to the corridor verification weight and the real-time perception weight to obtain the passability confidence. Arrange the spatial locations of all spatial units and their corresponding passability credibility according to spatial distribution to generate a local passability corridor credibility field.

[0012] Step S3 generates a local passable corridor credibility field corresponding to the candidate passable area by performing consistency analysis on the corridor verification data and real-time perception data with the candidate passable area, so that different spatial units in the candidate passable area have corresponding passable credibility. This step can integrate the corridor verification results from the map database data side and the perception consistency results from the real-time perception data side into the same credibility field, providing a common basis for passable corridor reconstruction, target trajectory generation and vehicle control command generation.

[0013] S4. Reconstruct the passageways of candidate passage areas based on the local passable corridor confidence field to generate passable corridors; specifically including: Receive the local passable corridor confidence field and candidate passable areas; Based on the passability confidence of each spatial unit in the local passable corridor confidence field, the candidate passable areas are divided into high-confidence passable areas, unconfirmed passable areas, and low-confidence passable areas. Remove low-confidence traffic areas from the candidate traffic areas and retain high-confidence traffic areas that are connected to the vehicle's current location and target direction of travel; The safety boundary of the area to be confirmed for passage is contracted. A continuous passable space is generated based on the highly reliable passable area and the unconfirmed passable area after the security boundary shrinkage process. Perform boundary smoothing and topological connectivity checks on continuous passable spaces, and remove local spaces that do not meet the requirements for minimum turning radius, vehicle width, and preset safety distance. When multiple consecutive passable spaces exist, the selection is based on the average passability confidence of each consecutive passable space, the directional deviation from the target driving direction, the connectivity in the road topology data, and the curvature change of the corresponding reference path. The consecutive passable spaces that meet the requirements are determined as passable corridors. When there is no continuous passable space that meets the requirements for normal passage, a restricted passable corridor is generated in the safe parking area adjacent to the vehicle's current location, and the restricted passable corridor is output as a passable corridor.

[0014] Step S4 reconstructs the travel corridors of candidate travel areas based on the local travel corridor confidence field, generating travel corridors. This allows high-confidence travel areas, unconfirmed travel areas, and low-confidence travel areas to be processed differently during the travel corridor generation process. This step is beneficial for filtering, narrowing, and connectivity processing of candidate travel areas under confidence constraints in areas where map database data and real-time perception data are locally inconsistent, thereby improving the matching degree between travel corridors and actual travel status.

[0015] S5. Generate the target trajectory based on the local traversable corridor confidence field and traversable corridors, specifically including: Receive accessible corridors, a confidence field of partially accessible corridors, the current vehicle status, and the autonomous driving task objective; Multiple candidate trajectories are generated within the passable corridor, and the candidate trajectories satisfy vehicle kinematic constraints, road boundary constraints, obstacle avoidance constraints, and traffic rule constraints. For each candidate trajectory, the passability confidence level of the spatial unit traversed by the candidate trajectory is calculated, and the average passability confidence level, minimum passability confidence level, degree of change in passability confidence level, and proportion of neighboring areas of low-confidence passability areas are calculated to obtain the trajectory confidence level characteristics. For each candidate trajectory, the trajectory length, curvature change, target direction deviation, obstacle distance margin, and trajectory reliability characteristics are calculated, and the comprehensive value of multiple candidate trajectories is ranked according to a unified preset evaluation rule. The candidate trajectory with the best comprehensive cost and that meets the preset constraints is determined as the target trajectory; When the accessible corridor is a restricted accessible corridor, a target trajectory is generated with the safe parking area as the target location.

[0016] Based on the local traversable corridor confidence field, the lateral and longitudinal control processes corresponding to the target trajectory are adjusted for control continuity to generate vehicle control commands; specifically including: Based on the curvature, lateral deviation, heading deviation, and local navigable corridor confidence field of the target trajectory, the range of target rotation angle change corresponding to the lateral control process is determined; When the target trajectory passes through a high-confidence passage area, normal lateral control parameters are generated according to the curvature of the target trajectory. When the target trajectory passes through the area to be confirmed, the rate of change of steering wheel angle is limited, and the lateral offset of the target trajectory is smoothed. When the target trajectory passes through an area where the rate of change of passability confidence reaches a preset rate of change threshold, the rate of change of steering wheel angle is limited, and the lateral offset of the target trajectory is smoothed. When the target trajectory is near the boundary of a low-confidence passage area, the lateral safety margin is increased, the lateral control target is adjusted, and lateral control parameters are generated. Based on the expected velocity of the target trajectory, the distance to the obstacle ahead, the road speed limit, the road gradient, and the confidence field of the local passable corridor, the target velocity and target acceleration corresponding to the longitudinal control process are determined. When the passability confidence along the target trajectory is stable and reaches the first confidence threshold, normal longitudinal control parameters are generated. When there are areas along the target trajectory that need to be confirmed for passage, areas adjacent to low-confidence passage areas, or areas where the credibility of passage decreases rapidly, the current target speed is reduced, and the rate of change of acceleration and the rate of change of deceleration are limited. When the local passable corridor confidence field indicates that there is a spatial unit ahead of the target trajectory with a passability confidence level lower than the second confidence threshold, a deceleration and parking control target is generated, and longitudinal control parameters are obtained. Receive target trajectory, lateral control parameters and longitudinal control parameters, convert lateral control parameters into steering execution commands and longitudinal control parameters into drive execution commands and braking execution commands; Time synchronization processing is performed on steering execution commands, drive execution commands, and braking execution commands; When there is a conflict between the lateral control parameters and the longitudinal control parameters, the target speed should be reduced according to the principle of safety priority, while maintaining the executability of the lateral control process; Output vehicle control commands, including steering execution commands, drive execution commands, and braking execution commands.

[0017] Step S5 generates the target trajectory based on the local traversable corridor confidence field and the traversable corridor, and adjusts the control continuity of the lateral and longitudinal control processes corresponding to the target trajectory based on the local traversable corridor confidence field to generate vehicle control commands. This step ensures that the target trajectory generation, lateral control process, and longitudinal control process are all constrained by the same local traversable corridor confidence field, which is beneficial for improving the coordination between trajectory planning and vehicle control when passing through areas to be confirmed, areas adjacent to low-confidence traversable areas, or areas with large changes in traversable confidence.

[0018] S6. Collect traffic feedback data generated by vehicles traveling within a local road area according to vehicle control commands, and update the map database data for the corresponding local road area in the big data map database based on the traffic feedback data; specifically including: During the process of the vehicle driving according to the vehicle control command, the actual driving trajectory, actual speed, actual acceleration, actual steering angle, actual braking status and actual driving status of the vehicle are collected; The system collects real-time identification results of vehicles in local road areas, including road boundaries, lane lines, obstacles, temporary closed areas, construction areas, and dynamic targets. Collect the trajectory deviation between the target trajectory and the actual driving trajectory of the vehicle, and collect the control response information during the execution of vehicle control commands; The actual driving trajectory, real-time recognition results, trajectory deviation and control response information are timestamped and coordinates are unified to form traffic feedback data. The traffic feedback data is subjected to integrity verification, confidence verification, and spatial consistency verification. Traffic feedback data that meets the requirements is marked as valid traffic feedback data, and traffic feedback data with partial missing parts but still able to reflect the road change status is marked as traffic feedback data to be merged. Based on valid traffic feedback data, identify road conditions in local road areas that are inconsistent with map database data to obtain map difference information; The map discrepancy information is matched with the map database data of the corresponding local road area in the big data map database to determine the data fields that need to be updated. When map discrepancy information is generated from valid single-vehicle passage feedback data and the corresponding confidence level is higher than the preset high confidence level threshold, the corresponding data field is temporarily updated, and the data source and update time are marked. When map discrepancy information is jointly verified by multi-vehicle traffic feedback data, the corresponding data fields are merged and updated, and the credibility level of the updated data fields is improved. When map discrepancy information conflicts with existing map database data and the corresponding confidence level does not reach the preset high confidence threshold, the corresponding data field is marked as pending confirmation, and the existing map database data remains unchanged. The update processing results are written into the map database data of the corresponding local road area in the big data map database; When the big data map library adopts a cache update mechanism, the update processing result is synchronously written to the update cache area associated with the big data map library, and the association between the update processing result and the map library data of the corresponding local road area in the big data map library is established, so that when the subsequent step S1 obtains the map library data of the corresponding local road area, the update processing result in the update cache area is called at the same time.

[0019] Step S6 collects traffic feedback data generated by the vehicle during its travel within a local road area according to vehicle control commands, and updates the map database data for the corresponding local road area in the big data map database based on the traffic feedback data. This step enables the use of traffic feedback data formed by the vehicle's actual driving trajectory, real-time recognition results, trajectory deviation, and control response information for map difference information identification and map database data update processing, which helps improve the consistency between the map database data obtained in subsequent step S1 and the actual state of the local road area.

[0020] S7. If the autonomous driving task has not ended, the execution will be repeated. If the execution is repeated, the execution will return to steps S1 to S6 based on the local road area corresponding to the updated vehicle location. If the execution is not repeated, the execution will end. Step S7 uses the condition that the autonomous driving task has not ended as the condition for loop execution. After the vehicle position is updated, it returns to execute steps S1 to S6, so that map database data acquisition, real-time perception data acquisition, local passable corridor confidence field generation, passable corridor reconstruction, vehicle control command generation, and map database data update processing form a loop process. This step is conducive to maintaining dynamic connection between local road area changes, actual vehicle traffic feedback, and subsequent autonomous driving decisions during continuous vehicle driving.

[0021] The second objective of this invention is to provide a system for an autonomous driving method based on a large-scale map database, comprising a local data acquisition module, a candidate traffic area generation module, a confidence field generation module, a traffic corridor reconstruction module, a trajectory planning and control module, a map feedback update module, and a cyclic scheduling module, wherein: The local data acquisition module acquires map data of the local road area corresponding to the vehicle's current location in the big data map library, and acquires real-time perception data of the vehicle of the local road area; The candidate traffic area generation module extracts road topology data and corridor verification data from the map library data, and generates candidate traffic areas based on the road topology data; The credibility field generation module performs consistency analysis between the corridor verification data and real-time perception data and the candidate passage areas, respectively, and generates a local passable corridor credibility field corresponding to the candidate passage areas. The passageway reconstruction module reconstructs passageways for candidate passage areas based on the local passable passageway confidence field, generating passable passageways; The trajectory planning and control module generates the target trajectory based on the local passable corridor confidence field and the passable corridor, and adjusts the control continuity of the lateral and longitudinal control processes corresponding to the target trajectory based on the local passable corridor confidence field to generate vehicle control commands. The map feedback update module collects traffic feedback data generated by the vehicle as it travels within a local road area according to vehicle control commands, and updates the map database data for the corresponding local road area in the big data map database based on the traffic feedback data. The loop scheduling module uses the incomplete autonomous driving task as the loop execution condition. If the loop execution condition is met, it returns to the local data acquisition module and then to the map feedback update module based on the local road area corresponding to the updated vehicle location. If the loop execution condition is not met, the execution ends.

[0022] Compared with the prior art, the beneficial effects of the present invention are: 1. Based on further analysis and research of related technical issues, this invention recognizes that when there are local differences between map database data and real-time perception data, the lack of unified credibility constraints and feedback connections between traversable corridor reconstruction, target trajectory generation, vehicle lateral and longitudinal control adjustment, and map database data update processing can easily lead to insufficient coordination between vehicle traffic decisions and vehicle control processes in local road areas. Therefore, this invention obtains map database data of the local road area corresponding to the vehicle's current location in a large-scale map database, as well as the vehicle's real-time perception data of the local road area. It extracts road topology data and corridor verification data from the map database data, generates candidate traversable areas based on the road topology data, and then performs consistency analysis between the corridor verification data and real-time perception data and the candidate traversable areas to generate a local traversable corridor credibility field corresponding to the candidate traversable areas. The credibility field of the local passable corridor is further used to reconstruct the passable corridor in the candidate passable area to generate passable corridors. The same credibility field of the local passable corridor is used to participate in the generation of the target trajectory and the control continuity adjustment of the lateral and longitudinal control processes corresponding to the target trajectory to generate vehicle control commands. At the same time, the passable feedback data generated by the vehicle driving according to the vehicle control commands is collected, and the map library data of the corresponding local road area in the big data map library is updated based on the passable feedback data. Therefore, the credibility field of local passable corridors serves as both the basis for reconstructing passable corridors and the basis for generating target trajectories and adjusting the continuity of vehicle lateral and longitudinal control. It also forms a loop with the data update processing of traffic feedback data and map database data, enabling better coordination between passable corridors, target trajectories, vehicle control commands and subsequently called map database data, thereby improving the ability of vehicles to travel continuously and stably in local road areas.

[0023] 2. This invention generates traffic feedback data by collecting actual vehicle driving trajectories, real-time identification results, trajectory deviations, and control response information. After performing integrity, confidence, and spatial consistency checks on the traffic feedback data, it is used to identify map discrepancies and update the map database data for corresponding local road areas in the big data map database. When the big data map database uses a caching update mechanism, the update processing results are also synchronously written to an update cache associated with the big data map database, establishing a correlation between the update processing results and the map database data for the corresponding local road areas in the big data map database. This ensures that when subsequent step S1 obtains the map database data for the corresponding local road areas, it simultaneously calls the update processing results in the update cache. Through this processing method, the feedback data generated during vehicle traffic can participate in subsequent map database data calls, thus creating a closer update connection between the previous round of traffic results and the subsequent round of local road area decisions. This is beneficial for improving the adaptability of the map database data to changes in the actual state of local road areas during the cyclic execution process. Attached Figure Description

[0024] Figure 1 This is a schematic diagram of the core process of the overall method steps of the present invention; Figure 2 This is a schematic diagram of the core process of step S1 of the present invention; Figure 3 This is a schematic diagram of the core process of step S2 of the present invention; Figure 4 This is a schematic diagram of the core process of step S3 of the present invention; Figure 5 This is a schematic diagram of the core process of step S4 of the present invention; Figure 6 This is a schematic diagram of the core process of step S5 of the present invention; Figure 7 This is a schematic diagram of the core process of step S6 of the present invention; Figure 8 This is a schematic diagram of the core process of step S7 of the present invention; Figure 9 This is a schematic diagram of the overall module flow of the present invention.

[0025] In the diagram: 100, Local Data Acquisition Module; 200, Candidate Passage Area Generation Module; 300, Credibility Field Generation Module; 400, Passage Corridor Reconstruction Module; 500, Trajectory Planning and Control Module; 600, Map Feedback Update Module; 700, Loop Scheduling Module. Detailed Implementation

[0026] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0027] Next, please refer to Figure 1 One objective of this embodiment is to provide an autonomous driving method based on a large-scale map library. During the autonomous driving process, this method combines map library data from the large-scale map library with real-time perception data of the vehicle on local road areas to generate candidate traffic areas, a local passable corridor confidence field, passable corridors, target trajectories, and vehicle control commands for the local road area corresponding to the vehicle's current location. It also collects traffic feedback data while the vehicle is driving according to the vehicle control commands, updating the map library data for the corresponding local road area in the large-scale map library. This forms a closed-loop processing flow between map library data, real-time perception data, vehicle control commands, and traffic feedback data. The method includes steps S1, S2, S3, S4, S5, S6, and S7; wherein: S1. Obtain map data of the local road area corresponding to the current location of the vehicle in the big data map library, and obtain real-time perception data of the vehicle of the local road area; S2. Extract road topology data and corridor verification data from the map database, and generate candidate traffic areas based on the road topology data; S3. Perform consistency analysis between the corridor verification data and real-time perception data and the candidate passage areas to generate a local passable corridor confidence field corresponding to the candidate passage areas. S4. Based on the local passable corridor confidence field, reconstruct the passable corridors of the candidate passable areas to generate passable corridors; S5. Generate the target trajectory based on the local passable corridor confidence field and the passable corridor, and adjust the control continuity of the lateral control process and longitudinal control process corresponding to the target trajectory based on the local passable corridor confidence field to generate vehicle control commands. S6. Collect traffic feedback data generated by the vehicle during its travel in a local road area according to the vehicle control command, and update the map database data of the corresponding local road area in the big data map database based on the traffic feedback data. S7. If the autonomous driving task has not ended, the execution will be repeated. If the execution is repeated, the execution will return to steps S1 to S6 based on the local road area corresponding to the updated vehicle location. If the execution is not repeated, the execution will end.

[0028] The specific steps are as follows: Please see Figure 2 S1. Obtain map data of the local road area corresponding to the vehicle's current location in the big data map database, and obtain real-time perception data of the vehicle for the local road area, specifically including: Obtain the vehicle's current location information, which includes the vehicle's current location coordinates, heading angle, speed, and road sign information. The local road area corresponding to the vehicle's current location is determined based on the vehicle's current location information. The vehicle's current location coordinates are used as the starting position of the local road area, and the driving direction corresponding to the vehicle's heading angle is used as the extension direction of the local road area. The basic length of the local road area along the driving direction is determined by multiplying the vehicle's current speed by the preset planning time domain, and a preset safety redundancy distance is added to this basic length to obtain the longitudinal range of the local road area along the driving direction. The lateral range of the local road area is obtained by adding preset lateral safety redundancy distances on both sides of the road width corresponding to the vehicle's current location. The road space range jointly defined by the vehicle's current location coordinates, longitudinal range, lateral range, and vehicle heading angle is determined as the local road area. Send a data request corresponding to the local road area to the big data map library. The data request includes the vehicle's current location coordinates, vehicle heading angle, road sign information, and local road area range information. Receive map library data returned by the big data map library. The map library data includes data used to describe the road structure, lane connection relationship, road boundary, historical traffic status and road change status of local road areas; The vehicle collects environmental data within the sensing coverage area of ​​a local road area through sensing devices, including cameras, millimeter-wave radar, lidar, ultrasonic radar, and positioning sensors. Real-time perception data is obtained by performing target detection, road boundary recognition, obstacle recognition, lane line recognition, traffic sign recognition, and dynamic target state estimation on environmental data. Specifically, environmental data undergoes preprocessing, feature extraction, target association, and multi-source fusion processing. This includes distortion correction, noise reduction, and time synchronization processing on image data acquired by cameras, and timestamp alignment and coordinate unification processing on data output from millimeter-wave radar, lidar, ultrasonic radar, and positioning sensors. Based on the preprocessed environmental data, lane line features, road boundary features, obstacle features, dynamic target features, and traffic facility features are extracted, and the positional and temporal relationships of these features in a unified vehicle coordinate system are considered. The system performs correlation matching to determine the spatial locations of lane lines, road boundaries, obstacles, dynamic targets, and traffic facilities; then, it determines the obstacle-occupied area based on the distribution range of obstacles and road boundaries; it estimates the position, speed, and direction of motion of dynamic targets based on the position, speed, and direction of motion of dynamic targets at continuous time intervals; and it extracts the real-time sensing passage area based on the comprehensive distribution results of lane lines, road boundaries, obstacles, and traffic facilities, thus obtaining real-time sensing data including lane lines, road boundaries, obstacles, obstacle-occupied areas, dynamic targets, dynamic target positions, dynamic target speeds, dynamic target directions of motion, traffic facilities, and the real-time sensing passage area. Map database data and real-time perception data are uniformly converted to the vehicle coordinate system for coordinate alignment, enabling subsequent processing of map database data and real-time perception data under the same spatial reference.

[0029] Through the above-mentioned processes of acquiring the vehicle's current location information, determining the local road area, requesting and receiving map database data, collecting environmental data, generating real-time perception data, and aligning coordinates, map database data and real-time perception data matching the local road area corresponding to the vehicle's current location are obtained. This provides input data for subsequent road topology data extraction, corridor verification data extraction, candidate traffic area generation, and consistency analysis.

[0030] Please see Figure 3 S2. Extract road topology data and corridor verification data from the map database, and generate candidate traffic areas based on the road topology data, specifically including: Step S2 is used to extract road topology data describing the road connection structure of a local road area from the map library data, and corridor verification data for verifying the traffic status of the local road area. Based on the road topology data, candidate traffic areas that vehicles can choose are formed. Specifically, it includes a road topology data extraction step, a corridor verification data extraction step, and a candidate traffic area generation step.

[0031] The specific steps of extracting road topology data include: Read the lane centerline, lane boundary line, road boundary, lane width, lane direction, lane connection relationship, intersection connection relationship, turning relationship and road restriction relationship within a local road area from the map database data; Based on the vehicle's current position and heading angle, filter road topology elements that are consistent with the vehicle's current direction of travel, and remove reverse lanes, prohibited lanes, and unreachable road branches that are inconsistent with the vehicle's direction of travel. A local road topology graph is constructed based on lane connection relationships and intersection connection relationships. The nodes in the local road topology graph include lane segment nodes, intersection connection segment nodes, and road branch point nodes. The edges in the local road topology graph represent the drivable connection relationships between adjacent nodes. The local road topology map, lane geometric boundaries, and road connection relationships are used as road topology data.

[0032] Through the above-mentioned processes of reading map library data, filtering road topology elements, constructing local road topology maps, and organizing road connection relationships, road topology data is obtained to describe the road connection structure of local road areas.

[0033] The specific steps of extracting corridor verification data include: Data fields related to the traffic status of local road areas are read from the map database. These data fields include historical traffic trajectories, road construction records, temporary closure records, accident high-incidence records, road congestion status, vehicle group perception results, and road change records. Filter valid data fields based on the vehicle's current location, the local road area range, and the data timestamp, and remove data fields that exceed the local road area range and data fields that exceed the valid time range; The filtered historical traffic trajectories, road change records, and vehicle group perception results are mapped to the spatial units and road topology elements corresponding to the local road area; Corridor verification data is generated based on the mapping results. This data is used for consistency analysis with candidate passage areas and real-time sensing data in subsequent steps.

[0034] Through the above-mentioned processing of reading traffic status-related data fields, filtering valid data fields, mapping historical traffic trajectories, mapping road change records, and mapping vehicle group perception results, corridor verification data for verifying the traffic status of local road areas is obtained.

[0035] The specific steps in generating candidate access areas include: Based on the lane centerlines, lane boundary lines, and road connection relationships in the road topology data, determine the lane segments and road connection segments that the vehicle can reach from its current location; Boundary correction is performed on reachable lane segments and road connection segments; when there is redundant space available for vehicle passage between the inner side of the lane boundary line and the road boundary, the lane boundary line is shifted outward by a preset correction distance in a direction away from the lane center line, and the preset outward correction distance does not exceed the width of the redundant space; when the lane boundary line is adjacent to the road boundary, fixed obstacle boundary, and prohibited area boundary, the lane boundary line is shifted inward by a preset correction distance in a direction closer to the lane center line, and the preset inward correction distance is determined according to the vehicle width and the preset safety distance; The road space after boundary correction is matched with vehicle motion constraints, road direction constraints and lane connection constraints, and the road space that meets the conditions of vehicle passage width, minimum turning radius and continuous connectivity is retained as candidate passage area; When there are multiple drivable branches in the road topology data, candidate traffic areas corresponding to each drivable branch are generated, and the topological connection relationship between each candidate traffic area is preserved. When there are prohibited, closed, or inaccessible areas in the road topology data, these areas will be removed from the candidate traffic areas.

[0036] Through the above processes of determining accessible lane segments, determining road connection segments, boundary correction, matching vehicle motion constraints, matching road direction constraints, matching lane connection constraints, and eliminating prohibited areas, candidate traffic areas are obtained for subsequent consistency analysis.

[0037] Please see Figure 4 S3. Perform consistency analysis between the corridor verification data and real-time sensing data and the candidate passage areas to generate a local passable corridor confidence field corresponding to the candidate passage areas, specifically including: Step S3 is used to analyze the consistency between corridor verification data and candidate passage areas, as well as the consistency between real-time perception data and candidate passage areas, and to integrate the consistency analysis results into a local passable corridor confidence field corresponding to the candidate passage areas. Specifically, it includes a corridor verification consistency analysis step, a real-time perception consistency analysis step, and a local passable corridor confidence field generation step.

[0038] The specific steps of the corridor verification consistency analysis process include: The candidate passage area is divided into spatial units, so that the credibility of the candidate passage area can be calculated in discrete spatial units; The corridor verification data is mapped to the spatial units corresponding to the candidate passage areas to obtain the historical passage characteristics, road change characteristics and vehicle group perception characteristics of each spatial unit. For each spatial unit, the number of historical travel trajectories that overlap with that spatial unit is counted, and the total number of historical travel trajectories that pass through the neighborhood of that spatial unit is also counted. The proportion of the number of historical travel trajectories that overlap with that spatial unit to the total number of historical travel trajectories that pass through the neighborhood of that spatial unit is determined as the historical travel consistency value of that spatial unit. The more historical travel trajectories that overlap with that spatial unit, and the higher the proportion, the higher the historical travel consistency value of that spatial unit. For each spatial unit, determine whether the spatial unit overlaps with road construction records, temporary closure records, and road change records; when the spatial unit does not overlap with road construction records, temporary closure records, and road change records, the road change consistency value of the spatial unit is determined to be high; when the spatial unit overlaps with road construction records, temporary closure records, and road change records, the road change consistency value of the spatial unit is determined to be low. For each spatial unit, the number of results in the vehicle group perception results that indicate that the spatial unit is passable is counted, and the total number of results in the vehicle group perception results that cover the spatial unit is counted. The proportion of the number of results that indicate that the spatial unit is passable to the total number of results that cover the spatial unit is determined as the vehicle group perception consistency value of the spatial unit. Based on the preset historical traffic weight, preset road change weight, and preset vehicle group perception weight, the historical traffic consistency value, road change consistency value, and vehicle group perception consistency value are weighted and fused to obtain the corridor verification consistency result of each spatial unit; the sum of the preset historical traffic weight, preset road change weight, and preset vehicle group perception weight is a unified preset value.

[0039] Through the above-mentioned processes of dividing candidate traffic areas into spatial units, mapping corridor verification data, calculating historical traffic consistency values, determining road change consistency values, calculating vehicle group perception consistency values, and weighted fusion, the corridor verification consistency results between candidate traffic areas and corridor verification data are obtained.

[0040] The real-time consistency analysis process specifically includes: Map lane lines, road boundaries, obstacles, dynamic targets, traffic facilities, and real-time perceived traffic areas from real-time sensing data to the spatial units corresponding to candidate traffic areas; For each spatial unit, the distance deviation between the road boundary identified by real-time sensing data and the candidate traffic area boundary is calculated; the preset maximum allowable boundary deviation is compared with the distance deviation; the smaller the distance deviation, the higher the boundary consistency value of the spatial unit; when the distance deviation reaches the preset maximum allowable boundary deviation, the boundary consistency value of the spatial unit is determined as the minimum value. For each spatial unit, determine whether the spatial unit overlaps with an obstacle-occupied area; when the spatial unit overlaps with an obstacle-occupied area, the obstacle occupancy impact value of the spatial unit is determined to be high; when the spatial unit does not overlap with an obstacle-occupied area, the obstacle occupancy impact value of the spatial unit is determined to be low; the obstacle occupancy impact value is negatively correlated with the passability confidence of the spatial unit. For each spatial unit, the trajectory of the dynamic target within a preset planning time domain is predicted based on the dynamic target's position, velocity, and direction of motion. The duration of the dynamic target's occupation of the spatial unit within the preset planning time domain is also determined. The shorter the duration of the dynamic target's occupation of the spatial unit, the higher the dynamic risk consistency value of the spatial unit. The longer the duration of the dynamic target's occupation of the spatial unit, the lower the dynamic risk consistency value of the spatial unit. For each spatial unit, determine whether the spatial unit is located within the real-time sensing access area; when the spatial unit is located within the real-time sensing access area, determine the real-time sensing access area consistency value of the spatial unit as a high value; when the spatial unit is located outside the real-time sensing access area, determine the real-time sensing access area consistency value of the spatial unit as a low value. The consistency values ​​of the boundary, dynamic risk, and real-time perception access areas are fused according to preset boundary weights, preset dynamic risk weights, and preset real-time perception access area weights. The fusion result is then suppressed based on the obstacle occupancy impact value to obtain the real-time perception consistency result for each spatial unit. The sum of the preset boundary weights, preset dynamic risk weights, and preset real-time perception access area weights is a unified preset value. The higher the obstacle occupancy impact value, the greater the reduction in the real-time perception consistency result of the corresponding spatial unit.

[0041] Through the above-mentioned real-time perception data mapping, boundary consistency value calculation, obstacle occupancy impact value determination, dynamic risk consistency value calculation, real-time perception passage area consistency value determination, and suppression fusion processing, the real-time perception consistency result between the candidate passage area and the real-time perception data is obtained.

[0042] The specific steps in generating the credibility field of a locally accessible corridor include: Using spatial units within the candidate passage area as the computational object, the system receives corridor verification consistency results and real-time perception consistency results. Based on the timeliness, reliability of data source, and spatial coverage integrity of corridor verification data, the corridor verification weight corresponding to the corridor verification consistency result is determined; the higher the timeliness, reliability of data source, and spatial coverage integrity of corridor verification data, the higher the corridor verification weight. Based on the sensor confidence, target recognition confidence, occlusion status, and perception coverage of the real-time perception data, the real-time perception weight corresponding to the real-time perception consistency result is determined; the higher the sensor confidence, the higher the target recognition confidence, the lower the occlusion degree, and the more complete the perception coverage of the real-time perception data, the higher the real-time perception weight. For each spatial unit, the corridor verification consistency result and the real-time perception consistency result are weighted and fused according to the corridor verification weight and the real-time perception weight to obtain the passability confidence of the spatial unit; the sum of the weights of the corridor verification weight and the real-time perception weight is a unified preset value. Arrange the spatial locations and corresponding passability confidence scores of all spatial units in the candidate passability area according to spatial distribution to generate a local passability corridor confidence field corresponding to the candidate passability area. For each spatial unit, the difference between the corridor verification consistency result and the real-time perception consistency result is calculated. When the difference reaches a preset conflict threshold, the spatial unit is determined to have a data conflict. When the spatial unit has a data conflict and the sensor confidence of the real-time perception data is higher than the preset sensor confidence threshold, the real-time perception weight corresponding to the spatial unit is increased and the corridor verification weight corresponding to the spatial unit is decreased. When the spatial unit has a data conflict and the sensor confidence of the real-time perception data is lower than the preset sensor confidence threshold, the corridor verification weight corresponding to the spatial unit is increased and the real-time perception weight corresponding to the spatial unit is decreased. When real-time sensing data is obstructed, the spatial unit corresponding to the obstructed area is marked as pending confirmation; when real-time sensing data is missing, the spatial unit corresponding to the missing area is marked as pending confirmation; when the recognition confidence of real-time sensing data is lower than the preset confidence threshold, the corresponding spatial unit is marked as pending confirmation.

[0043] Through the aforementioned processes of corridor verification weight determination, real-time perception weight determination, passability confidence fusion, spatial location arrangement, data conflict determination, and pending confirmation status marking, a local passability corridor confidence field corresponding to the candidate passable area is generated. This local passability corridor confidence field is used for subsequent passability corridor generation, target trajectory generation, and control continuity adjustment of the lateral and longitudinal control processes.

[0044] Please see Figure 5 S4. Based on the local passable corridor confidence field, reconstruct the passable corridors of the candidate passable areas to generate passable corridors, specifically including: Receive the local passable corridor confidence field generated in step S3, and the candidate passable area generated in step S2; A first confidence threshold and a second confidence threshold are preset, with the first confidence threshold being higher than the second confidence threshold; Based on the passability confidence of each spatial unit in the local passable corridor confidence field, the candidate passable areas are divided into high-confidence passable areas, unconfirmed passable areas, and low-confidence passable areas. Spatial units whose passability confidence reaches the first confidence threshold are defined as high-confidence passable areas; Spatial units with a passability confidence level below the first confidence level threshold but reaching the second confidence level threshold are identified as areas to be confirmed for passage. Spatial units with a passability confidence level below the second confidence threshold are identified as low-confidence passability areas; Remove low-confidence traffic areas from the candidate traffic areas and retain high-confidence traffic areas that are connected to the vehicle's current location and target direction of travel; The safety boundary shrinkage process is performed on the area to be confirmed for passage. The boundary shrinkage distance of the area to be confirmed for passage is determined according to the braking distance corresponding to the current vehicle speed, the perception uncertainty corresponding to the real-time perception data, and the lateral control margin corresponding to the road curvature. The boundary shrinkage distance is translated inward along the boundary of the area to be confirmed for passage to obtain the area to be confirmed for passage after the safety boundary shrinkage process. A continuous passable space is generated based on the highly reliable passable area and the unconfirmed passable area after the security boundary shrinkage process. Perform boundary smoothing and topological connectivity checks on continuous passable spaces; remove local spaces that do not meet the requirements for minimum turning radius, vehicle width, and preset safety distance. When multiple consecutive passable spaces exist, the passability confidence of each spatial unit within each consecutive passable space is calculated, and the average passability confidence of each consecutive passable space is calculated. The directional deviation between each consecutive passable space and the target driving direction is determined, the connectivity of each consecutive passable space in the road topology data is determined, and the curvature change of the reference path corresponding to each consecutive passable space is calculated. Continuous passable spaces with high average passability confidence, small directional deviation, high road topology connectivity, and small curvature change are identified as passable corridors. When there is no continuous passable space that meets the requirements for normal passage, a restricted passable corridor is generated in the safe parking area near the current position of the vehicle, and the restricted passable corridor is output as a passable corridor to step S5, so that step S5 generates the target trajectory corresponding to deceleration and parking based on the restricted passable corridor.

[0045] Through the above-mentioned processes of grading the passability credibility, eliminating low-credibility passable areas, shrinking the safety boundary of unconfirmed passable areas, generating continuous passable spaces, smoothing the boundary, checking the topological connectivity, and sorting and selecting continuous passable spaces, the passable corridors of candidate passable areas are reconstructed, and passable corridors for subsequent target trajectory generation are generated.

[0046] Please see Figure 6 S5. Generate the target trajectory based on the local traversable corridor confidence field and the traversable corridor, and adjust the control continuity of the lateral and longitudinal control processes corresponding to the target trajectory based on the local traversable corridor confidence field to generate vehicle control commands, specifically including: Step S5 is used to generate a target trajectory based on the passable corridor generated in step S4 and the local passable corridor confidence field generated in step S3, and to adjust the control continuity of the lateral control process and the longitudinal control process corresponding to the target trajectory. Specifically, it includes a target trajectory generation step, a control continuity adjustment step for the lateral control process, a control continuity adjustment step for the longitudinal control process, and a vehicle control command generation step.

[0047] The specific steps in generating the target trajectory include: The system receives the passable corridor, the credibility field of the partially passable corridor, the current state of the vehicle, and the autonomous driving task objective. The current state of the vehicle includes the vehicle position, vehicle speed, vehicle acceleration, vehicle heading angle, and vehicle yaw rate. Multiple candidate trajectories are generated within the passable corridor, and the candidate trajectories satisfy vehicle kinematic constraints, road boundary constraints, obstacle avoidance constraints, and traffic rule constraints. For each candidate trajectory, the passability confidence level of the spatial unit traversed by the candidate trajectory is calculated, and the average passability confidence level, minimum passability confidence level, degree of change in passability confidence level, and proportion of neighboring areas of low-confidence passability areas are calculated for the regions traversed by the candidate trajectory, so as to obtain the trajectory confidence characteristics of each candidate trajectory. For each candidate trajectory, the trajectory length, curvature variation, target direction deviation, obstacle distance margin, and trajectory reliability characteristics are calculated. When the trajectory length is short, the curvature variation is small, the target direction deviation is small, the obstacle distance margin is large, the average passability reliability is high, the minimum passability reliability is high, the passability reliability variation is small, and the proportion of adjacent areas of low-reliability passability areas is low, the comprehensive cost value of the candidate trajectory is determined to be better. The comprehensive cost values ​​of multiple candidate trajectories are ranked according to a unified preset evaluation rule, and the candidate trajectory with the best comprehensive cost value and that meets the preset constraints is determined as the target trajectory. When the accessible corridor is a restricted accessible corridor, a target trajectory is generated with the safe parking area as the target location.

[0048] Through the above-mentioned processes of candidate trajectory generation, trajectory credibility feature calculation, trajectory length calculation, curvature change degree calculation, target direction deviation degree calculation, obstacle distance margin calculation, and comprehensive cost ranking, a target trajectory matching the credibility field of the passable corridor and the local passable corridor is obtained.

[0049] The specific steps of the control continuity adjustment process in the lateral control process include: Based on the curvature, lateral deviation, heading deviation, and local navigable corridor confidence field of the target trajectory, the range of target rotation angle change corresponding to the lateral control process is determined; When the target trajectory passes through a high-confidence passage area, normal lateral control parameters are generated according to the curvature of the target trajectory. When the target trajectory passes through the area to be confirmed, the steering wheel angle change rate is limited according to the preset first angle change rate threshold, and the lateral offset of the target trajectory is smoothed. When the target trajectory passes through an area where the rate of change of passability confidence reaches a preset rate of change threshold, the rate of change of steering wheel angle is limited according to the preset second rate of change threshold, and the lateral offset of the target trajectory is smoothed; the preset second rate of change threshold is lower than the preset first rate of change threshold. When the target trajectory is near the boundary of the low-confidence traffic zone, increase the lateral safety margin and adjust the lateral control target to keep the vehicle away from the low-confidence traffic zone. The lateral control parameters are generated based on the adjusted lateral control target. The lateral control parameters include the target steering wheel angle, the rate of change of steering wheel angle, and the lateral deviation correction.

[0050] By determining the target angle change range, generating normal lateral control parameters, limiting the steering wheel angle change rate, smoothing lateral offset, increasing lateral safety margin, and adjusting the lateral control target, the control continuity adjustment of the lateral control process corresponding to the target trajectory is completed.

[0051] The specific steps of the control continuity adjustment process in the longitudinal control process include: Based on the expected velocity of the target trajectory, the distance to the obstacle ahead, the road speed limit, the road gradient, and the confidence field of the local passable corridor, the target velocity and target acceleration corresponding to the longitudinal control process are determined. When the passability confidence along the target trajectory is stable and reaches the first confidence threshold, normal longitudinal control parameters are generated according to the expected speed corresponding to the target trajectory. When there are unconfirmed passage areas along the target trajectory, the current target speed is reduced according to the preset first speed adjustment coefficient, and the rate of change of acceleration and the rate of change of deceleration are limited; When there are low-confidence passage areas adjacent to the target trajectory, the current target speed is reduced according to the preset second speed adjustment coefficient, and the rate of change of acceleration and deceleration are limited; the preset second speed adjustment coefficient is lower than the preset first speed adjustment coefficient; When there are areas along the target trajectory where the passability confidence rapidly decreases, the current target speed is reduced according to the preset third speed adjustment coefficient, and the rate of change of acceleration and deceleration is limited; the preset third speed adjustment coefficient is lower than the preset first speed adjustment coefficient; When the local passable corridor confidence field indicates that there is a spatial unit ahead of the target trajectory with a passability confidence level lower than the second confidence threshold, a deceleration and parking control target is generated. Based on the adjusted longitudinal control target, longitudinal control parameters are generated, including target speed, target acceleration, braking intensity, and drive torque request.

[0052] Through the above-mentioned processes of target velocity determination, target acceleration determination, normal longitudinal control parameter generation, target velocity graded reduction, acceleration rate of change limitation, deceleration rate of change limitation, and deceleration and parking control target generation, the control continuity adjustment of the longitudinal control process corresponding to the target trajectory is completed.

[0053] The specific steps in generating vehicle control commands include: Receive target trajectory, lateral control parameters, and longitudinal control parameters; Convert lateral control parameters into steering execution commands; The longitudinal control parameters are converted into drive execution commands and braking execution commands; The steering, drive, and braking commands are synchronized in time to ensure that the lateral and longitudinal movements of the vehicle are executed in a coordinated manner within the same control cycle. When there is a conflict between the lateral control parameters and the longitudinal control parameters, the target speed should be reduced according to the principle of safety priority, while maintaining the executability of the lateral control process; Output vehicle control commands, including steering execution commands, drive execution commands, and braking execution commands.

[0054] Through the aforementioned processing of target trajectory reception, lateral control parameter reception, longitudinal control parameter reception, steering execution command conversion, drive execution command conversion, braking execution command conversion, and time synchronization, vehicle control commands are generated to control the vehicle to travel along the target trajectory.

[0055] Please see Figure 7 S6. Collect traffic feedback data generated by the vehicle during its movement within a local road area according to vehicle control commands, and update the map database data for the corresponding local road area in the big data map database based on the traffic feedback data. Specifically, this includes: Step S6 is used to collect traffic feedback data during the process of the vehicle executing vehicle control commands, and to use the traffic feedback data to update the map library data of the corresponding local road area in the big data map library. Specifically, it includes a traffic feedback data collection step, a traffic feedback data verification step, and a map library data update processing step.

[0056] The specific steps of the data collection process for traffic feedback include: During the process of the vehicle driving according to the vehicle control command, the actual driving trajectory, actual speed, actual acceleration, actual steering angle, actual braking status and actual driving status of the vehicle are collected; The system collects real-time identification results of vehicles in local road areas, including road boundaries, lane lines, obstacles, temporary closed areas, construction areas, and dynamic targets. The trajectory deviation between the target trajectory and the actual driving trajectory of the vehicle is collected. The trajectory deviation includes lateral deviation, longitudinal deviation, heading deviation and speed deviation. The system collects control response information during the execution of vehicle control commands. This control response information includes steering response delay, braking response delay, drive response delay, and control stability information. The actual driving trajectory, real-time recognition results, trajectory deviation, and control response information are timestamped and coordinates are unified to form traffic feedback data.

[0057] Through the above-mentioned processing of collecting actual vehicle driving status, real-time road condition recognition results, trajectory deviation, control response information, timestamp alignment, and coordinate unification, traffic feedback data reflecting the actual traffic status of vehicles in a local road area is obtained.

[0058] The specific steps of the pass feedback data verification process include: Perform integrity verification on the traffic feedback data to determine whether the traffic feedback data contains location, time, road conditions, and vehicle response information; The confidence level of the passage feedback data is verified to determine whether the corresponding perception results meet the preset confidence level requirements. Perform spatial consistency verification on the traffic feedback data to determine whether the location corresponding to the traffic feedback data is located within the local road area; When the traffic feedback data is complete, the corresponding perception result meets the preset reliability requirements, and the corresponding location is within a local road area, the traffic feedback data will be marked as valid traffic feedback data. When traffic feedback data has partial missing parts but can still reflect the road change status, and the corresponding location is within a local road area, the traffic feedback data is marked as traffic feedback data to be merged. When the traffic feedback data does not meet the preset reliability requirements and cannot reflect the effective road conditions, the corresponding traffic feedback data will be stopped from being used for map database data update processing.

[0059] Through the above-mentioned processes of integrity verification, confidence verification, spatial consistency verification, marking of valid access feedback data, marking of access feedback data to be merged, and exclusion of invalid access feedback data, valid access feedback data and access feedback data to be merged are obtained for map library data update processing.

[0060] The specific steps of updating map database data include: Based on valid traffic feedback data, identify road conditions in local road areas that are inconsistent with map database data to obtain map difference information; Map discrepancies include road boundary offsets, missing lane lines, temporary obstructions, construction zones, changes in road closures, changes in passable areas, and changes in historical traffic patterns. The map discrepancy information is matched with the map database data of the corresponding local road area in the big data map database to determine the data fields that need to be updated. When map discrepancy information is generated from valid single-vehicle passage feedback data and the corresponding confidence level is higher than the preset high confidence level threshold, the corresponding data field is temporarily updated, and the data source and update time are marked. When map discrepancy information is jointly verified by multi-vehicle traffic feedback data, the corresponding data fields are merged and updated, and the credibility level of the updated data fields is improved. When map discrepancy information conflicts with existing map database data and the corresponding confidence level does not reach the preset high confidence threshold, the corresponding data field is marked as pending confirmation, and the existing map database data remains unchanged. The update processing result is written to the map database data of the corresponding local road area in the big data map database. When the big data map database adopts a cache update mechanism, the update processing result is synchronously written to the update cache area associated with the big data map database, and the association between the update processing result and the map database data of the corresponding local road area in the big data map database is established, so that when the subsequent step S1 obtains the map database data of the corresponding local road area, the update processing result in the update cache area is called at the same time.

[0061] Through the above-mentioned processes of map difference information identification, map difference information matching, determination of updated processing data fields, temporary update processing of single-vehicle valid passage feedback data, fusion update processing of multi-vehicle passage feedback data, pending confirmation status marking, writing of update processing results, and associated calls to the update cache area, the update processing of the passage feedback data on the map library data of the corresponding local road area in the big data map library is completed, and the updated map library data of the corresponding local road area in the big data map library is obtained.

[0062] Please see Figure 8 S7. Using the condition that the autonomous driving task has not ended as the loop execution condition, if the loop execution condition is met, return to steps S1 to S6 based on the local road area corresponding to the updated vehicle location; if the loop execution condition is not met, terminate the execution, specifically including: Acquire the autonomous driving task status, which includes task progress status, task completion status, manual takeover status, system exit status, and malfunction stop status. Determine whether the loop execution condition that the autonomous driving task has not ended is met; When the autonomous driving task is in the task in progress state, the loop execution condition of determining that the autonomous driving task has not ended is met. If the loop execution condition of the autonomous driving task not ending is met, obtain the updated location of the vehicle, and redetermine the corresponding local road area based on the updated location of the vehicle. Based on the redefined local road area, return to steps S1 to S6, so that the updated map database data, the reacquired real-time perception data, and the updated vehicle position can jointly participate in the next round of candidate traffic area generation, local trafficable corridor confidence field generation, trafficable corridor generation, target trajectory generation, vehicle control command generation, and map database data update processing. When the autonomous driving task is in the task completion state, the loop execution condition of determining that the autonomous driving task has not ended is not met. When the autonomous driving task is in a manual takeover state, the loop execution condition that determines the autonomous driving task has not ended is not met. When the autonomous driving task is in the system exit state, the loop execution condition that determines that the autonomous driving task has not ended is not met. When the autonomous driving task is in a fault-prone parking state, the loop execution condition that determines the autonomous driving task has not ended is not met. If the condition that the autonomous driving task has not ended is not met, stop the execution of the loop from step S1 to step S6, and save the map library data update processing results, traffic feedback data, vehicle control commands and task status data generated in the current loop. When the autonomous driving task is in the task completion state, record the information that the vehicle has reached the target location and end the execution; When the autonomous driving task is in a state of manual takeover, stop automatically generating vehicle control commands, switch vehicle control to manual driving mode, and then end the execution. When the autonomous driving task is in the system exit state, stop the data processing flow from step S1 to step S6, save the state data before the system exit, and then end the execution. When the autonomous driving task is in a fault-locked state, execute safe parking control, save the traffic feedback data and vehicle status data before and after the fault-locked stop, and then end the execution.

[0063] Through the above-mentioned processes of acquiring the autonomous driving task status, judging the loop execution conditions, acquiring the updated vehicle location, redefining the local road area, returning to execution in steps S1 to S6, stopping the loop execution, and saving the task status data, the loop execution judgment of the autonomous driving task is completed. A closed-loop rolling processing flow based on the updated vehicle location is formed before the autonomous driving task ends. When the loop execution condition of the autonomous driving task not ending is not met, the process ends and the status is saved.

[0064] In this embodiment, the thresholds include a preset time boundary corresponding to the planning time domain, a preset maximum allowable boundary deviation, a preset conflict threshold, a preset sensor confidence threshold, a preset confidence threshold, a first confidence threshold, a second confidence threshold, a preset rate of change threshold, a preset first angle rate of change threshold, a preset second angle rate of change threshold, and a preset high confidence threshold; all of the above thresholds are taken within the effective range of their corresponding physical quantities, confidence measures, or normalized evaluation quantities; wherein: The first confidence threshold and the second confidence threshold are used to divide the spatial units in the local passable corridor confidence field into high confidence passable areas, unconfirmed passable areas and low confidence passable areas. Both are within the valid value range of passable confidence, and the first confidence threshold is higher than the second confidence threshold. The preset sensor confidence threshold, preset confidence threshold, and preset high confidence threshold are all within the valid range of perception confidence, and are used to determine the credibility of real-time perception data, traffic feedback data, and map difference information. The preset conflict threshold is within the effective range of the difference between the corridor verification consistency result and the real-time perception consistency result, and is used to determine whether there is a data conflict between the corridor verification data and the real-time perception data. The preset maximum allowable boundary deviation is determined based on vehicle width, lane width, positioning error, sensor ranging error, and road boundary recognition error; The preset rate of change threshold is determined based on the smoothness requirement of spatial change along the target trajectory for passability credibility; both the preset first angle rate of change threshold and the preset second angle rate of change threshold are within the range of steering wheel angle change rate allowed by the vehicle steering actuator, and the preset second angle rate of change threshold is lower than the preset first angle rate of change threshold; The determination of the above thresholds is based on factors including vehicle structural parameters, vehicle dynamics performance, braking capacity, steering execution capacity, sensor accuracy, positioning accuracy, road grade, road curvature, lane width, autonomous driving safety margin, historical operation statistics, and calibration verification results.

[0065] The coefficients in this embodiment include a preset first speed adjustment coefficient, a preset second speed adjustment coefficient, and a preset third speed adjustment coefficient, and involve preset historical traffic weights, preset road change weights, preset vehicle group perception weights, preset boundary weights, preset dynamic risk weights, preset real-time perception traffic area weights, corridor verification weights, and real-time perception weights; wherein: The preset first speed adjustment coefficient, the preset second speed adjustment coefficient, and the preset third speed adjustment coefficient are all greater than zero and less than one, and are used to reduce the current target speed. Among them, the preset second speed adjustment coefficient is lower than the preset first speed adjustment coefficient, and the preset third speed adjustment coefficient is lower than the preset first speed adjustment coefficient. When there are low confidence passable areas or areas where the passable confidence decreases rapidly along the target trajectory, the longitudinal control process can generate a more conservative target speed. All weights are within the valid range of normalized weights, and the sum of weights participating in the same weighted fusion process is a unified preset value; the preset historical traffic weight, preset road change weight, and preset vehicle group perception weight are determined based on the reliability of the data source, the timeliness of the data, and the spatial coverage integrity of the corridor verification data; The preset boundary weight, preset dynamic risk weight, and preset real-time perceived traffic area weight are determined based on the confidence level of road boundary identification, the reliability of dynamic target prediction, and the confidence level of real-time perceived traffic area identification in the real-time perceived data. The corridor verification weight and real-time perception weight are determined based on the reliability of the corridor verification data and real-time perception data within the corresponding spatial unit, and are adjusted according to the sensor confidence level of the real-time perception data when there is a data conflict.

[0066] The determination of the above coefficients and weights is based on factors including vehicle driving safety margin, target trajectory smoothness requirements, sensor stability, road environment complexity, historical traffic statistics, real vehicle test results, and simulation calibration results.

[0067] To further illustrate the working process of this embodiment, a specific operation process is given below: During the execution of the autonomous driving task, step S1 obtains the vehicle's current position coordinates, vehicle heading angle, vehicle speed information, and road sign information. Based on the vehicle's current position coordinates, vehicle heading angle, longitudinal range, and lateral range, a local road area is determined. The system requests and receives map data for the local road area from the big data map library. At the same time, environmental data is collected through cameras, millimeter-wave radar, lidar, ultrasonic radar, and positioning sensors to generate real-time perception data. The map library data and real-time perception data are then uniformly converted to the vehicle coordinate system. Step S2 reads lane center lines, lane boundary lines, road boundaries, lane connection relationships, intersection connection relationships, and road restriction relationships from the map database to construct a local road topology map and obtain road topology data. It also reads historical traffic trajectories, road construction records, temporary closure records, vehicle group perception results, and road change records to obtain corridor verification data. Based on the road topology data, it determines the lane segments and road connection segments that vehicles can reach, corrects their boundaries, and removes prohibited areas, closed areas, and inaccessible areas to generate candidate traffic areas. Step S3 divides the candidate passage area into spatial units, maps the corridor verification data and real-time perception data to each spatial unit, obtains the corridor verification consistency result and the real-time perception consistency result respectively, and performs weighted fusion according to the corridor verification weight and the real-time perception weight to generate a local passable corridor confidence field corresponding to the candidate passage area. Step S4 divides the candidate access areas into high-confidence access areas, unconfirmed access areas, and low-confidence access areas according to the first confidence threshold and the second confidence threshold. Low-confidence access areas are eliminated, and the unconfirmed access areas are subjected to security boundary shrinkage processing. The resulting continuous accessible space is then subjected to boundary smoothing processing, topological connectivity checks, and sorting selection to generate accessible corridors. Step S5 generates multiple candidate trajectories within the passable corridor, determines the target trajectory based on trajectory confidence characteristics, trajectory length, curvature change degree, target direction deviation degree and obstacle distance margin, and adjusts the control continuity of the lateral control process and longitudinal control process based on the local passable corridor confidence field, generating vehicle control commands including steering execution commands, drive execution commands and braking execution commands. Step S6 involves collecting the vehicle's actual driving trajectory, real-time identification results, trajectory deviation, and control response information during the vehicle's operation according to vehicle control commands, forming traffic feedback data, performing integrity verification, confidence verification, and spatial consistency verification on the traffic feedback data, identifying map difference information based on valid traffic feedback data, and performing temporary update processing, fusion update processing, and pending confirmation status marking on the map library data of the corresponding local road area in the big data map library. Step S7 obtains the autonomous driving task status. When the autonomous driving task is in the task in progress state, it is determined that the loop execution condition of the autonomous driving task not being completed is met, and steps S1 to S6 are returned to be executed based on the local road area corresponding to the updated vehicle location. When the autonomous driving task is in the task completed state, manual takeover state, system exit state, or fault parking state, it is determined that the loop execution condition of the autonomous driving task not being completed is not met, and the loop execution of steps S1 to S6 is stopped and the execution is terminated.

[0068] As can be seen from the above description, the autonomous driving method based on a big data map database provided in this embodiment has the following technical effects: This embodiment provides an autonomous driving method based on a big data map library. It acquires map library data of the local road area corresponding to the vehicle's current location in the big data map library, along with the vehicle's real-time perception data of the local road area. Road topology data and corridor verification data are extracted to generate candidate traversable areas. Consistency analysis is performed between the corridor verification data and the real-time perception data and the candidate traversable areas to generate a local traversable corridor confidence field. Furthermore, based on the local traversable corridor confidence field, traversable corridor reconstruction, target trajectory generation, and control continuity adjustment of the lateral and longitudinal control processes corresponding to the target trajectory are completed. The map library data of the corresponding local road area in the big data map library is updated using traffic feedback data generated during vehicle operation. This improves the synergy between traversable corridor reconstruction, target trajectory generation, vehicle lateral and longitudinal control adjustment, and map library data update processing when there are local differences between the map library data and real-time perception data. It also enhances the vehicle's ability to continuously and stably traverse local road areas.

[0069] Please see Figure 9 The second objective of this embodiment is to provide a system for an autonomous driving method based on a large-scale map database, including a local data acquisition module 100, a candidate traffic area generation module 200, a confidence field generation module 300, a traffic corridor reconstruction module 400, a trajectory planning and control module 500, a map feedback update module 600, and a cyclic scheduling module 700, wherein: The local data acquisition module 100 acquires the map library data of the local road area corresponding to the current location of the vehicle in the big data map library, and acquires the real-time perception data of the vehicle of the local road area. The candidate traffic area generation module 200 extracts road topology data and corridor verification data from the map library data, and generates candidate traffic areas based on the road topology data; The credibility field generation module 300 performs consistency analysis between the corridor verification data and real-time perception data and the candidate passage area, respectively, and generates a local passable corridor credibility field corresponding to the candidate passage area. The passageway reconstruction module 400 reconstructs the passageway in the candidate passage area based on the local passable passageway confidence field, and generates passable passageways; The trajectory planning and control module 500 generates the target trajectory based on the local passable corridor confidence field and the passable corridor, and adjusts the control continuity of the lateral control process and longitudinal control process corresponding to the target trajectory based on the local passable corridor confidence field, and generates vehicle control commands. The map feedback update module 600 collects traffic feedback data generated by the vehicle as it travels within a local road area according to the vehicle control command, and updates the map database data of the corresponding local road area in the big data map database based on the traffic feedback data. The loop scheduling module 700 uses the fact that the autonomous driving task has not ended as the loop execution condition. If the loop execution condition is met, it returns to the local data acquisition module 100 to the map feedback update module 600 based on the local road area corresponding to the updated vehicle location. If the loop execution condition is not met, the execution ends.

[0070] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred embodiments of the present invention and are not intended to limit the invention. Various changes and modifications can be made to the present invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. An autonomous driving method based on a large-scale map database, characterized in that, The methods and steps include the following: S1. Obtain map data of the local road area corresponding to the current location of the vehicle in the big data map library, and obtain real-time perception data of the vehicle on the local road area; S2. Extract road topology data and corridor verification data from the map database data, and generate candidate traffic areas based on the road topology data; S3. Perform consistency analysis between the corridor verification data and the real-time perception data and the candidate passage area respectively to generate a local passable corridor confidence field corresponding to the candidate passage area; S4. Based on the local passable corridor confidence field, reconstruct the passable corridor in the candidate passable area to generate a passable corridor; S5. Generate a target trajectory based on the local passable corridor confidence field and the passable corridor, and adjust the control continuity of the lateral control process and longitudinal control process corresponding to the target trajectory based on the local passable corridor confidence field to generate vehicle control commands. S6. Collect traffic feedback data generated during the vehicle's travel in the local road area according to the vehicle control command, and update the map database data corresponding to the local road area in the big data map database based on the traffic feedback data; S7. Using the condition that the autonomous driving task has not ended as the condition for cyclic execution, if the condition is met, return to steps S1 to S6 based on the local road area corresponding to the updated vehicle location; if the condition is not met, end the execution.

2. The autonomous driving method based on a big data map database according to claim 1, characterized in that, The acquisition process in step S1 specifically includes: Obtain the vehicle's current location information, which includes the vehicle's current position coordinates, heading angle, speed information, and road sign information. The local road area corresponding to the vehicle's current location is determined based on the vehicle's current location information; Send a data request corresponding to the local road area to the big data map library, and receive map library data returned by the big data map library; Environmental data within the sensing coverage area of ​​the local road region is collected by the sensing devices mounted on the vehicle; The environmental data is preprocessed, feature extracted, target associated, and multi-source fusion processed to obtain real-time perception data including lane lines, road boundaries, obstacles, obstacle-occupied areas, dynamic targets, dynamic target positions, dynamic target speeds, dynamic target movement directions, traffic facilities, and real-time perception passage areas. The map database data and the real-time perception data are uniformly converted to the vehicle coordinate system for coordinate alignment.

3. The autonomous driving method based on a big data map database according to claim 2, characterized in that, The extraction process in step S2 specifically includes: Read the lane centerline, lane boundary line, road boundary, lane width, lane direction, lane connection relationship, intersection connection relationship, turning relationship and road restriction relationship within the local road area from the map database data; Filter road topology elements that are consistent with the vehicle's current direction of travel based on the vehicle's current position and heading angle; A local road topology map is constructed based on the lane connection relationships and the intersection connection relationships, and the local road topology map, lane geometric boundaries, and road connection relationships are used as the road topology data. Read data fields related to the traffic status of the local road area from the map database, and filter valid data fields based on the vehicle's current location, the range of the local road area, and the data timestamp; The filtered historical traffic trajectories, road change records, and vehicle group perception results are mapped to the spatial units and road topology elements corresponding to the local road area to form the corridor verification data.

4. The autonomous driving method based on a big data map database according to claim 3, characterized in that, The process of generating the candidate passage area specifically includes: Based on the lane centerlines, lane boundary lines, and road connection relationships in the road topology data, determine the lane segments and road connection segments that the vehicle can reach from its current location. Boundary correction processing is performed on the reachable lane sections and road connection sections; The road space after boundary correction is matched with vehicle motion constraints, road direction constraints and lane connection constraints, and the road space that meets the conditions of vehicle passage width, minimum turning radius and continuous connectivity is retained as the candidate passage area; When there are multiple drivable branches in the road topology data, candidate traffic areas corresponding to each drivable branch are generated respectively, and the topological connection relationship between each candidate traffic area is preserved. When the road topology data contains prohibited areas, closed areas, and inaccessible areas, the prohibited areas, closed areas, and inaccessible areas are removed from the candidate traffic areas.

5. The autonomous driving method based on a big data map database according to claim 1, characterized in that, The generation process of the local passable corridor confidence field specifically includes: The candidate traffic area is divided into spatial units, and the corridor verification data is mapped to the spatial units corresponding to the candidate traffic area to obtain the historical traffic characteristics, road change characteristics and vehicle group perception characteristics corresponding to each spatial unit. For each spatial unit, the historical traffic consistency value, road change consistency value, and vehicle group perception consistency value are calculated, and the historical traffic consistency value, the road change consistency value, and the vehicle group perception consistency value are weighted and fused to obtain the corridor verification consistency result. The lane lines, road boundaries, obstacles, dynamic targets, traffic facilities, and real-time perceived traffic areas in the real-time perception data are mapped to the spatial units corresponding to the candidate traffic areas. For each spatial unit, the boundary consistency value, obstacle occupancy impact value, dynamic risk consistency value, and real-time perceived passage area consistency value are calculated. The boundary consistency value, the dynamic risk consistency value, and the real-time perceived passage area consistency value are then fused. Suppression processing is performed based on the obstacle occupancy impact value to obtain the real-time perception consistency result. The corridor verification weight is determined based on the timeliness, data source reliability, and spatial coverage integrity of the corridor verification data, and the real-time perception weight is determined based on the sensor confidence, target recognition confidence, occlusion status, and perception coverage of the real-time perception data. For each spatial unit, the corridor verification consistency result and the real-time perception consistency result are weighted and fused according to the corridor verification weight and the real-time perception weight to obtain the passability confidence level. Arrange the spatial locations of all spatial units and their corresponding passability credibility according to spatial distribution to generate the local passability corridor credibility field.

6. The autonomous driving method based on a big data map database according to claim 5, characterized in that, The process of generating the passable corridor specifically includes: Receive the local passable corridor confidence field and the candidate passable area; Based on the passability confidence of each spatial unit in the local passable corridor confidence field, the candidate passable area is divided into a high-confidence passable area, a passable area to be confirmed, and a low-confidence passable area. The low-confidence traffic areas are removed from the candidate traffic areas, and the high-confidence traffic areas that are connected to the vehicle's current position and the target driving direction are retained; The safety boundary of the unconfirmed passage area is contracted. Based on the highly reliable passage area and the passage area to be confirmed after security boundary shrinkage, a continuous passable space is generated; The continuous passable space is subjected to boundary smoothing and topological connectivity checks, and local spaces that do not meet the requirements of minimum turning radius, vehicle width and preset safety distance are removed. When multiple consecutive passable spaces exist, the passable corridor is determined by selecting the consecutive passable spaces that meet the requirements based on the average passability confidence of each consecutive passable space, the directional deviation from the target driving direction, the connectivity in the road topology data, and the curvature change of the corresponding reference path. When there is no continuous passable space that meets the requirements for normal passage, a restricted passable corridor is generated in the safe parking area adjacent to the vehicle's current location, and the restricted passable corridor is output as the passable corridor.

7. The autonomous driving method based on a big data map database according to claim 6, characterized in that, The process of generating the target trajectory specifically includes: Receive the accessible corridor, the credibility field of the locally accessible corridor, the current vehicle status, and the autonomous driving task objective; Multiple candidate trajectories are generated within the passable corridor, and the candidate trajectories satisfy vehicle kinematic constraints, road boundary constraints, obstacle avoidance constraints, and traffic rule constraints. For each candidate trajectory, the passability confidence level of the spatial unit traversed by the candidate trajectory is statistically analyzed, and the average passability confidence level, the minimum passability confidence level, the degree of change in passability confidence level, and the proportion of adjacent areas of low-confidence passable areas are calculated to obtain the trajectory confidence level characteristics. For each candidate trajectory, the trajectory length, curvature change, target direction deviation, obstacle distance margin, and trajectory reliability characteristics are calculated, and the comprehensive value of multiple candidate trajectories is ranked according to a unified preset evaluation rule. The candidate trajectory with the best comprehensive cost value and that meets the preset constraints is determined as the target trajectory; When the accessible corridor is a restricted accessible corridor, a target trajectory is generated with the safe parking area as the target location.

8. The autonomous driving method based on a big data map database according to claim 6, characterized in that, The process of generating the vehicle control commands specifically includes: Based on the curvature, lateral deviation, heading deviation of the target trajectory and the confidence field of the local passable corridor, the range of target angle change corresponding to the lateral control process is determined; When the target trajectory passes through a high-confidence passage area, normal lateral control parameters are generated according to the curvature of the target trajectory. When the target trajectory passes through the area to be confirmed, the rate of change of the steering wheel angle is limited, and the lateral offset of the target trajectory is smoothed. When the target trajectory passes through an area where the rate of change of passability confidence reaches a preset rate of change threshold, the rate of change of steering wheel angle is limited, and the lateral offset of the target trajectory is smoothed. When the target trajectory is near the boundary of a low-confidence passage area, the lateral safety margin is increased, the lateral control target is adjusted, and lateral control parameters are generated. Based on the expected speed of the target trajectory, the distance to the obstacle ahead, the road speed limit, the road gradient, and the confidence field of the local passable corridor, the target speed and target acceleration corresponding to the longitudinal control process are determined. When the passability confidence along the target trajectory is stable and reaches the first confidence threshold, normal longitudinal control parameters are generated. When there are areas along the target trajectory that need to be confirmed for passage, areas adjacent to low-confidence passage areas, or areas where the credibility of passage decreases rapidly, the current target speed is reduced, and the rate of change of acceleration and the rate of change of deceleration are limited. When the local passable corridor confidence field indicates that there is a spatial unit in front of the target trajectory with a passability confidence level lower than the second confidence threshold, a deceleration and parking control target is generated, and longitudinal control parameters are obtained. Receive the target trajectory, the lateral control parameters, and the longitudinal control parameters; convert the lateral control parameters into steering execution commands; and convert the longitudinal control parameters into drive execution commands and braking execution commands. The steering execution command, the drive execution command, and the braking execution command are time-synchronized. When there is a conflict between the lateral control parameters and the longitudinal control parameters, the target speed is reduced according to the principle of safety priority, while maintaining the executability of the lateral control process; The vehicle control commands are output, including the steering execution command, the drive execution command, and the braking execution command.

9. The autonomous driving method based on a big data map database according to claim 1, characterized in that, The update process in step S6 specifically includes: During the process of the vehicle driving according to the vehicle control command, the actual driving trajectory, actual speed, actual acceleration, actual steering angle, actual braking status and actual driving status of the vehicle are collected; The system collects real-time identification results of vehicles within the local road area, including road boundaries, lane lines, obstacles, temporary closures, construction areas, and dynamic targets. The trajectory deviation between the target trajectory and the actual driving trajectory of the vehicle is collected, and the control response information during the execution of the vehicle control command is also collected; The actual driving trajectory, real-time recognition results, trajectory deviation and control response information are timestamped and coordinates are unified to form traffic feedback data. The traffic feedback data is subjected to integrity verification, confidence verification, and spatial consistency verification. Traffic feedback data that meets the requirements is marked as valid traffic feedback data, and traffic feedback data that has partial missing parts but can still reflect the road change status is marked as traffic feedback data to be merged. Based on the valid traffic feedback data, identify the road conditions in a local road area that are inconsistent with the map database data, and obtain map difference information; The map difference information is matched with the map database data corresponding to the local road area in the big data map database to determine the data fields that need to be updated. When the map difference information is formed by the effective passage feedback data of a single vehicle and the corresponding confidence level is higher than the preset high confidence level threshold, the corresponding data field is temporarily updated and the data source and update time are marked. When the map difference information is jointly verified by multi-vehicle traffic feedback data, the corresponding data fields are fused and updated, and the credibility level of the updated data fields is improved. When the map difference information conflicts with existing map library data and the corresponding confidence level does not reach the preset high confidence threshold, the corresponding data field is marked as pending confirmation, and the existing map library data remains unchanged. The update processing result is written into the map database data corresponding to the local road area in the big data map database; When the big data map library adopts a caching update mechanism, the update processing result is synchronously written to the update cache area associated with the big data map library, and an association is established between the update processing result and the map library data corresponding to the local road area in the big data map library, so that when the subsequent step S1 obtains the map library data corresponding to the local road area, the update processing result in the update cache area is called at the same time.

10. A system using an autonomous driving method based on a large data map database as described in any one of claims 1-9, characterized in that, It includes a local data acquisition module, a candidate passage area generation module, a confidence field generation module, a passage corridor reconstruction module, a trajectory planning and control module, a map feedback update module, and a cyclic scheduling module, among which: The local data acquisition module acquires map library data of the local road area corresponding to the vehicle's current location in the big data map library, and acquires real-time perception data of the vehicle on the local road area. The candidate traffic area generation module extracts road topology data and corridor verification data from the map database data, and generates candidate traffic areas based on the road topology data; The credibility field generation module performs consistency analysis between the corridor verification data and the real-time perception data and the candidate passage area, respectively, to generate a local passable corridor credibility field corresponding to the candidate passage area. The passageway reconstruction module reconstructs passageways for the candidate passage areas based on the local passable passageway confidence field, thereby generating passable passageways. The trajectory planning and control module generates a target trajectory based on the local passable corridor confidence field and the passable corridor, and adjusts the control continuity of the lateral control process and longitudinal control process corresponding to the target trajectory based on the local passable corridor confidence field to generate vehicle control commands. The map feedback update module collects traffic feedback data generated by the vehicle during its travel in the local road area according to the vehicle control command, and updates the map database data corresponding to the local road area in the big data map database based on the traffic feedback data. The cyclic scheduling module uses the incomplete autonomous driving task as the condition for cyclic execution. When the cyclic execution condition is met, it returns to the local data acquisition module and then to the map feedback update module based on the local road area corresponding to the updated vehicle location. When the cyclic execution condition is not met, the execution ends.

Images