Trajectory correction method, medium, device and vehicle
By selecting zero-point frames in autonomous vehicles and using the Ackerman steering model and filtering algorithm to correct the trajectory, the problems of high cost of high-precision maps and low accuracy of maps with heavy perception are solved, achieving low-cost and efficient trajectory recording and improved positioning accuracy.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- WUHAN IDRIVERPLUS TECH CO LTD
- Filing Date
- 2022-09-20
- Publication Date
- 2026-06-05
AI Technical Summary
In existing autonomous driving technologies, relying on high-precision maps is costly and slow to update, while the approach of emphasizing perception over maps has low positioning accuracy and is prone to positioning jumps, resulting in insufficient precision in the positions of obstacles and the vehicle itself.
In the perception-heavy, map-light approach, a theoretical position is generated by selecting a zero-point frame, using the Ackerman steering model and filtering algorithm, and then trajectory correction is performed by combining the vehicle's status and map perception information to establish the accurate trajectory of the vehicle and obstacles.
It enables efficient and accurate recording of autonomous vehicle trajectory information at low cost, improves positioning accuracy, reduces positioning jumps, and is suitable for scenarios lacking high-precision maps.
Smart Images
Figure CN117782083B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of data processing technology, and in particular to a trajectory correction method, medium, device, and vehicle. Background Technology
[0002] With the widespread application of autonomous driving technology, the development process of autonomous driving algorithms requires the collection of information on various obstacles encountered by road users, which is then used as data to train various models and simulate testing systems. The accuracy of this data is crucial for the development and testing of autonomous driving algorithms.
[0003] Currently, there are two main methods for generating trajectories for autonomous vehicles and obstacles. The first method relies on high-precision maps. In this method, the autonomous vehicle can achieve centimeter-level positioning accuracy using high-precision maps and can fully utilize its location to obtain detailed map element information, such as lanes and pedestrian crossings. The position information of each obstacle in each frame can be precisely converted from its relative position to the autonomous vehicle's absolute position in the map coordinate system for recording. When using the data, the required absolute position of the obstacle is converted to the necessary local coordinate system for calculation, and the corresponding map element information can be retrieved from the high-precision map based on its position. The second method emphasizes perception over mapping. In this method, the autonomous vehicle uses information from the Global Positioning System (GPS) and inertial sensors for coarse positioning. Simultaneously, it uses surrounding map information perceived by the perception module and navigation-level maps for matching to correct the autonomous vehicle's positioning information. Furthermore, based on the corrected positioning information, the obstacle's position is converted to absolute map coordinates and stored. In the absence of detailed map element information provided by high-precision maps, the map element information provided by the perception module in each frame needs to be saved simultaneously.
[0004] Regarding the first method, which relies on high-precision maps, the production cost of high-precision maps is high, the data volume is large, and the update of map elements is slow, requiring the map to be built again before updates can be made. Furthermore, positioning is inaccurate in scenarios lacking significant landmarks, such as tunnels and long straight roads. Regarding the second method, which emphasizes perception over maps, the positioning accuracy using GPS and inertial sensors can only reach the meter level, with significant jitter. Even combining perception information to correct positioning information cannot achieve the positioning accuracy of high-precision maps, and it can also cause positioning jumps, resulting in inaccurate recordings of both the vehicle's and obstacle's positions. Summary of the Invention
[0005] The purpose of this invention is to address the shortcomings of existing technologies by providing a trajectory correction method, medium, device, and vehicle to achieve low-cost, efficient, and accurate correction of the trajectory information of autonomous vehicles.
[0006] To achieve the above objectives, the first aspect of the present invention provides a trajectory correction method, the method comprising:
[0007] Select the zero-point frame of the collected multi-frame driving information;
[0008] Based on the vehicle state information of the driving information of the zero-point frame, the theoretical position information of the adjacent frames of the zero point is generated by the Ackermann steering model.
[0009] Based on the theoretical position information, the vehicle state information of the driving information of the zero-point adjacent frames, and the map perception information, the corrected position information of the zero-point adjacent frames is generated by a filtering algorithm.
[0010] The vehicle positions in adjacent frames are corrected based on the corrected position information to generate the first self-vehicle correction trajectory.
[0011] Furthermore, the step of correcting the vehicle positions in adjacent frames based on the corrected position information to generate a first vehicle correction trajectory specifically involves:
[0012] Based on the corrected position information, the vehicle position in adjacent frames is corrected frame by frame through the Ackerman steering model and filtering algorithm to generate multiple corrected position information. Based on the multiple corrected position information and the zero-point frame position information, a first self-vehicle correction trajectory is generated.
[0013] Furthermore, the selection of the zero-point frame of the multi-frame driving information specifically includes:
[0014] When the coarse position confidence of the multi-frame driving information reaches a preset value, a zero-point frame is set.
[0015] Alternatively, zero-point frames can be set periodically according to a preset period.
[0016] Furthermore, after correcting the vehicle positions in adjacent frames based on the corrected position information to generate the first vehicle correction trajectory, the method further includes:
[0017] The first self-correction trajectory is smoothed to generate the second self-correction trajectory.
[0018] Furthermore, after correcting the vehicle positions in adjacent frames based on the corrected position information to generate the first vehicle correction trajectory, the method further includes:
[0019] The first obstacle correction trajectory is generated based on the obstacle status information of the first self-driving vehicle correction trajectory and driving information.
[0020] Furthermore, after generating the first obstacle correction trajectory based on the obstacle state information of the first vehicle correction trajectory and driving information, the method further includes:
[0021] The first obstacle correction trajectory is smoothed to generate the second obstacle correction trajectory.
[0022] Furthermore, after smoothing the first vehicle correction trajectory to generate the second vehicle correction trajectory, the method further includes:
[0023] The second vehicle correction trajectory and the second obstacle correction trajectory are associated with map information.
[0024] Furthermore, the step of associating the second vehicle correction trajectory with map information specifically includes:
[0025] Based on the map perception information of multi-frame driving information, a local map in the zero-point frame coordinate system is drawn, and the positions of the second self-vehicle correction trajectory and the second obstacle correction trajectory are drawn in the local map.
[0026] The location of the local map information in the offline map is calculated using a map matching algorithm;
[0027] The local map information is completed based on the matching relationship between map elements in the local map and the offline map.
[0028] Furthermore, the map matching algorithm is one of the following: template matching algorithm, feature point matching algorithm, and point set registration algorithm.
[0029] A second aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the trajectory correction method as described in any of the first aspects above.
[0030] A third aspect of the present invention provides a computer device including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the trajectory correction method as described in any of the first aspects above.
[0031] A fourth aspect of the present invention provides a vehicle including the computer device described in the third aspect above.
[0032] The present invention provides a trajectory correction method, medium, device, and vehicle. Under the approach of emphasizing perception over mapping, a zero-point frame coordinate system is established. The theoretical positions of adjacent frames of the vehicle are generated through the Ackerman steering model. The positions of adjacent frames are corrected by filtering algorithms in combination with the vehicle state information and map perception information of adjacent frames. The corrected trajectory of the vehicle is then derived. This achieves low-cost, efficient, and accurate recording of trajectory information of autonomous vehicles without relying on high-precision maps. Attached Figure Description
[0033] Figure 1This is one of the flowcharts for the trajectory correction method provided in Embodiment 1 of the present invention;
[0034] Figure 2 This is a schematic diagram of trajectory correction according to Embodiment 1 of the present invention;
[0035] Figure 3 This is the second flowchart of the trajectory correction method provided in Embodiment 1 of the present invention;
[0036] Figure 4 This is the third flowchart of the trajectory correction method provided in Embodiment 1 of the present invention;
[0037] Figure 5 This is the fourth flowchart of the trajectory correction method provided in Embodiment 1 of the present invention;
[0038] Figure 6 This is a schematic diagram illustrating the association between the correction trajectory and the map in Embodiment 1 of the present invention;
[0039] Figure 7 This is a schematic diagram of the computer device structure provided in Embodiment 3 of the present invention. Detailed Implementation
[0040] The present application will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. Furthermore, it should be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings.
[0041] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.
[0042] The trajectory correction method, medium, device, and vehicle provided by this invention acquire multiple frames of driving information. In the zero-point frame coordinate system, the theoretical positions of adjacent frames are generated by the Ackerman steering model based on the coarse positioning information of the zero-point frame. Combined with the coarse positioning information of the vehicle's state information and map perception information of the adjacent frames, the vehicle position of the adjacent frames is corrected by a filtering algorithm. Then, the vehicle position of the next adjacent frame is recursively corrected by the corrected vehicle position. This process is repeated to complete the correction of the vehicle position of multiple frames. Furthermore, the trajectory of the obstacle is corrected based on the relative positional relationship between the obstacle and the vehicle.
[0043] Example 1
[0044] Figure 1 This is one of the flowcharts for the trajectory correction method provided in Embodiment 1 of the present invention, such as... Figure 1 As shown, the trajectory correction method specifically includes:
[0045] Step 110: Select the zero-point frame of the collected multi-frame driving information.
[0046] Specifically, the system acquires multiple frames of driving information synchronously during the driving process. During driving, the autonomous vehicle acquires and records driving information in real time at fixed time intervals. Each frame of driving information acquired at a fixed interval constitutes one frame of driving information. This driving information may include information about surrounding obstacles perceived by the perception module, such as vehicles, pedestrians, and non-motorized vehicles, including their position coordinates relative to the autonomous vehicle, heading angle, and the length, width, and height of the detection box; map element information provided by the perception module, such as lane lines, pedestrian crossings, and traffic lights, including the position information of these map elements relative to the autonomous vehicle; and the autonomous vehicle's own state information, such as coarse positioning information, radial velocity, tangential velocity, acceleration, and steering angle. Each frame includes surrounding obstacle information, map element information, and the autonomous vehicle's own state information at the time of data acquisition.
[0047] A zero-point frame is selected from multiple frames of driving information to establish a zero-point frame coordinate system. The driving information is stored as a multi-frame sequence. Within this sequence, one frame is selected as the baseline zero-point frame. Frames before the zero-point frame are designated as historical frames, and frames after the zero-point frame are designated as future frames. For example, if 9 seconds of driving information are recorded with a time interval of 0.1 seconds, and the frame recorded at the 3rd second is selected as the zero-point frame, then the frame data recorded from 0 to 2.9 seconds before the 3rd second is the historical frame driving information, and the frame data recorded from 3.1 to 9 seconds after the 3rd second is the future frame driving information.
[0048] In one possible implementation, the zero-point frame of multiple frames of driving information is selected as follows:
[0049] When the coarse location confidence of multiple frames of driving information reaches a preset value, a zero-point frame is set. Since the confidence of coarse positioning is higher at certain specific locations during the driving process of an autonomous vehicle, the coarse location confidence is calculated and generated by the positioning module based on preset parameters. When the coarse location confidence of one frame of the multiple frames of driving information reaches the preset value, that frame is set as the zero-point frame. In an optional implementation, the multiple frame sequence is segmented according to its length, with each segment containing an equal number of frames. A zero-point frame is set in each segment. The coarse location confidence of the zero-point frame in each segment must reach a preset value. Preferably, the frame with the highest coarse location confidence in each segment is taken as the zero-point frame of that segment. If no frame in a segment has a coarse location confidence that reaches the preset value, that segment is merged with the adjacent segment. Dividing a multi-frame sequence into segments and recursively extrapolating can avoid correction errors caused by excessively long recursion frames. Since a vehicle position with a high coarse position confidence level is a prerequisite for ensuring the accuracy of the extrapolated vehicle trajectory in each segment, setting a zero-point frame based on the coarse position confidence level can improve the position correction accuracy of other frames in that segment during the extrapolation process.
[0050] In another possible implementation, the method for selecting the zero-point frame of multiple frames of driving information is as follows:
[0051] According to the preset period, zero-point frames are set periodically. For example, if 60 minutes of driving information is recorded in multiple frames, one frame of driving information is recorded every 0.1 seconds. The period of the zero-point frame is set to 1 minute, that is, 60 zero-point frames are set at 60 time points in the 0th, 1st, 2nd, 3rd...59th minute.
[0052] Step 120: Based on the vehicle state information of the driving information of the zero-point frame, generate the theoretical position of the adjacent frame through the Ackerman steering model.
[0053] Specifically, the theoretical positions of vehicles in adjacent frames in the zero-point frame coordinate system are calculated using the Ackermann steering model based on the radial velocity, tangential velocity, acceleration, and steering angle of the vehicle's state information at the zero-point frame.
[0054] In one possible implementation, the method for selecting the zero-point frame of multi-frame driving information is as follows: when the coarse position confidence of the multi-frame driving information reaches a preset value, a zero-point frame is set; when the zero-point frame is in the middle of the multi-frame sequence, it is preceded and followed by historical frames and future frames, respectively. Specifically, the above-mentioned calculation of the theoretical position of the vehicle in the zero-point frame coordinate system based on the radial velocity, tangential velocity, acceleration, and steering angle of the vehicle's state information at the zero-point frame time using the Ackerman steering model involves calculating the theoretical positions of the adjacent historical and future frames based on the radial velocity, tangential velocity, acceleration, and steering angle of the vehicle's state information at the zero-point frame time using the Ackerman steering model.
[0055] In one possible implementation, the method for selecting the zero-point frame of multi-frame driving information is as follows: Zero-point frames are periodically set according to a preset period, where the zero-point frame is either the first or last frame of each multi-frame sequence. Then, the calculation of the theoretical position of the vehicle in the zero-point frame coordinate system based on the radial velocity, tangential velocity, acceleration, and steering angle of the vehicle's state information at the zero-point frame time using the Ackermann steering model specifically involves: calculating the theoretical position of the adjacent historical frame or future frame based on the radial velocity, tangential velocity, acceleration, and steering angle of the vehicle's state information at the zero-point frame time using the Ackermann steering model. The calculation of the adjacent historical frame occurs when the zero-point frame is the last frame of each multi-frame sequence, and the calculation of the adjacent future frame occurs when the zero-point frame is the first frame of each multi-frame sequence.
[0056] Step 130: Based on the theoretical position information, the vehicle state information of the driving information of the zero-point adjacent frames, and the map perception information, the corrected position information of the zero-point adjacent frames is generated by the filtering algorithm.
[0057] Among them, the vehicle status information of the driving information of the zero-point adjacent frames includes the coarse positioning information of the zero-point adjacent frames, and the map perception information includes the relative positional relationship between the vehicle and map elements. The coarse positioning information of the zero-point adjacent frames and the relative positional relationship between the vehicle and map elements can be understood as the observed value of the vehicle position.
[0058] Since calculating the vehicle's theoretical position in the next frame using the Ackerman steering model is a discrete-state inference, the vehicle's driving state may remain consistent throughout the time interval. Furthermore, due to dynamic factors, the vehicle's actual position may differ from its theoretical position. Therefore, using a filtering algorithm to combine the observed and theoretical vehicle positions yields a more reliable vehicle position.
[0059] Specifically, based on the calculated theoretical positions of adjacent frames, the actual coarse positioning information of adjacent frames, and the relative positional relationship between the vehicle and map elements in the frame, that is, based on the theoretical and observed positions, a filtering algorithm is used to correct the coarse positioning position of the adjacent frames, generating a more reliable corrected position for the vehicle. This frame is marked as the corrected frame. In possible implementations, the filtering algorithm can be, but is not limited to, one of Kalman filtering, particle filtering, Bayesian estimation, and maximum likelihood estimation.
[0060] In a specific example Figure 2 This is a schematic diagram of trajectory correction according to Embodiment 1 of the present invention, as shown below. Figure 2 As shown, this includes vehicle position 1, vehicle position 2, vehicle position 3, and lane line 4.
[0061] Here, the time at which vehicle position 1 is located is assumed to be t, which can be any frame time. When it is the zero frame time, the position of vehicle 1 is the coarse positioning position. When it is any other time, the position of vehicle 1 is the corrected position. Vehicle position 2 is the theoretical position of vehicle position 1 generated by the Ackerman steering model based on the radial velocity, tangential velocity, acceleration and steering angle of vehicle position 1. Vehicle position 3 is the corrected position generated by Kalman filtering based on the coarse positioning information of vehicle position 2 and time t+1 and the relative position information of the vehicle and the lane line.
[0062] Step 140: Correct the vehicle positions in adjacent frames based on the correction position information to generate the first vehicle correction trajectory.
[0063] Specifically, based on the corrected position information, the vehicle position in adjacent frames is corrected frame by frame on the basis of the corrected position of the already corrected frame through the Ackerman steering model and filtering algorithm, generating multiple corrected position information. Based on the multiple corrected position information and the zero-point frame position information, the first vehicle correction trajectory is generated.
[0064] When calculating the theoretical position of the next frame using the vehicle's state information in the current frame, the coarse positioning position is only needed when calculating the theoretical position of the adjacent frame in the zero-point frame. The theoretical position of the adjacent frame is calculated using the corrected position in other frames. Since the coarse positioning position has a large error, the theoretical position of the adjacent frame is calculated only when the coarse positioning position has a high reliability in the zero-point frame. The theoretical position of the other recursively calculated frames no longer depends on the coarse positioning position to ensure the accuracy of trajectory correction.
[0065] In a preferred embodiment, such as Figure 3 As shown, after step 140, the following steps are also included:
[0066] Step 210: Smooth the first self-driving correction trajectory to generate the second self-driving correction trajectory.
[0067] The first self-correcting trajectory is generated by connecting multiple discrete self-correcting positions, considering only the motion state and constraints between adjacent points, without considering the smoothness constraints of the entire curve. However, the actual trajectory of the vehicle should be a relatively smooth curve. Smoothing can further correct the first self-correcting trajectory. In possible real-time methods, smoothing can employ algorithms such as polynomial smoothing, exponential smoothing, sine curve smoothing, and approximation.
[0068] In another preferred embodiment, such as Figure 4 As shown, after step 140, the following steps are also included:
[0069] Step 310: Generate the first obstacle correction trajectory based on the obstacle status information of the first self-driving vehicle correction trajectory and driving information.
[0070] The obstacle state information includes the relative position information between the vehicle and the obstacles. Due to the limited accuracy of the perception algorithm, the acquisition error of information such as the speed, acceleration, and wheel steering angle of obstacles other than the vehicle is relatively large, making it impossible to directly use the Ackerman motion model combined with Kalman filtering for trajectory correction. However, the relative position information between the vehicle and the obstacles is highly accurate. Based on multiple vehicle position correction points in the first vehicle correction trajectory in the zero-point frame coordinate system and the relative position information between the vehicle and the obstacles, multi-frame obstacle correction positions in the zero-point frame coordinate system are generated. Connecting the multi-frame obstacle correction positions generates the first obstacle correction trajectory.
[0071] Step 320: Smooth the first obstacle correction trajectory to generate the second obstacle correction trajectory.
[0072] The first obstacle correction trajectory is generated by connecting multiple discrete obstacle correction positions, considering only the motion state and constraints between adjacent points, without considering the smoothness constraints of the entire curve. However, the actual trajectory of obstacles such as motor vehicles and high-speed non-motorized vehicles should be a relatively smooth curve. Smoothing can further correct the first obstacle correction trajectory. In possible real-time methods, smoothing can employ algorithms such as polynomial smoothing, exponential smoothing, sine curve smoothing, and approximation.
[0073] In another preferred embodiment, such as Figure 5 As shown, after steps 320 and 210, the following steps are also included:
[0074] Step 410: Associate the second vehicle correction trajectory and the second obstacle correction trajectory with the map information.
[0075] Once the generated second vehicle correction trajectory and second obstacle correction trajectory are correlated with map information, they can be effectively used as model training data and for more comprehensive simulation testing.
[0076] Based on map perception information from multiple frames of driving data, a local map in the zero-point frame coordinate system is drawn, and the positions of the second vehicle correction trajectory and the second obstacle correction trajectory in the local map are plotted. The map perception information is generated by a perception module, such as a camera, sensing map elements around the vehicle. The local map is drawn in the zero-point frame coordinate system, using the zero-point frame as a reference. Map elements around the vehicle mainly include road markings such as lane lines, pedestrian crossings, and stop lines, as well as traffic lights and signs. The second vehicle correction trajectory and the second obstacle correction trajectory are trajectories in the zero-point frame coordinate system, and the correction trajectories are directly plotted in the local map.
[0077] The location of the local map information in the offline map is calculated using a map matching algorithm. Since the sensor's sensing range is limited, and the map information required in subsequent trajectory prediction and planning modules often exceeds the sensing range, it is necessary to find usable map elements in the offline map to complete the information based on the current location. In the absence of a high-precision map, a navigation-level map is used as the offline map. Available map elements are selected from the navigation-level map. Based on the coarse positioning of the zero-point frame, a preset range is selected in the offline map. Map element vectors are obtained from this preset range and saved to form a first vector set. Simultaneously, a second vector set of surrounding map elements is generated in the local map, centered on the zero-point frame. The second vector set is a subset of the first vector set. The location of the local map information in the offline map is calculated using a map matching algorithm based on the first and second vector sets. In possible implementations, the map matching algorithm can be a template matching algorithm, a feature point matching algorithm, or a point set registration algorithm, etc.
[0078] Based on the matching relationship between map elements in the local map and the offline map, complete the local map information. Obtain the location of the local map information in the offline map, and fill in the missing map elements in the local map relative to the offline map.
[0079] Figure 6 This is a schematic diagram illustrating the association between the correction trajectory and the map in Embodiment 1 of the present invention, as shown below. Figure 6 As shown, the map includes the vehicle correction trajectory 5, obstacle correction trajectory 6, perception map element 7, and supplementary map element 8. The map information includes lane lines, stop lines, and pedestrian crossings. Perception map element 7 is represented by thick lines, while supplementary map element 8, completed from the offline map, is represented by thin lines. Figure 6 The system extends lane lines and completes pedestrian crossings using offline maps.
[0080] In this embodiment, under the heavy perception and light map approach, the coarse positioning trajectory of the vehicle is corrected by using the Ackerman steering model, filtering algorithm and trajectory smoothing, and the obstacle trajectory is indirectly corrected by using the precise relative position of the obstacle and the vehicle.
[0081] Example 2
[0082] Embodiment 2 of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the trajectory correction methods provided in Embodiment 1 above.
[0083] Example 3
[0084] Embodiment 3 of the present invention provides a computer device, Figure 7 This is a schematic diagram of the computer device structure provided in Embodiment 3 of the present invention. Figure 7As shown, it includes a memory 500, a processor 600, and a computer program stored in the memory. The processor executes the computer program to implement any of the trace correction methods provided in Embodiment 1 above.
[0085] Example 4
[0086] Embodiment 4 of the present invention provides a vehicle including the computer device described in Embodiment 3 above.
[0087] The term "vehicle" as used in this application can refer to vehicles with passenger carrying function (such as passenger cars, buses, etc.), cargo carrying function (such as ordinary trucks, box trucks, trailer trucks, enclosed trucks, tank trucks, flatbed trucks, container trucks, dump trucks, special structure trucks, etc.), tool function (such as logistics delivery vehicles, automated guided vehicles (AGVs), patrol vehicles, cranes, excavators, bulldozers, loaders, road rollers, off-road engineering vehicles, armored engineering vehicles, sewage treatment vehicles, sanitation vehicles, vacuum trucks, floor scrubbers, water sprinkler trucks, sweeping robots, lawnmowers, golf carts), entertainment function (such as recreational vehicles, amusement park autopilots, balance bikes), or special rescue function (such as fire trucks, ambulances, power repair vehicles, engineering emergency rescue vehicles, etc.).
[0088] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.
[0089] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented in hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.
[0090] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.
Claims
1. A trajectory correction method, characterized in that, The method includes: Select the zero-point frame of the collected multi-frame driving information; Based on the vehicle state information of the driving information of the zero-point frame, the theoretical position information of the adjacent frames of the zero point is generated by the Ackermann steering model. Based on the theoretical position information, the vehicle state information of the driving information of the zero-point adjacent frames, and the map perception information, the corrected position information of the zero-point adjacent frames is generated by a filtering algorithm. The vehicle positions in adjacent frames are corrected based on the corrected position information to generate the first vehicle correction trajectory; The step of correcting the vehicle positions in adjacent frames based on the corrected position information to generate a first vehicle correction trajectory specifically involves: Based on the corrected position information, the vehicle position of adjacent frames is corrected frame by frame through the Ackerman steering model and filtering algorithm to generate multiple corrected position information. Based on the multiple corrected position information and the zero-point frame position information, the first vehicle correction trajectory is generated. After correcting the vehicle positions in adjacent frames based on the corrected position information to generate the first vehicle correction trajectory, the method further includes: The first self-correction trajectory is smoothed to generate the second self-correction trajectory; After smoothing the first vehicle correction trajectory to generate the second vehicle correction trajectory, the method further includes: The second vehicle correction trajectory and the second obstacle correction trajectory are associated with map information; The step of associating the second vehicle correction trajectory with map information specifically includes: Based on the map perception information of multi-frame driving information, a local map in the zero-point frame coordinate system is drawn, and the positions of the second self-vehicle correction trajectory and the second obstacle correction trajectory are drawn in the local map. The location of local map information in the offline map is calculated using a map matching algorithm; The local map information is completed based on the matching relationship between map elements in the local map and the offline map.
2. The trajectory correction method according to claim 1, characterized in that, Selecting the zero-point frame of the multiple frames of driving information specifically includes: When the coarse position confidence of the multi-frame driving information reaches a preset value, a zero-point frame is set. Alternatively, zero-point frames can be set periodically according to a preset period.
3. The trajectory correction method according to claim 1, characterized in that, The map matching algorithm is one of the following: template matching algorithm, feature point matching algorithm, and point set registration algorithm.
4. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the steps of the method according to any one of claims 1-3.
5. A computer device comprising a memory, a processor, and a computer program stored in the memory, characterized in that, The processor executes the computer program to implement the steps of the method according to any one of claims 1-3.
6. A vehicle, characterized in that, Includes the computer device as described in claim 5.