Method and terminal for realizing U-turn of unmanned vehicle based on point cloud

Abstract

The application provides a method and a terminal for realizing U-turn of an unmanned vehicle based on point cloud, acquires point cloud map data and generates an initial map, acquires point cloud data of a current environment and matches the point cloud data with the initial map, determines current position information and attitude information of the vehicle based on motion information of the vehicle, records continuous position information of the vehicle in time sequence, and obtains a motion trajectory of the vehicle based on the continuous position information, and when it is detected that the vehicle is about to reach a U-turn position based on a historical data set, the U-turn is realized within a safe range by using point cloud tracking. The application mainly acquires point cloud map data and generates an initial map, ensures that the vehicle can smoothly perform a U-turn operation in a narrow road environment, and effectively solves the problem that the unmanned vehicle frequently needs to adjust position when performing a U-turn in a narrow road, reduces the operation times, and improves the U-turn efficiency.

Description

Technical Field

[0001] This invention relates to the field of autonomous driving, and in particular to a method and terminal for realizing U-turns of autonomous vehicles based on point clouds. Background Technology

[0002] With the continuous development of autonomous driving technology, driverless vehicles have gradually moved towards practical application. However, in real-world road environments, especially in situations with limited space such as narrow urban roads, alleys, or parking lots, U-turns still face significant challenges. While traditional autonomous driving systems can perform simple driving tasks well, during U-turns on narrow roads, due to limited space, irregular obstacle distribution, and diverse environmental changes, the vehicle needs to frequently adjust its position to complete the turn, resulting in low operational efficiency and increased system instability.

[0003] Traditional methods typically rely on static maps or single sensor information for path planning and obstacle detection. However, these methods have certain limitations. First, static maps cannot reflect real-time environmental changes and may fail to accurately describe real-time obstacle locations and road conditions. Second, in dynamic environments, traditional path planning methods often fail to accurately predict vehicle trajectory and attitude changes, causing vehicles to frequently adjust their positions in confined spaces, increasing the complexity and uncertainty of U-turns. Furthermore, point cloud matching based on traditional algorithms suffers from low matching accuracy in complex environments, with severe occlusion, or high noise levels, also affecting the accuracy of vehicle localization and attitude estimation. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a method and terminal for realizing U-turns of unmanned vehicles based on point clouds, so as to solve the problem of the large number of U-turn operations of unmanned vehicles on narrow roads.

[0005] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: A method for enabling autonomous vehicles to make U-turns based on point clouds includes the following steps: S1. Acquire point cloud map data and generate an initial map; S2. Acquire point cloud data of the current environment and match it with the initial map, and determine the current position and attitude information of the vehicle based on the vehicle's motion information; S3. Record the vehicle's continuous position information in time sequence, and obtain the vehicle's motion trajectory based on the continuous position information; S4. Based on historical datasets, when a vehicle is detected to be approaching a U-turn location, point cloud tracking is used to enable the U-turn.

[0006] To solve the above-mentioned technical problems, another technical solution adopted by the present invention is as follows: A terminal for enabling autonomous vehicles to turn around based on point clouds includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it performs the following steps: S1. Acquire point cloud map data and generate an initial map; S2. Acquire point cloud data of the current environment and match it with the initial map, and determine the current position and attitude information of the vehicle based on the vehicle's motion information; S3. Record the vehicle's continuous position information in time sequence, and obtain the vehicle's motion trajectory based on the continuous position information; S4. Based on historical datasets, when a vehicle is detected to be approaching a U-turn location, point cloud tracking is used to enable the U-turn.

[0007] The beneficial effects of this invention are as follows: It provides a method and terminal for realizing U-turns of unmanned vehicles based on point clouds. It mainly obtains point cloud map data and generates an initial map to ensure that the vehicle can perform U-turn operations smoothly in narrow road environments, thereby effectively solving the problem that unmanned vehicles need to frequently adjust their position when making U-turns on narrow roads, reducing the number of operations and improving U-turn efficiency. Attached Figure Description

[0008] Figure 1 This is a flowchart illustrating a method for enabling an autonomous vehicle to turn around based on point clouds, as described in an embodiment of the present invention. Figure 2 This is a flowchart illustrating the motion trajectory recording process in a method for realizing a U-turn of an unmanned vehicle based on point clouds, as described in an embodiment of the present invention. Figure 3 This is a flowchart of point cloud tracking in a method for realizing U-turn of an unmanned vehicle based on point cloud in an embodiment of the present invention; Figure 4 This is a schematic diagram of a terminal that enables an unmanned vehicle to turn around based on point cloud in an embodiment of the present invention; Label Explanation: 1. A terminal for enabling autonomous vehicles to turn around based on point clouds; 2. Memory; 3. Processor. Detailed Implementation

[0009] To explain in detail the technical content, objectives, and effects of the present invention, the following description is provided in conjunction with the embodiments and accompanying drawings.

[0010] Please refer to Figure 1 A method for enabling autonomous vehicles to make U-turns based on point clouds, comprising the following steps: S1. Acquire point cloud map data and generate an initial map; S2. Acquire point cloud data of the current environment and match it with the initial map. Based on the vehicle's motion information, determine the vehicle's current position and attitude information. S3. Record the vehicle's continuous position information in time sequence, and obtain the vehicle's motion trajectory based on the continuous position information; S4. Based on historical datasets, when a vehicle is detected to be approaching a U-turn location, point cloud tracking is used to enable the U-turn.

[0011] As described above, the beneficial effects of this invention are as follows: by acquiring environmental point cloud data and generating an initial map, the vehicle can accurately perceive its surroundings, thereby enabling efficient U-turns in confined spaces. Preferably, this method utilizes high-precision three-dimensional point cloud data provided by LiDAR (Light Detection and Ranging) and combines it with the vehicle's motion information, attitude information, historical data, and trajectory. Through precise point cloud matching and tracking technology, it can efficiently complete U-turns within limited spaces, reducing the frequency of manual intervention and improving the degree of automation. By combining point cloud maps and motion information, the system can accurately predict the vehicle's trajectory and position, ensuring accuracy and safety during the U-turn process, thus solving the problems of low operational efficiency and instability caused by frequent adjustments in traditional methods.

[0012] In some implementations, step S2 specifically includes the following steps: S21. Optimize the point cloud data of the current environment based on the normal distribution transformation algorithm; S22. Match the optimized point cloud data of the current environment with the initial map, and estimate the initial position and attitude information of the generated vehicle. S23. Based on the vehicle's motion information and the estimated initial position and attitude information, determine the vehicle's current position and attitude information.

[0013] As described above, by introducing the Normal Distribution Transform (NDT) algorithm to optimize the point cloud data of the current environment, the accuracy and robustness of point cloud matching are further improved. Specifically, the NDT algorithm transforms point cloud data into a grid map or Gaussian mixture model. The principle behind this is that the NDT algorithm can model the point cloud using a Gaussian distribution, improving the noise problem in the point cloud data. This allows for accurate acquisition of vehicle positioning and attitude information even in complex environments or with significant occlusion. Simultaneously, combining vehicle motion information for matching effectively reduces matching errors caused by instantaneous motion, improving the system's real-time performance and accuracy, and avoiding the position drift problem that may occur when relying on static maps.

[0014] Specifically, step S23 includes the following steps: S231. Using the vehicle's motion information and the estimated initial position and attitude information, predict the vehicle's current position and attitude information; S232. Determine the search range based on the predicted current position and attitude information; S233. Within the search range, point cloud matching is performed using the iterative nearest point algorithm, and the current position and attitude information of the vehicle are determined based on the matching results.

[0015] As described above, by combining vehicle motion information with initially estimated position and attitude information, accurate search range prediction is achieved. Vehicle motion information, such as IMU (Inertial Measurement Unit) data and odometer information, provides real-time speed and acceleration data. Based on this information, the system can predict the vehicle's current position and automatically adjust the search range, thereby performing point cloud matching more efficiently. Traditional point cloud matching methods often rely solely on static or known maps, neglecting the impact of real-time vehicle motion. This invention, by dynamically adjusting the search range, enables vehicles to perform more flexible and accurate positioning and navigation in confined environments, significantly reducing operational complexity and computational burden.

[0016] Preferably, the process for determining the search range is as follows: The search area is based on the vehicle's length, width, and wheelbase during a U-turn on a narrow road. Centered on the vehicle, the area in front and behind covers 1.5 to 2 times the vehicle's length, and the area to the left and right covers 1.5 to 2 times the vehicle's width. Based on this data, a rectangular area can be marked on the map.

[0017] Adjust the search area based on obstacle information. The search area should avoid obstacles, with a safety boundary of 0.3 to 0.5 meters around each obstacle.

[0018] Preferably, step S233 specifically includes the following steps: Using the vehicle's motion information as the initial estimation data and auxiliary constraints for the iterative nearest point algorithm, the source point cloud data of the current environment and the target point cloud data of the initial gradient are read. The iterative nearest point algorithm is used to perform point cloud matching, and the current position and attitude information of the vehicle are output based on the matching results.

[0019] As described above, the Iterative Closest Point (ICP) algorithm, combined with vehicle motion information as initial estimation and constraints, accurately updates the vehicle's position and attitude by matching the source point cloud data with the target point cloud data of the current environment. The ICP algorithm can iteratively adjust and optimize the vehicle's trajectory in dynamic environments. By incorporating vehicle motion information such as IMU data, the algorithm effectively avoids errors caused by environmental complexity or dynamic changes, thereby improving the system's stability and robustness.

[0020] In the ICP algorithm iteration, the rotation and translation in the calculated results are corrected based on IMU data to improve the stability and reliability of the matching, especially when the vehicle's movement changes rapidly, thus avoiding errors caused by relying solely on power matching.

[0021] After reading the source point cloud data of the current environment and the target point cloud data of the initial gradient, preprocessing is performed on the source and target point clouds. The process involves setting an initial transformation matrix and time interval, initiating multiple iterations, acquiring current IMU data (vehicle motion data) in each iteration, fusing the IMU data into the initial transformation matrix, performing ICP registration, and updating the initial transformation matrix. The final transformation is then applied to the source point cloud.

[0022] In addition, the NDT algorithm is used to optimize the selected region to improve the matching effect.

[0023] In some implementations, step S4 specifically includes: S41. When the vehicle is detected to be about to reach the U-turn position, predict the vehicle's path and attitude changes based on historical datasets and the vehicle's trajectory. S42. Using point cloud tracking methods, adjust the vehicle's movement within the search range according to the predicted path to ensure that the vehicle can perform a U-turn within a safe range. S43. During point cloud tracking, monitor the environment and obstacle information around the vehicle in real time, and adjust the vehicle's steering wheel, throttle and braking control parameters to ensure the accuracy and safety of the U-turn operation. S44. After the vehicle completes the U-turn, update the vehicle's current position and attitude information, and record the motion trajectory data at this time to ensure that the system can continuously track the vehicle's position and attitude.

[0024] As described above, by utilizing historical datasets for point cloud tracking, the system can accurately predict the vehicle's U-turn path when it is detected approaching a U-turn location, ensuring the vehicle completes the U-turn within a safe range. This method not only combines the vehicle's real-time trajectory but also monitors surrounding obstacles in real time, dynamically adjusting the path to enable the vehicle to make a smooth U-turn in confined spaces. Through precise control of point cloud tracking, the system reduces U-turn failures caused by space constraints, improving the success rate of vehicle U-turn operations.

[0025] Specifically, please refer to Figure 2 The specific process for recording the motion trajectory in step S44 is as follows: 1) Start Operation Operate the vehicle (manually) to enter the section of road where a U-turn is required, and ensure that the vehicle is driving smoothly. Begin recording the U-turn process.

[0026] 2) Determine if a U-turn operation can be completed. Based on the vehicle's current position and the surrounding environment, determine if there is enough space to make a U-turn. If there is enough space, proceed to the next step; if there is not enough space, stop the operation and exit.

[0027] 3) Path data storage If the path space meets the conditions, the system stores the corresponding point cloud data along the vehicle's driving path at 0.1-meter intervals, including the vehicle's coordinates, orientation status, and position data.

[0028] 4) Determine whether the vehicle can continue forward. Determine if the vehicle can proceed smoothly to the U-turn position. If there is an obstacle or insufficient space ahead, the vehicle will pause and return to the previous position; if there is enough space ahead, the U-turn will be executed.

[0029] 5) Determine if there is sufficient space at the rear of the vehicle. Determine if there is enough space at the rear of the vehicle to make a U-turn. If there is space, proceed with the U-turn. If there is insufficient space at the rear, adjust the vehicle's position to ensure there is enough space for the U-turn.

[0030] 6) Execution of route adjustment and U-turn After confirming that there is sufficient space at the rear of the vehicle, the system automatically adjusts the steering wheel and performs a U-turn, with the vehicle turning around according to the predetermined path.

[0031] 7) Complete the U-turn operation and record it. After completing the U-turn operation, the system records the complete path and related information for future reference and optimization. Once the operation is finished, the system exits the U-turn process and completes the recording.

[0032] Preferably, please refer to Figure 3 The steps for point cloud tracing in step S42 are as follows: 1) Activate the vehicle U-turn loop function Activate the vehicle's automatic U-turn system to ensure that the system can determine the route and perform the U-turn operation based on the current road conditions and the vehicle's position.

[0033] 2) Choose a suitable starting point for the loop. The system iterates through all available road segments and selects the valid starting point closest to the vehicle's current location as the first loop point for the U-turn.

[0034] 3) Determine if the loop point is valid. Based on the selected starting point, the system determines whether the point is suitable for continuing the operation. If it is valid, the system proceeds to the next step; if it is invalid, the current operation ends.

[0035] 4) The vehicle moves forward, using the front wheel alignment mode in a loop. After confirming the loop point is valid, the vehicle begins to move forward and performs a U-turn using the front wheel alignment mode. At this time, the vehicle will automatically adjust according to the set path.

[0036] 5) Determine if the correct cycle phase has been entered. The system determines whether the vehicle has entered the correct U-turn cycle. If it has, the system continues the U-turn operation; if it has not, the system adjusts the route.

[0037] 6) Determine if the last loop point is valid. As the vehicle continues moving forward, the system checks the validity of the last loop point. If valid, execution continues; if invalid, a path switch is required to select the next suitable loop point.

[0038] 7) Loop ends Once the system confirms the U-turn operation is complete, it determines whether to end the loop. If the loop is complete, the operation is successful, and the U-turn ends.

[0039] 8) Switch to the next search point If the current loop point is invalid or the turn fails, the system switches to the next search point and continues the loop operation until a valid loop point is found and the turn is completed.

[0040] 9) End operation After successfully completing the U-turn operation, the loop ends, the system returns to the initial state, and prepares for the next operation.

[0041] Please refer to Figure 4 A terminal 1 for realizing U-turn of an unmanned vehicle based on point cloud includes a memory 2, a processor 3, and a computer program stored in the memory and executable on the processor 3. When the processor 3 executes the computer program, it completes the steps in a method for realizing U-turn of an unmanned vehicle based on point cloud.

[0042] In summary, this invention provides a method and terminal for enabling autonomous vehicles to turn around based on point clouds. By fusing high-precision point cloud data, vehicle motion information, historical data, and point cloud tracking technology, it significantly improves the accuracy and efficiency of vehicle turning around in complex and narrow environments. Specifically, this invention exhibits significant beneficial effects in the following aspects: This invention utilizes high-precision 3D point cloud data provided by LiDAR (Light Detection and Ranging) and combines it with the vehicle's real-time motion information, attitude information, and historical data to enable the vehicle to perform efficient U-turns in narrow or obstacle-filled environments. Through point cloud matching and point cloud tracking technology, the vehicle's trajectory can be accurately predicted and controlled, thereby reducing operational delays and inefficiencies caused by frequent position adjustments. Traditional autonomous driving systems typically require frequent human intervention, especially when making U-turns in confined spaces. This invention, however, utilizes precise point cloud data for perception and path prediction to automatically identify and adjust the vehicle's position and attitude, reducing reliance on human intervention and enabling the vehicle to automatically and efficiently complete U-turns in complex environments. By introducing the Normal Distribution Transform (NDT) algorithm to optimize point cloud data, the vehicle's localization and attitude information can still be accurately acquired even in complex environments or with significant occlusion. Combined with vehicle motion information (such as IMU and odometer data), the system can dynamically adjust the search range and optimize point cloud matching, improving matching accuracy and system real-time performance. By combining the Iterative Closest Point (ICP) algorithm with vehicle motion information, the point cloud matching process in dynamic environments is optimized through continuous iterative adjustments. Especially when vehicle motion changes rapidly, the introduction of IMU data allows for precise correction of rotation and translation calculations, thereby improving the stability and robustness of point cloud matching and reducing errors caused by environmental changes. This invention provides an intelligent decision-making process for path selection and loop judgment, automatically identifying and selecting appropriate loop points, and executing loop operations until a successful U-turn is achieved. The system can automatically switch to the next search point based on real-time feedback and adjust the path in case of failure, minimizing the risk of U-turn failure and improving the adaptability of the autonomous driving system in confined spaces.

[0043] The above description is merely an embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent modifications made based on the content of the present invention specification and drawings, or direct or indirect applications in related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A method for enabling autonomous vehicles to turn around based on point clouds, characterized in that: Including the following steps: S1. Acquire point cloud map data and generate an initial map; S2. Acquire point cloud data of the current environment and match it with the initial map, and determine the current position and attitude information of the vehicle based on the vehicle's motion information; S3. Record the vehicle's continuous position information in time sequence, and obtain the vehicle's motion trajectory based on the continuous position information; S4. Based on historical datasets, when a vehicle is detected to be approaching a U-turn location, a U-turn is achieved using point cloud tracking, specifically including: S41. When the vehicle is detected to be about to reach the U-turn position, predict the vehicle's path and attitude changes based on historical datasets and the vehicle's trajectory. S42. Using point cloud tracking methods, adjust the vehicle's movement within the search range according to the predicted path to ensure that the vehicle can perform a U-turn within a safe range. S43. During point cloud tracking, monitor the environment and obstacle information around the vehicle in real time, and adjust the vehicle's steering wheel, throttle and braking control parameters to ensure the accuracy and safety of the U-turn operation. S44. After the vehicle completes the U-turn, update the vehicle's current position and attitude information, and record the motion trajectory data at this time.

2. The method for realizing the U-turn of an unmanned vehicle based on point cloud as described in claim 1, characterized in that: Step S2 specifically includes the following steps: S21. Optimize the point cloud data of the current environment based on the normal distribution transformation algorithm; S22. Match the optimized point cloud data of the current environment with the initial map, and estimate the initial position and attitude information of the generated vehicle. S23. Based on the vehicle's motion information and the estimated initial position and attitude information, determine the vehicle's current position and attitude information.

3. The method for realizing the U-turn of an unmanned vehicle based on point cloud as described in claim 2, characterized in that: Step S23 specifically includes the following steps: S231. Using the vehicle's motion information and the estimated initial position and attitude information, predict the vehicle's current position and attitude information; S232. Determine the search range based on the predicted current position and attitude information; S233. Within the search range, point cloud matching is performed using the iterative nearest point algorithm, and the current position and attitude information of the vehicle are determined based on the matching results.

4. The method for realizing the U-turn of an unmanned vehicle based on point cloud as described in claim 3, characterized in that: Step S233 specifically includes the following steps: Using the vehicle's motion information as the initial estimation data and auxiliary constraints for the iterative nearest point algorithm, the source point cloud data of the current environment and the target point cloud data of the initial gradient are read. The iterative nearest point algorithm is used to perform point cloud matching, and the current position and attitude information of the vehicle are output based on the matching results.

5. A terminal for realizing U-turns of unmanned vehicles based on point clouds, characterized in that: Includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, performs the following steps: S1. Acquire point cloud map data and generate an initial map; S2. Acquire point cloud data of the current environment and match it with the initial map, and determine the current position and attitude information of the vehicle based on the vehicle's motion information; S3. Record the vehicle's continuous position information in time sequence, and obtain the vehicle's motion trajectory based on the continuous position information; S4. Based on historical datasets, when a vehicle is detected to be approaching a U-turn location, a U-turn is achieved using point cloud tracking, specifically including: S41. When the vehicle is detected to be about to reach the U-turn position, predict the vehicle's path and attitude changes based on historical datasets and the vehicle's trajectory. S42. Using point cloud tracking methods, adjust the vehicle's movement within the search range according to the predicted path to ensure that the vehicle can perform a U-turn within a safe range. S43. During point cloud tracking, monitor the environment and obstacle information around the vehicle in real time, and adjust the vehicle's steering wheel, throttle and braking control parameters to ensure the accuracy and safety of the U-turn operation. S44. After the vehicle completes the U-turn, update the vehicle's current position and attitude information, and record the motion trajectory data at this time.

6. A terminal for realizing U-turn of an unmanned vehicle based on point cloud as described in claim 5, characterized in that: Step S2 specifically includes the following steps: S21. Optimize the point cloud data of the current environment based on the normal distribution transformation algorithm; S22. Match the optimized point cloud data of the current environment with the initial map, and estimate the initial position and attitude information of the generated vehicle. S23. Based on the vehicle's motion information and the estimated initial position and attitude information, determine the vehicle's current position and attitude information.

7. A terminal for realizing U-turn of an unmanned vehicle based on point cloud as described in claim 6, characterized in that: Step S23 specifically includes the following steps: S231. Using the vehicle's motion information and the estimated initial position and attitude information, predict the vehicle's current position and attitude information; S232. Determine the search range based on the predicted current location and attitude information; S233. Within the search range, point cloud matching is performed using the iterative nearest point algorithm, and the current position and attitude information of the vehicle are determined based on the matching results.

8. A terminal for realizing U-turn of an unmanned vehicle based on point cloud as described in claim 7, characterized in that: Step S233 specifically includes the following steps: Using the vehicle's motion information as the initial estimation data and auxiliary constraints for the iterative nearest point algorithm, the source point cloud data of the current environment and the target point cloud data of the initial gradient are read. The iterative nearest point algorithm is used to perform point cloud matching, and the current position and attitude information of the vehicle are output based on the matching results.

Images