Automatic parking method of two-stage environmental sanitation road sweeper

Abstract

The invention discloses an automatic parking method of a two-stage environmental sanitation road sweeper, and belongs to the technical field of automatic driving. The method comprises the following steps: constructing a double-map system comprising an RTK global map and an SLAM local high-precision map; unifying multi-source positioning data through coordinate transformation; in the global navigation stage, a path planning algorithm is adopted to generate a path from the starting point to the parking point; when the switching condition is met, a local parking stage is turned to, and a parking track conforming to vehicle kinematics constraints is generated; carrying out smoothing processing on the path joint through spline interpolation; and finally, trajectory tracking control is realized through a bottom layer controller. According to the invention, through double-map cooperation and a two-stage decoupling architecture, the contradiction between the planning efficiency and the track feasibility in the prior art is solved, and the technical effects of improving the parking end point precision, reducing the planning time consumption and improving the control stability are realized.

Description

Technical Field

[0001] This invention relates to a two-stage automatic parking method for sanitation sweepers, belonging to the field of autonomous driving technology. Background Technology

[0002] Automated parking technology is a key component in improving vehicle intelligence and operational efficiency. Current mainstream solutions generally adopt a unified path planning architecture, such as the parking path generation method based on the hybrid AlphaGo algorithm proposed in invention patent CN120621342A, and the two-stage planning scheme combining hybrid AlphaGo and geometric algorithms disclosed in CN120773723A.

[0003] However, these existing technologies have significant limitations in the practical application scenarios of sanitation sweepers. Sanitation work environments are characterized by large spatial spans and complex local areas, requiring both efficient navigation within a range of hundreds of meters and precise parking in narrow, non-standard spaces. Existing technical solutions mainly suffer from the following problems: 1. The algorithm architecture is too simplistic, failing to balance the efficiency requirements of long-distance navigation with the accuracy requirements of near-field parking. Using a hybrid AlphaGo algorithm for the entire path results in excessive computational load, while employing a simple geometric algorithm ignores vehicle kinematic constraints, leading to infeasible trajectories.

[0004] 2. The map system is rigid; a single-resolution map cannot simultaneously meet the dual requirements of large-scale coverage and local detail perception. High-resolution global maps have huge computational costs, while low-resolution maps cannot accurately model parking space boundaries and obstacles.

[0005] 3. The lack of an intelligent phase switching mechanism leads to discontinuous path connections during the transition from global navigation to local parking, resulting in abrupt changes in control commands and affecting operational stability.

[0006] 4. Insufficient environmental adaptability; lacks effective fault-tolerant mechanisms for typical sanitation operation scenarios such as GNSS signal obstruction and dynamic obstacle interference.

[0007] Therefore, there is an urgent need for an automated parking solution that can dynamically adjust the planning strategy according to the characteristics of the task, taking into account efficiency, accuracy and robustness. Summary of the Invention

[0008] The purpose of this invention is to overcome the shortcomings of the prior art and provide a two-stage automatic parking method for sanitation sweepers, overcoming the technical defects of existing automatic parking technology in sanitation sweeper application scenarios, such as low planning efficiency, insufficient trajectory feasibility, and poor system adaptability.

[0009] To achieve the above objectives, the present invention is implemented using the following technical solution: In a first aspect, the present invention provides a two-stage automatic parking method for a sanitation sweeper, comprising the following steps: Step 1: Build a global map; Step 2: Based on the global map, use the first path planning algorithm to generate a global path from the vehicle's current location to the preset parking spot; Step 3: When a vehicle is detected to enter the switching area set based on the waiting parking point, a local high-precision map is constructed in real time, and a local trajectory from the waiting parking point to the target parking space is generated based on the local high-precision map using the second path planning algorithm. Step 4: Perform path smoothing processing on the connection points between the global path and the local trajectory to generate a continuous and executable overall trajectory; Step 5: Control the vehicle to complete the parking operation based on the overall trajectory.

[0010] Furthermore, a global map is constructed, including: The global map is constructed based on real-time dynamic positioning (RTK) technology, covering the road network of the entire work area; Furthermore, the grid resolution of the global map is set to 0.5 meters per grid.

[0011] Furthermore, real-time construction of local high-precision maps includes: The local high-precision map is constructed in real time based on laser simultaneous localization and mapping (SLAM) technology, which can represent parking space boundaries, curbs and obstacle information.

[0012] Furthermore, the grid precision of the local high-precision map reaches the centimeter level.

[0013] Furthermore, it also includes the coordinate system step: By using the coordinate transformation tf tree in the robot operating system ROS, the RTK world coordinate system of the global map and the SLAM local coordinate system of the local high-precision map are aligned in real time to ensure spatial consistency of path planning.

[0014] Furthermore, the first path planning algorithm is either the A* algorithm or the Dijkstra algorithm; Based on the global map, a first path planning algorithm is used to generate a global path from the vehicle's current location to a preset parking spot, including: Starting from the vehicle's current RTK pose and ending at the pre-defined "parking point" in front of the target parking space entrance, run the A* algorithm or Dijkstra algorithm on the global map to generate the shortest feasible path that ignores vehicle kinematic constraints.

[0015] Furthermore, the second path planning algorithm is a hybrid A* algorithm or a planning algorithm based on the Reeds-Shepp curve; A second path planning algorithm is used to generate a local trajectory from the parking spot to the target parking space, including: The search is performed in the continuous state space. The hybrid A* algorithm or the planning algorithm based on the Reeds-Shepp curve is launched. Combined with the nonholonomic constraints of the vehicle, an executable parking trajectory that satisfies the Reeds-Shepp motion model is generated, and a local trajectory with the destination being the center of the target parking space and the specified orientation is obtained. The vehicle incomplete constraints include minimum turning radius constraints and forward / reverse gear switching constraints.

[0016] Furthermore, the switching area set based on the parking spot includes: The distance between the vehicle and the parking spot is less than or equal to 10 meters, and the deviation between the vehicle's heading angle and the preset heading angle at the parking spot is less than or equal to 15 degrees.

[0017] Furthermore, the path smoothing process employs an interpolation optimization method, specifically including: The end of the global path and the beginning of the local trajectory are optimized to ensure that the position, heading and curvature at the splicing point change continuously.

[0018] Further optimization methods include: By using cubic spline interpolation or Bézier curve fitting, we ensure that the position, heading, and curvature at the splice point change continuously.

[0019] Furthermore, when generating the local trajectory from the parking spot to the target parking space using the second path planning algorithm, a location anomaly handling step is also included: Continuously monitor the reliability of SLAM positioning data; When the difference between the SLAM pose and the reliable RTK pose exceeds a preset value, the system automatically reverts to the RTK pose and uses the RTK positioning data as the planning reference. Re-trigger local high-precision map creation and local trajectory planning.

[0020] Furthermore, the preset value is 0.3m.

[0021] Furthermore, controlling the vehicle to complete the parking operation includes: The overall trajectory is sent to the underlying vehicle controller via CAN bus or ROS driver interface; The trajectory is tracked and controlled in real time using either a pre-aiming tracking algorithm or a pure tracking algorithm until the vehicle is parked in the target parking space and the endpoint positioning accuracy reaches the preset accuracy range.

[0022] Furthermore, the preset accuracy range is within ±5 cm.

[0023] Secondly, the present invention provides a two-stage path planning system for automatic parking of sanitation sweepers, comprising: A high-precision RTK-GNSS receiver and inertial measurement unit are used to collect environmental information; LiDAR is used to detect faulty components. The onboard computing unit is connected to the lidar, the high-precision RTK-GNSS receiver, and the inertial measurement unit, respectively, and is used to execute the two-stage automatic parking method for the sanitation sweeper as described in the first aspect based on the environmental information and fault information.

[0024] Compared with the prior art, the beneficial effects achieved by the present invention are as follows: 1. By decoupling tasks and dividing algorithms, the computationally complex kinematic planning is restricted to a necessary local range, which greatly improves system efficiency; the collaborative work of the dual map system ensures the feasibility of global navigation and meets the accuracy requirements of local docking; intelligent stage switching and smooth processing ensure the stability of the control process; multiple fault-tolerant mechanisms enhance the system's adaptability in complex operating environments.

[0025] 2. This invention achieves collaborative optimization based on a dual-map system and algorithmic division of labor. Due to the adoption of the aforementioned dual-map system architecture, coupled with the defined algorithmic division of labor strategy, the A algorithm, which ignores kinematic constraints, is used in the global navigation phase to perform efficient searches on the global map, keeping the long-distance path planning time within a preset range.

[0026] 3. This invention strictly limits the parking phase to a defined switching area and employs a hybrid A algorithm that considers vehicle kinematic constraints to generate a feasible trajectory on a centimeter-level high-precision map, ensuring that the parking endpoint accuracy is within ±5 centimeters. This "division of labor and cooperation" architecture fundamentally solves the technical contradiction in existing "one-size-fits-all" algorithms that cannot balance efficiency and accuracy.

[0027] 4. Through intelligent switching and smoothing, this invention significantly improves the quantification of switching conditions for control stability settings (distance ≤ 10 meters and heading deviation ≤ 15 degrees). This ensures that the vehicle begins local parking under ideal conditions using cubic spline interpolation or Bézier curve fitting, effectively eliminating curvature abrupt changes at path junctions in traditional methods. This allows the specially designed pre-aiming and tracking algorithm to execute smoothly. Actual tests show that this smoothing reduces the fluctuation amplitude of control commands, significantly improving the smoothness and comfort of the parking process.

[0028] 5. The multi-source positioning fusion mechanism of this invention enhances system robustness. Based on the coordinate unification scheme and anomaly handling mechanism, the coordinate system implemented through ROS tf trees provides the technical foundation for data fusion of RTK and SLAM, while the drift threshold triggering mechanism can detect and correct SLAM positioning anomalies in a timely manner. This feature enables this invention to effectively cope with GNSS signal obstruction scenarios such as underground parking garages and under overpasses, improving system availability.

[0029] 6. The engineering applicability based on parametric design is significantly expanded. The combination of 0.5-meter global resolution and centimeter-level local accuracy optimizes the allocation of computing resources while ensuring navigation performance; the fault tolerance threshold balances the sensitivity of fault detection and the requirements of system stability. These optimized parameter systems enable the present invention to be quickly adapted to different models of sanitation sweepers, greatly reducing the debugging costs for engineering applications.

[0030] 7. This invention forms a complete technical closed loop from overall architecture to final execution. This system-level optimization improves the overall parking success rate while shortening the average parking time, providing a comprehensively optimized technical solution for the automatic parking of special vehicles such as sanitation sweepers. Attached Figure Description

[0031] Figure 1 This is an architecture diagram of an automatic parking system based on "dual maps + two-stage planning"; Figure 2 This is a schematic diagram of the splicing of the global A-satellite path and the local hybrid A-satellite trajectory. Detailed Implementation

[0032] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and should not be used to limit the scope of protection of the present invention. Example 1

[0033] This invention provides a two-stage path planning method for automatic parking of sanitation sweepers, such as... Figure 2 As shown, the specific steps include: (1) Construction of a dual-map system Global map: Built based on RTK high-precision positioning data, covering the entire parking lot road network, with a grid resolution of 0.5 meters / grid, used for large-scale route navigation; Local map: When the vehicle enters the preset parking area (that is, the switching area), the laser SLAM system (such as LOAM or Cartographer) is activated to build a high-precision local map at the centimeter level in real time, accurately modeling the parking space boundary, pillars, curbs and static obstacles; Coordinate System 1: The RTK world coordinate system, SLAM local coordinate system and unified map coordinate system are spatiotemporally aligned using the tf tree in ROS to ensure the consistency of path planning across different maps.

[0034] (2) Two-stage path planning Phase 1 (Global Navigation): Starting from the vehicle's current RTK pose and ending at the pre-set "parking point" in front of the target parking space entrance, the standard A* algorithm is run on the global map to generate the shortest feasible path that ignores vehicle kinematic constraints, pursuing planning efficiency. The second stage (partial parking): When the vehicle approaches the parking spot (e.g., distance ≤ 10 meters and heading deviation ≤ 15°, enter the switching area), switch to the local high-precision map, start the hybrid A* algorithm, and combine the vehicle's non-holonomic constraints (minimum turning radius, forward / reverse gear switching) to generate an executable parking trajectory that satisfies the Reeds-Shepp motion model, with the endpoint being the center of the target parking space and the specified orientation.

[0035] (3) Path smoothing and state connection At the two-stage transition point, the path smoothing module is invoked to perform interpolation optimization between the end of the global path and the start of the local path to ensure the continuity of position, heading angle and curvature. If SLAM positioning drifts due to dynamic obstacle interference, the system can revert to RTK pose and re-trigger local mapping to ensure the reliability of the planning starting point.

[0036] (4) System integration and control closed loop Develop a ROS path selection node that determines the current stage based on the vehicle's real-time pose and automatically calls the corresponding planner. The final planned path is smoothed and then published to the underlying vehicle controller (via CAN or ROS driver interface) to achieve closed-loop tracking; RViz supports visualization of global paths, local search trees, high-precision maps, and parking trajectories, facilitating debugging and verification. RViz is a 3D visualization tool in ROS used to display robot sensor data, models, path planning, and other information in real time. Example 2

[0037] This embodiment uses a sanitation sweeper equipped with an RTK-GNSS receiver and a 16-line lidar as a platform, and a robot operating system as the software framework to describe in detail the implementation of the present invention.

[0038] Reference Figure 1 The system architecture shown, and the implementation environment of this invention, includes the following core components: The localization and perception system consists of an RTK localization module and a laser SLAM system. RTK localization provides global absolute coordinates; laser SLAM, based on the Cartographer algorithm, is responsible for building high-precision environmental maps in local areas.

[0039] The map management system maintains a dual-map structure: the global map is generated based on pre-collected RTK data, using a 0.5-meter resolution raster format, covering the entire road network of the parking lot; the local high-precision map is built in real time by SLAM, with a resolution of 5 centimeters, including detailed information such as parking space boundaries, curbs, and pillars.

[0040] The coordinate unification framework is implemented using ROS's tf tree, establishing the transformation relationship between rtk_map, slam_map and the unified map coordinate system, ensuring that spatial data from different sources can be correctly aligned.

[0041] The planning system consists of a global path planner and a local parking planner. The global planner uses the A* algorithm to search on the global map; the local planner uses a hybrid A* algorithm to solve for trajectories that meet kinematic constraints in a continuous state space.

[0042] The control execution system communicates with the vehicle's underlying controller via a CAN bus interface and uses a pre-aiming tracking algorithm to achieve accurate trajectory tracking.

[0043] Reference Figure 2 The workflow shown is illustrated, and the specific implementation steps of this invention are as follows: After the system starts, it first performs initialization: loading the global map, starting the SLAM system, and establishing coordinate transformation relationships. When a parking command is received, the system enters the working state.

[0044] During the global navigation phase, the planner uses the vehicle's current RTK pose as the starting point and the parking spot as the ending point to perform the A* algorithm search. The parking spot is preset 5-10 meters in front of the target parking space entrance, with its heading aligned with the ideal parking direction. The path generated by the A* algorithm consists of a series of discrete grid points, without considering the vehicle's kinematic constraints, prioritizing planning efficiency.

[0045] The system continuously monitors the relative position and orientation of the vehicle and the parking spot. When the distance is less than or equal to 10 meters and the heading deviation is less than or equal to 15 degrees, the switching conditions are met, and the system enters the partial parking phase.

[0046] During the local parking phase, the system switches to a local high-precision map constructed using SLAM and launches the Hybrid A* planner. The Hybrid A* algorithm considers the minimum turning radius of the sanitation sweeper (set to 5 meters) and forward / reverse switching constraints, and generates a feasible trajectory from the current position to the center of the target parking space based on the Reeds-Shepp motion model.

[0047] Because the A-satellite path and the hybrid A-satellite trajectory differ in their representation and constraints, direct splicing would lead to control difficulties. This invention employs cubic spline interpolation to jointly optimize the connecting region of the two path segments, focusing on ensuring the continuity of position, heading, and curvature to generate a smooth and trackable complete trajectory.

[0048] During local planning, the system continuously evaluates the positioning quality. When the difference between the SLAM pose and the reliable RTK pose exceeds 0.3 meters, it is considered that a positioning drift has occurred, and the system immediately switches back to RTK positioning and re-performs local planning to ensure the reliability of the system.

[0049] Finally, the smoothed trajectory is published to the underlying controller via the ROS node. The controller receives commands via the CAN bus and uses a pure tracking algorithm to track the trajectory in real time until the vehicle accurately parks in the target parking space.

[0050] The implementation results of this embodiment show that, in typical sanitation operation scenarios, the global planning time is less than 0.1 seconds, and the average local planning time is 0.5-1 seconds. The entire process from starting automatic parking to completing parking is smooth and stable, with the parking endpoint position error controlled within ±3 cm and the heading error less than 2 degrees, fully verifying the effectiveness of the invention.

[0051] This embodiment demonstrates the flexibility and scalability of the invention in its technical implementation. The core architecture of this invention is also applicable to automated scheduling scenarios for container trucks in ports, but adjustments can be made to the specific technology selection: The global map can be automatically generated using high-resolution satellite imagery combined with deep learning road recognition technology, providing a global path network for the port area; the local map can be constructed using visual SLAM and UWB positioning fusion to meet the precise docking requirements of container trucks.

[0052] Global path planning can use Dijkstra's algorithm, which is more suitable for the network structure characteristics of port area roads; local planning can use trajectory generation methods based on optimal control, which can better handle the complex dynamic characteristics of container trucks.

[0053] Path smoothing can be achieved by using Bézier curve fitting, which provides better curvature characteristics while ensuring continuity, thus meeting the control requirements of large vehicles.

[0054] Although the specific technical implementations differ, the core architecture of "dual-map collaboration, two-stage decoupling, intelligent switching and smooth connection" is consistent with that of this invention, and can achieve similar technical effects, proving that this invention has good versatility and adaptability.

[0055] This invention achieves synergistic optimization of efficiency, accuracy, and robustness in automatic parking through a "dual-map system" and a "two-stage path planning" architecture. Specific beneficial effects are as follows: (1) Significantly improve planning efficiency: The computationally complex hybrid A-Star algorithm is strictly limited to a local small range (<10 meters), and a lightweight A-Star algorithm is used for long-distance navigation. The overall planning time is controlled within 1 second (local stage), which meets the real-time requirements of high-frequency operation scenarios such as sanitation vehicles. (2) Enhance parking accuracy and safety: Local SLAM maps provide centimeter-level environmental modeling capabilities, accurately identify non-standard parking spaces and small obstacles, and generate executable trajectories by combining kinematic constraints, with parking endpoint accuracy reaching ±5cm; (3) Improve system environmental adaptability: The dynamic switching between RTK and SLAM dual positioning sources can rely on global absolute coordinates when GNSS signal is good, and seamlessly switch to local relative positioning in signal-blocked areas (such as underground garages and under overpasses) to adapt to complex indoor and outdoor operating environments; (4) Ensure trajectory continuity and control stability: Through the parking point anchoring and path smoothing mechanism, pose jumps during stage switching are effectively avoided, ensuring that the underlying controller receives smooth and continuous trajectory instructions, thereby improving the stability and safety of the parking process.

[0056] In other embodiments, the present invention may employ the following alternatives: (1) Alternatives to dual-map systems: High-resolution satellite images or aerial images can be used as the basic data source for the global map, combined with deep learning algorithms for automatic identification and updating of parking lot road networks. For local high-precision maps, in addition to laser SLAM, visual SLAM (such as ORB-SLAM), RGB-D SLAM, or UWB-based indoor navigation systems can be used to construct maps with parking space-level precision. (2) Alternatives to path planning algorithms: In the global navigation phase, in addition to the A* algorithm, other efficient graph search algorithms such as Dijkstra's algorithm and RRT (Rapid Random Tree) can also be used. For local parking planning, in addition to hybrid A* algorithms, Reeds-Shepp curves can be used to directly generate executable trajectories, or reinforcement learning methods can be applied to train the agent to autonomously learn the optimal parking strategy. (3) Alternatives to path smoothing: In addition to existing interpolation optimization methods, mathematical models such as Bézier curves and spline interpolation can be introduced to smooth the path at the splicing point, ensuring that the vehicle's trajectory is smoother and more natural during execution, and reducing the control difficulty caused by curvature abrupt changes.

[0057] The technical solution of this invention is not only applicable to the field of automatic parking of sanitation sweepers, but can also be extended to other scenarios requiring high-precision path planning and environmental perception, such as: (1) Autonomous delivery vehicles: By using a dual-map system and two-stage path planning technology, the efficiency and safety of logistics delivery vehicles in complex warehousing environments are improved; (2) Port container trucks: In the automated terminal, the precise docking and loading / unloading operations of container trucks are achieved by using a dual-map system and two-stage path planning technology.

[0058] The key technical point of this invention is: (1) Dual map system construction technology: A low-resolution global map is constructed using RTK high-precision positioning data to cover the entire parking lot road network for large-scale navigation; at the same time, after the vehicle enters the preset parking area, the laser SLAM system is activated to construct a centimeter-level high-precision local map in real time, accurately modeling parking space boundaries, pillars, curbs and static obstacles, and unifying different coordinate systems through tf trees to ensure the consistency of path planning. (2) Two-stage path planning technology: In the first stage, the standard A* algorithm is used to quickly generate a coarse path on the global map, ignoring vehicle kinematic constraints and pursuing planning efficiency; in the second stage, the local high-precision map is switched to start the hybrid A* algorithm, combined with vehicle nonholonomic constraints (minimum turning radius, forward / reverse gear switching), to generate an executable parking trajectory that satisfies the Reeds-Shepp motion model, with the endpoint being the center of the target parking space and the specified orientation; (3) Path smoothing technology: At the two-stage switching point, the path smoothing module is called to perform interpolation optimization on the global path end and the local path start point to ensure the continuity of position, heading angle and curvature, avoid sudden changes in control commands due to discontinuous trajectory, and improve the smoothness and safety of the parking process. Example 3

[0059] This embodiment provides a two-stage path planning system for automatic parking of sanitation sweepers, including: The main body of the sanitation sweeper; High-precision RTK-GNSS receivers and inertial measurement units (IMUs) are used to achieve centimeter-level absolute positioning outdoors and pose estimation when signals are lost; LiDAR serves as the primary environmental perception sensor, supporting SLAM in building local high-precision maps and detecting parking spaces and obstacles. The onboard computing unit (such as an embedded domain controller supporting ROS) is used to execute the method described in Example 1, and is responsible for running algorithms such as dual map management, two-stage path planning (A-star and hybrid A-star), coordinate system 1 (TF tree), and path smoothing. The drive-by-wire chassis and CAN bus interface module are used to receive planned trajectory commands and control the vehicle's steering, speed, and gear shifting to achieve closed-loop execution.

[0060] Definitions of abbreviations and key terms: Abbreviations:

[0061] RTK --- Real-Time Kinematic (RTK) is a real-time dynamic positioning technology. SLAM --- Simultaneous Localization and Mapping; Key term definitions:

[0062] The A-Star Algorithm is a classic heuristic graph search algorithm used to find the shortest path from the starting point to the target point in a discrete grid map. In this invention, it is used in the global navigation phase, ignoring vehicle kinematic constraints in order to pursue planning efficiency.

[0063] Hybrid A-Star Algorithm --- A path planning algorithm that considers nonholonomic kinematic constraints of the vehicle (such as minimum turning radius, forward / reverse switching). It generates feasible trajectories that conform to the vehicle's dynamic characteristics by introducing continuous control variables (such as steering wheel angle and velocity direction) into the state space.

[0064] Parking waiting point --- Preset in a safe location in front of the entrance to the target parking space (usually 5-10 meters away from the parking space), serving as the end point of the global path and the starting point of the local parking planning, and is the connecting anchor point between the two-stage path planning.

[0065] Path smoothing module --- A software module used to interpolate or optimize the trajectory generated by two-stage planning at the splicing point, ensuring continuity of position, heading angle and curvature, and improving the tracking performance of the underlying controller and the smoothness of the parking process.

[0066] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0067] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0068] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0069] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0070] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A two-stage automatic parking method for a sanitation sweeper, characterized in that, Includes the following steps: Build a global map; Based on the global map, a first path planning algorithm is used to generate a global path from the vehicle's current location to a preset parking spot. When a vehicle is detected to enter the switching area set based on the waiting parking point, a local high-precision map is constructed in real time, and a local trajectory from the waiting parking point to the target parking space is generated based on the local high-precision map using a second path planning algorithm. At the junction of the global path and the local trajectory, path smoothing is performed to generate a continuous and executable overall trajectory. The vehicle completes the parking operation based on the overall trajectory control.

2. The two-stage automatic parking method for sanitation sweepers according to claim 1, characterized in that, Building a global map includes: The global map is constructed based on real-time dynamic positioning (RTK) technology, covering the road network of the entire work area; The grid resolution of the global map is set to 0.5 meters per grid.

3. The two-stage automatic parking method for sanitation sweepers according to claim 1, characterized in that, Real-time construction of local high-precision maps, including: The local high-precision map is constructed in real time based on laser simultaneous localization and mapping (SLAM) technology, which can represent parking space boundaries, curbs and obstacle information. The local high-precision map has a grid precision down to the centimeter level.

4. The two-stage automatic parking method for sanitation sweepers according to claim 1, characterized in that, It also includes the coordinate system step: By using the coordinate transformation tf tree in the robot operating system ROS, the RTK world coordinate system of the global map and the SLAM local coordinate system of the local high-precision map are aligned in real time to ensure spatial consistency of path planning.

5. The two-stage automatic parking method for sanitation sweepers according to claim 1, characterized in that, The first path planning algorithm is either the A* algorithm or the Dijkstra algorithm; Based on the global map, a first path planning algorithm is used to generate a global path from the vehicle's current location to a preset parking spot, including: Starting from the vehicle's current RTK pose and ending at the pre-defined "parking point" in front of the target parking space entrance, run the A* algorithm or Dijkstra algorithm on the global map to generate the shortest feasible path that ignores vehicle kinematic constraints.

6. The two-stage automatic parking method for sanitation sweepers according to claim 1, characterized in that, The second path planning algorithm is either a hybrid A* algorithm or a planning algorithm based on the Reeds-Shepp curve; A second path planning algorithm is used to generate a local trajectory from the parking spot to the target parking space, including: The search is performed in the continuous state space. The hybrid A* algorithm or the planning algorithm based on the Reeds-Shepp curve is launched. Combined with the nonholonomic constraints of the vehicle, an executable parking trajectory that satisfies the Reeds-Shepp motion model is generated, and a local trajectory with the destination being the center of the target parking space and the specified orientation is obtained. The vehicle incomplete constraints include minimum turning radius constraints and forward / reverse gear switching constraints.

7. The two-stage automatic parking method for sanitation sweepers according to claim 1, characterized in that, The switching area set based on the parking spot includes: In the switching area, the distance between the vehicle and the parking spot is less than or equal to 10 meters, and the deviation between the vehicle's heading angle and the preset heading angle at the parking spot is less than or equal to 15 degrees.

8. The two-stage automatic parking method for sanitation sweepers according to claim 1, characterized in that, The path smoothing method employs an interpolation optimization method, including: The ends of the global path and the beginning segments of the local trajectory are optimized to ensure that the position, heading, and curvature at the splicing point change continuously. Optimization methods include: By using cubic spline interpolation or Bézier curve fitting, we ensure that the position, heading, and curvature at the splice point change continuously.

9. The two-stage automatic parking method for sanitation sweepers according to claim 1, characterized in that, When generating the local trajectory from the parking spot to the target parking space using the second path planning algorithm, the method also includes a localization anomaly handling step: Continuously monitor the reliability of SLAM positioning data; When the difference between the SLAM pose and the reliable RTK pose exceeds a preset value, the system automatically reverts to the RTK pose and uses the RTK positioning data as the planning reference. Re-trigger local high-precision map creation and local trajectory planning; And / or, Controlling the vehicle to complete the parking operation includes: The overall trajectory is sent to the underlying vehicle controller; The trajectory is tracked and controlled in real time using either a pre-aiming tracking algorithm or a pure tracking algorithm until the vehicle is parked in the target parking space and the endpoint positioning accuracy reaches the preset accuracy range.

10. A two-stage path planning system for automatic parking of sanitation sweepers, characterized in that, include: A high-precision RTK-GNSS receiver and inertial measurement unit are used to collect environmental information; LiDAR is used to detect faulty components. The on-board computing unit is connected to the lidar, the high-precision RTK-GNSS receiver, and the inertial measurement unit, respectively, and is used to execute the two-stage automatic parking method for sanitation sweepers as described in any one of claims 1-9 based on the environmental information and fault information.

Images