A path planning method for narrow scenes and related devices

Abstract

The present disclosure provides a path planning method for narrow scenarios and related equipment. The method comprises: analyzing an initial planning path based on point cloud map data, determining narrow road sections and non-narrow road sections in the initial planning path; determining a non-narrow road target path and corresponding non-narrow road target speed data of the non-narrow road section based on a first initial sampling point of the non-narrow road section and preset speed data; and determining a narrow road target path and corresponding narrow road target speed that satisfy a preset condition based on a second initial sampling point of the narrow road section; obtaining a target path based on the non-narrow road target path and the narrow road target path, and driving on the target path based on the non-narrow road target speed data and the narrow road target speed.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to the field of path planning, and in particular to a path planning method for narrow scenarios and related equipment. BACKGROUND

[0002] Intelligent inspection technology has been widely applied in many fields. By integrating advanced visual recognition, sensor technology and navigation algorithms, it realizes effective inspection and monitoring of various scenes. This technology is crucial for robots that need to perform automated work, such as in power facilities, warehouse logistics, industrial production, etc. Intelligent inspection robots can perform periodic inspections, fault detection, safety monitoring, etc., thereby improving work efficiency and reducing labor costs.

[0003] However, in some specific narrow scenarios, existing intelligent inspection methods face many challenges. First, existing technical solutions usually do not have differentiated strategies for narrow scenarios and non-narrow scenarios. In narrow environments, robots need to move flexibly in limited space, while existing path planning algorithms often fail to fully consider the needs of such special environments. Second, in narrow scenarios, existing methods often only focus on path planning, ignoring the impact of observation errors and control errors. Due to the limited space, even a small error can cause the robot to collide with surrounding objects, affecting the safety and effectiveness of the inspection. In addition, the grid map used by existing methods has the problem of large storage overhead. In large-scale scenarios, if a high-resolution of centimeters is used to represent the map environment in detail, it will occupy a large amount of storage resources.

[0004] Therefore, there are problems such as lack of robustness and adaptability, and large map storage overhead in existing path planning techniques. SUMMARY

[0005] The present disclosure proposes a path planning method for narrow scenarios and related equipment to solve the technical problems of lack of robustness and adaptability, and large map storage overhead in some path planning methods to some extent.

[0006] In a first aspect, the present disclosure provides a path planning method for narrow scenarios, comprising:

[0007] analyzing an initial planning path based on point cloud map data to determine narrow road segments and non-narrow road segments in the initial planning path;

[0008] determining a non-narrow path target and corresponding non-narrow target speed data for the non-narrow road segment based on the first initial sampling point of the non-narrow road segment and the preset speed data; and

[0009] determine a narrow-path target path and a corresponding narrow-path target speed that satisfy a preset condition based on the second initial sampling points of the narrow road section;

[0010] obtain a target path based on the non-narrow-path target path and the narrow-path target path, and drive on the target path based on the non-narrow-path target speed data and the narrow-path target speed.

[0011] In a second aspect, the present disclosure provides a path planning device for a narrow scene, comprising:

[0012] a narrow-path identification module configured to analyze an initial planning path based on point cloud map data, and determine a narrow road section and a non-narrow road section in the initial planning path;

[0013] a non-narrow-path planning module configured to determine a non-narrow-path target path and corresponding non-narrow-path target speed data of the non-narrow road section based on first initial sampling points of the non-narrow road section and preset speed data; and

[0014] a narrow-path planning module configured to determine a narrow-path target path and a corresponding narrow-path target speed that satisfy a preset condition based on second initial sampling points of the narrow road section;

[0015] a target path module configured to obtain a target path based on the non-narrow-path target path and the narrow-path target path;

[0016] a control module configured to drive on the target path based on the non-narrow-path target speed data and the narrow-path target speed.

[0017] In a third aspect, the present disclosure provides an electronic device, comprising one or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and executed by the one or more processors, and the programs comprise instructions for executing the method according to the first aspect.

[0018] In a fourth aspect, the present disclosure provides a non-volatile computer readable storage medium containing a computer program, which, when executed by one or more processors, causes the processors to execute the method according to the first aspect.

[0019] In a fifth aspect, the present disclosure provides a computer program product, comprising computer program instructions, which, when executed on a computer, cause the computer to execute the method according to the first aspect.

[0020] As can be seen from the above, the path planning method for narrow scenes and the related device provided by the present disclosure, the adaptive path planning algorithm proposed performs more robust effect in narrow lane environment, and can realize efficient path planning in non-narrow lane environment, so that the inspection system can ensure safety in narrow lane environment and maintain high efficiency in the whole inspection process. In addition, through efficient map storage management, high-resolution map acquisition is realized with small storage overhead, so that the system can adapt to the inspection demand in a large range of scenes, further improving the practicability and expansibility of the system. BRIEF DESCRIPTION OF DRAWINGS

[0021] In order to more clearly illustrate the technical solutions in the present disclosure or the related art, the drawings needed to be used in the embodiments or the related art description will be briefly introduced. Obviously, the drawings in the following description are only embodiments of the present disclosure, and other drawings can be obtained by those skilled in the art without creative labor.

[0022] Figure 1 The schematic diagram of the path planning architecture for narrow scenes of the embodiment of the present disclosure.

[0023] Figure 2 The hardware structure schematic diagram of the exemplary electronic device of the embodiment of the present disclosure.

[0024] Figure 3 The flowchart of the path planning method for narrow scenes of the embodiment of the present disclosure.

[0025] Figure 4 The schematic diagram of the path planning method for narrow scenes of the embodiment of the present disclosure.

[0026] Figure 5 The schematic diagram of the path planning in narrow lane mode of the embodiment of the present disclosure.

[0027] Figure 6 The schematic diagram of the path planning of the A*(in-circle expansion) method.

[0028] Figure 7 The schematic diagram of the path planning of the A*(out-circle expansion) method.

[0029] Figure 8 The schematic diagram of the path planning of the hybrid A* method.

[0030] Figure 9 The schematic diagram of the path planning of the adaptive hybrid A* method of the embodiment of the present disclosure.

[0031] Figure 10 The schematic diagram of the path planning device for narrow scenes of the embodiment of the present disclosure. Detailed Implementation

[0032] To make the objectives, technical solutions, and advantages of this disclosure clearer, the following detailed description is provided in conjunction with specific embodiments and the accompanying drawings.

[0033] It should be noted that, unless otherwise defined, the technical or scientific terms used in the embodiments of this disclosure should have the ordinary meaning understood by one of ordinary skill in the art to which this disclosure pertains. The terms "first," "second," and similar terms used in the embodiments of this disclosure do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0034] It is understood that before using the technical solutions disclosed in the various embodiments of this disclosure, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in this disclosure in an appropriate manner in accordance with relevant laws and regulations, and user authorization should be obtained.

[0035] For example, upon receiving a user's active request, a prompt message is sent to the user to explicitly inform them that the requested operation will require the acquisition and use of the user's personal information. This allows the user to independently choose whether to provide personal information to the software or hardware, such as the electronic device, application, server, or storage medium performing the operations of this disclosed technical solution, based on the prompt message.

[0036] It is understood that the above notification and user authorization process are merely illustrative and do not constitute a limitation on the implementation of this disclosure. Other methods that comply with relevant laws and regulations may also be applied to the implementation of this disclosure.

[0037] Figure 1 A schematic diagram of a path planning architecture for narrow scenarios according to an embodiment of this disclosure is shown. (Reference) Figure 1The path planning architecture 100 for narrow scenarios may include a server 110, a terminal 120, and a network 130 providing a communication link. The server 110 and the terminal 120 can be connected via a wired or wireless network 130. The server 110 can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, security services, and CDN.

[0038] Terminal 120 can be implemented in hardware or software. For example, when terminal 120 is implemented in hardware, it can be various electronic devices with a display screen and support page display, including but not limited to smartphones, tablets, e-book readers, laptops, and desktop computers. When terminal 120 is implemented in software, it can be installed in the electronic devices listed above; it can be implemented as multiple software programs or software modules (e.g., software programs or software modules used to provide distributed services) or as a single software program or software module, without specific limitations.

[0039] It should be noted that the path planning method for narrow scenarios provided in this application embodiment can be executed by the terminal 120 or by the server 110. It should be understood that... Figure 1 The number of terminals, networks, and servers shown is for illustrative purposes only and is not intended to be a limitation. Any number of terminals, networks, and servers can be used depending on implementation needs.

[0040] Figure 2 A schematic diagram of the hardware structure of an exemplary electronic device 200 provided in an embodiment of this disclosure is shown. For example... Figure 2 As shown, the electronic device 200 may include: a processor 202, a memory 204, a network module 206, a peripheral interface 208, and a bus 210. The processor 202, memory 204, network module 206, and peripheral interface 208 are interconnected within the electronic device 200 via the bus 210.

[0041] Processor 202 may be a Central Processing Unit (CPU), a Neural Processing Unit (NPU), a Microcontroller (MCU), a programmable logic device, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), or one or more integrated circuits. Processor 202 can be used to perform functions related to the techniques described in this disclosure. In some embodiments, processor 202 may also include multiple processors integrated as a single logic component. For example, such as... Figure 2 As shown, processor 202 may include multiple processors 202a, 202b and 202c.

[0042] Memory 204 can be configured to store data (e.g., instructions, computer code, etc.). Figure 2 As shown, the data stored in memory 204 may include program instructions (e.g., program instructions for implementing a path planning method for narrow scenarios according to embodiments of this disclosure) and data to be processed (e.g., the memory may store configuration files of other modules, etc.). Processor 202 may also access the program instructions and data stored in memory 204 and execute the program instructions to operate on the data to be processed. Memory 204 may include volatile or non-volatile storage devices. In some embodiments, memory 204 may include random access memory (RAM), read-only memory (ROM), optical disk, magnetic disk, hard disk, solid-state drive (SSD), flash memory, memory stick, etc.

[0043] Network module 206 can be configured to provide communication with other external devices to electronic device 200 via a network. This network can be any wired or wireless network capable of transmitting and receiving data. For example, the network can be a wired network, a local wireless network (e.g., Bluetooth, WiFi, Near Field Communication (NFC), etc.), a cellular network, the Internet, or a combination thereof. It is understood that the type of network is not limited to the specific examples described above. In some embodiments, network module 206 may include any combination of any number of network interface controllers (NICs), radio frequency modules, transceivers, modems, routers, gateways, adapters, cellular network chips, etc.

[0044] The peripheral interface 208 can be configured to connect the electronic device 200 to one or more peripheral devices to enable information input and output. For example, peripheral devices may include input devices such as keyboards, mice, touchpads, touch screens, microphones, and various sensors, as well as output devices such as displays, speakers, vibrators, and indicator lights.

[0045] Bus 210 can be configured to transfer information between various components of electronic device 200 (e.g., processor 202, memory 204, network module 206, and peripheral interface 208), such as internal buses (e.g., processor-memory bus), external buses (USB port, PCI-E bus), etc.

[0046] It should be noted that although the architecture of the above-described electronic device 200 only shows the processor 202, memory 204, network module 206, peripheral interface 208, and bus 210, in specific implementations, the architecture of the electronic device 200 may also include other components necessary for normal operation. Furthermore, those skilled in the art will understand that the architecture of the above-described electronic device 200 may only include the components necessary for implementing the embodiments of this disclosure, and does not necessarily include all the components shown in the figures.

[0047] Currently, some path planning methods for intelligent inspection technologies employ an improved Dijkstra algorithm to generate a guiding path from the starting pose to the target pose that is as far away from obstacles as possible or as close to the center of a narrow passage. Quadratic programming is used to smooth the initial guiding path. Inferential sampling is used to detect whether the robot will collide with fixed obstacles while following the global path, and collision-prone sections are marked as requiring replanning. Gradient descent is then used to adjust these replanning sections. This invention uses an improved Dijkstra algorithm to generate a guiding path from the starting pose to the target pose that is as far away from obstacles as possible or as close to the center of a narrow passage. Quadratic programming is used to smooth the initial guiding path. Using the proposed algorithm, a smooth global path that is far from obstacles can be generated in narrow spaces without setting an expansion radius, meeting the navigation requirements of subway inspection robots.

[0048] Some path planning methods involve configuring robot mapping particles on road mapping lines in a virtual scene reference map, determining the first travel speed of the robot mapping particles on the road mapping lines based on the actual operating speed of the inspection robot, recording the power equipment nodes passed by the test trajectory within a preset time period, and determining the total inspection parameters of the test trajectory based on the inspection parameter change curve associated with each power equipment node. Based on the total inspection parameters of each test trajectory, the test trajectories are sorted, and the top-ranked test trajectories are identified as the optimal robot paths. The technical solution disclosed in this invention realizes the automatic generation of robot travel trajectories based on the operating status of power equipment, thereby improving the efficiency of effective inspection of power equipment.

[0049] Some path planning methods first perform EKF filtering on the encoder and IMU data to obtain high-precision odometer data before fusing it into the SLAM module. Separate coordinate alignment and coordinate transformation modules are constructed. In the coordinate alignment module, the SLAM output position and RTK output are nonlinearly optimized to obtain their coordinate transformation matrix. This transformation matrix is ​​then loaded into the coordinate transformation module to transform the RTK-GNSS data before it is incorporated into the SLAM output fusion. Furthermore, there is a high-precision positioning system based on multi-sensor fusion. This invention uses multiple sensors for data fusion, leveraging the advantages of different sensors. In environments where one sensor is disadvantageous, data from other sensors is used to complement it, ensuring positioning accuracy while effectively improving positioning stability and reducing safety incidents.

[0050] These intelligent inspection methods often use the same path planning algorithm, but the inspection objectives differ for narrow and non-narrow passage scenarios. In narrow passages, safety is paramount, while in non-narrow passages, efficiency is more important. A single strategy often cannot simultaneously address both safety and efficiency in different scenarios. For example, in a wide warehouse, quickly completing the inspection task is the primary goal, while avoiding collisions is more crucial in narrow corridors or passageways. Alternatively, existing technologies may only perform path planning for narrow passage scenarios without considering the tracking control issues. Since most planning algorithms treat the Automated Guided Vehicle (AGV) as a circular object, the collision radius is typically set to the radius of the vehicle's inscribed circle to enable the AGV to plan a path in narrow passages. However, in actual operation, collisions may occur due to control errors, observation noise, and other factors. This indicates that existing methods are insufficient in terms of safety in narrow passage environments. Traditional methods typically store global maps using grid maps. For narrow passage scenarios, centimeter-level high-resolution maps are required to ensure accurate navigation. However, using such high-resolution maps in large-scale scenarios leads to a significant increase in storage overhead. Therefore, how to achieve efficient map storage that meets the needs of narrow-lane environments in large-scale scenarios has become an urgent problem to be solved. Existing technologies often struggle to balance high resolution and storage efficiency when processing large-scale maps.

[0051] Therefore, how to ensure map accuracy while reducing storage overhead has become an urgent technical problem to be solved.

[0052] In view of this, the embodiments of this disclosure provide a path planning method and related equipment for narrow scenarios. The proposed adaptive path planning algorithm exhibits more robust performance in narrow-path environments, while achieving efficient path planning in non-narrow-path environments. This allows the inspection system to ensure safety in narrow-path environments while maintaining high efficiency throughout the entire inspection process. Furthermore, through efficient map storage management, high-resolution map acquisition is achieved with minimal storage overhead, enabling the system to adapt to inspection needs in a wide range of scenarios, further improving the system's practicality and scalability.

[0053] See Figure 3 , Figure 3 A schematic flowchart of a path planning method for narrow scenarios according to an embodiment of the present disclosure is shown. The path planning method for narrow scenarios according to an embodiment of the present disclosure can be deployed on a terminal. Figure 3 In the path planning method 300 for narrow scenarios, the following steps may be further included.

[0054] In step S310, the initial planned path is analyzed based on point cloud map data to determine narrow and non-narrow road segments within the initial planned path. Point cloud map data refers to a dataset of three-dimensional points reflecting the target area's environment and terrain, which can be obtained from the environment using methods such as laser scanning and stereo vision. The initial planned path refers to the expected travel route from the starting point to the ending point calculated from the map data based on path algorithms (e.g., A*, Dijkstra's algorithm). Narrow road identification can be performed on the initial path planning, detecting obstacles on both sides of the path to determine if it is within a narrow road. If a non-narrow road is detected, an efficient planning strategy is adopted, optimizing the path using B-splines and generating control commands using DWA (Dynamic Window Method). If a narrow road is detected, the path is replanned, prioritizing safety and reducing error impact through multiple observations and planning. The control unit selects local target points based on path characteristics and generates control commands to ensure the robot's motion path closely follows the planned trajectory, such as... Figure 4 As shown, Figure 4 A schematic diagram of path planning for narrow scenarios according to an embodiment of the present disclosure is shown.

[0055] In some embodiments, the initial planned path is analyzed based on point cloud map data to determine narrow road segments and non-narrow road segments in the initial planned path, including:

[0056] The initial planned path is sampled uniformly to obtain initial sampling points;

[0057] Determine the nearest valuable raster points on both sides of the path direction for the initial sampling point in the point cloud map;

[0058] In response to the fact that the distance between the valuable grid points is less than a preset distance, the initial sampling point is determined as the first candidate narrow channel sampling point;

[0059] Traverse all the initial sampling points and determine a preset number of sampling points adjacent to the first candidate narrow channel sampling point as the second candidate narrow channel sampling point;

[0060] The narrow road segment is obtained based on the first candidate narrow road sampling point and the second candidate narrow road sampling point, and the non-narrow road segment in the initial planned path is determined as the non-narrow road segment.

[0061] Uniform path sampling refers to selecting a series of points at fixed intervals or proportions along the initially planned path as sampling points for detailed path analysis. Valued grid points are grid points in map data assigned specific values ​​(such as obstacle height, occupancy status, etc.) that represent obstacles or other important features in the environment. Preset distances can be used to determine whether a location is a narrow passage. For example, if the distance between two valued grid points is less than this preset distance, the area between them may be considered a narrow passage.

[0062] Specifically, an initial path can be generated using the A* algorithm, and the input preliminary planned path is sampled uniformly. To determine if a sampled point is a narrow passage, the sampled points are traversed, and the nearest valuable grid points on both sides of the path direction for each sampled point are selected in the expanded map, with the distance between the two grid points calculated. If the distance is less than a threshold, the sampled point is considered a narrow passage; otherwise, it is considered a non-narrow passage. Narrow passage completion is then performed: the sampled points are traversed, and if a sampled point exists within the n preceding and following sampled points, it is marked as a candidate narrow passage point. After traversal, all candidate narrow passage points are marked as narrow passage points. Trajectory cutting points are calculated: the sampled points are traversed, and the boundaries between narrow and non-narrow passage points are marked as trajectory cutting points. The trajectory is divided into narrow and non-narrow passage segments based on these cutting points. Further trajectory optimization is then performed on the narrow and non-narrow passage segments respectively.

[0063] For example, an initial planned path from the starting point to the ending point is calculated using an algorithm. This initial planned path can be uniformly sampled, for example, by selecting a point every meter or at certain percentages of the path length. For each sampling point, the nearest valuable grid points on both sides of that point along the path direction can be found (these grid points may represent path boundaries, walls, buildings, or other obstacles). If the distance between the valuable grid points on both sides of a sampling point is less than a preset narrowway threshold (e.g., 3 meters), then this sampling point is considered a first candidate narrowway sampling point. All initial sampling points are iterated through, and a preset number (e.g., 3) of the sampling points immediately before and after those marked as first candidate narrowway sampling points are also considered second candidate narrowway sampling points to ensure the continuity and robustness of narrowway segment identification. Based on all first and second candidate narrowway sampling points, the narrowway segments can be determined. Path portions not marked as narrowway sampling points are considered non-narrowway segments.

[0064] In step S320, the non-narrow road target path and corresponding non-narrow road target speed data for the non-narrow road segment are determined based on the first initial sampling point and preset speed data of the non-narrow road segment. Specifically, a reasonable speed value is assigned to each path segment according to the preset speed data and path characteristics to avoid sudden speed changes and improve driving stability and safety.

[0065] In some embodiments, determining the non-narrow road target path and corresponding non-narrow road target speed data of the non-narrow road segment based on a first initial sampling point and preset speed data of the non-narrow road segment includes:

[0066] The first initial sampling point is subjected to a first smoothing process to obtain a first smooth non-narrow road segment;

[0067] The first smoothed non-narrow road segment is subjected to a second smoothing process based on the smoothing cost function to obtain the second smoothed non-narrow road segment.

[0068] A sampling trajectory is generated for the preset speed data and the second smooth non-narrow road segment, and the trajectory cost function of the sampling trajectory is obtained;

[0069] The sampled trajectory corresponding to the minimum value of the trajectory cost function is determined as the non-narrow-path target path, and the preset speed data corresponding to the minimum value of the trajectory cost function is the non-narrow-path target speed data.

[0070] The first initial sampling point can refer to the initial set of sampling points obtained by sampling a uniform path on a non-narrow road segment. The preset speed data can be pre-set speed data related to the path characteristics. For example, a higher speed might be set on straight sections, and a lower speed on curves or complex sections. The first smoothing process can refer to the initial smoothing of the sampling points to reduce path fluctuations and irregularities, making the path smoother and more continuous. This can be achieved, for example, through polynomial fitting, Bézier curves, or Kalman filtering. The smoothing cost function can refer to a function used to evaluate the smoothness of the path. This function typically considers factors such as path curvature, length, and distance to obstacles, and provides a corresponding cost. The second smoothing process can refer to further optimization and smoothing of the path based on the smoothing cost function to find the path with the minimum cost, making it more suitable for actual driving needs. This may include adjusting parameters such as path curvature and length to ensure path continuity and feasibility. The sampling trajectory can refer to a series of possible driving trajectories generated on the smoothed path based on the preset speed data. These trajectories consider factors such as path shape and speed variations. The trajectory cost function refers to a function used to evaluate the quality of a sampled trajectory. This function may consider multiple factors such as trajectory length, travel time, smoothness of speed changes, and distance to obstacles. By combining the smoothed path and the matched speed data, a non-narrow road target path is generated for the non-narrow road segment. By evaluating indicators such as the smoothness, continuity, and reasonableness of the speed data, the final non-narrow road target path and the corresponding non-narrow road target speed data are determined.

[0071] In some embodiments, the first initial sampling point is subjected to a first smoothing process to obtain a first smooth non-narrow road segment, including:

[0072] Starting from the first starting point of the first initial sampling point, determine whether there is a straight line among the first and last sampling points of the three adjacent first initial sampling points;

[0073] If a straight line exists between the first and last sampling points of three adjacent first initial sampling points, the middle sampling point among the three adjacent first initial sampling points is removed.

[0074] By traversing the first initial sampling points, the first smooth non-narrow road segment is obtained.

[0075] In the first smoothing process of the first initial sampling point, the goal is to reduce unnecessary fluctuations in the path while preserving its basic shape and features. This can be achieved using a simplified method based on line detection. Line detection involves starting from the first starting point of the first initial sampling point and sequentially checking three adjacent sampling points (denoted as A, B, and C). It determines whether a straight line relationship exists between the first and last sampling points A and C. This straight line relationship can be approximated by calculating the slopes of points A, B, and C or using methods such as the cross product of vectors. If the angle between the vector from A to B and the vector from B to C is very small (close to 0 degrees or 180 degrees), or their slopes are almost equal, then points A, B, and C can be considered approximately collinear. If a straight line relationship exists between the first and last sampling points A and C of the three adjacent first initial sampling points, then the intermediate point B can be considered redundant because it does not increase the complexity or information content of the path. Therefore, the intermediate sampling point B is removed, retaining only the first and last sampling points A and C. This simplifies the path and reduces the number of unnecessary sampling points. The above steps are repeated to traverse the entire set of first initial sampling points. Each time a straight line relationship is detected, intermediate sampling points are removed. This results in a simplified and smoothed non-narrow road segment, known as the first smoothed non-narrow road segment. This first smoothed non-narrow road segment consists of fewer sampling points but still retains the basic shape and features of the original path (such as turning points and key points), ensuring the accuracy and feasibility of the path. By removing unnecessary intermediate sampling points, the amount of data required for path representation is significantly reduced, improving the performance of real-time path planning. The smoother path, due to the removal of highly volatile intermediate points, helps reduce instability during driving.

[0076] Specifically, the main goal of path planning in non-narrow lane mode is to generate more efficient planned trajectories and to generate control commands more quickly, which mainly includes the following steps:

[0077] Step 1: Perform reverse smoothing on the initial planned path to make the trajectory smoother. First, initialize the iterators and use three iterators, it1, it2, and it3, to traverse the path points. it1 represents the starting point of the current check point, and it2 and it3 are the two subsequent points, respectively. Check if there is a straight line of sight between path points, traversing the path points until it3 reaches the end of the path. During the traversal, check if there is a straight line of sight between it1 and it3. If there is, remove point it2, as it2 is redundant. If there is no straight line of sight, move all three iterators to the next point and continue checking. Continue the loop until it3 reaches the end of the path, and the path smoothing operation is complete.

[0078] Step 2: Using the B-spline optimization algorithm, the trajectory is optimized into a smooth and efficient path that conforms to kinematics. The cost function of B-spline optimization is:

[0079] f=λ1f s +λ2f c +λ3(f v +f w (1)

[0080] Among them, f s For the cost of smoothness, f c For the cost of collision, f v and f w For the soft constraints of linear velocity and angular velocity, λ1, λ2, and λ3 are the weights of each cost, respectively;

[0081] Step 3: Using the DWA algorithm, generate sampling trajectories for different selectable angular velocities and linear velocities, evaluate the cost of each sampling trajectory, and evaluate the cost function of the sampling trajectory as: G(v, w)=α·head(v, w)+β·dist(v, w)+γ·velocity(v, w)+δ·path(v, w) (2).

[0082] In this framework, head(v, w) is the azimuth evaluation function, which is the angle difference between the final pose of the sampled trajectory and the target. dist(v, w) is the safety evaluation function, which is the distance between the sampled trajectory and the nearest obstacle. velocity(v, w) evaluates the trajectory velocity, and path(v, w) evaluates the degree of overlap between the sampled trajectory and the subsequent direction of the planned trajectory, ensuring that the final orientation of the selected sampled trajectory better matches the subsequent orientation of the global trajectory. α, β, γ, and δ are the weights of each cost. After calculating the cost of all sampled trajectories, the optimal path with the minimum cost is selected, and the corresponding linear and angular velocities are generated.

[0083] In step S330, the target path and the corresponding target speed of the narrow road section that meet the preset conditions are determined based on the second initial sampling point of the narrow road section.

[0084] For the second initial sampling point on the narrow road segment, an adaptive evaluation function can be used to generate one or more possible adaptive trajectories, which strike a balance between safety, efficiency, and comfort. The adaptive trajectory can consider factors such as the planned distance, the predicted distance to the target, the number of nearby obstacles, the cost of in-place rotation, and the cost of zero speed. By combining the adaptive evaluation function and the local navigation evaluation function, we can generate optimal path and speed planning based on the specific conditions of the narrow road segment and the vehicle's performance characteristics.

[0085] In some embodiments, determining a target path and a corresponding target speed for a narrow road segment that meet preset conditions based on a second initial sampling point of the narrow road segment includes:

[0086] An adaptive trajectory is generated for the narrow road segment based on an adaptive evaluation function; wherein, the adaptive evaluation function f(n) = g(n) + h(n) + s(n) + r(n) + z(n), where g(n) is the planned distance, h(n) is the predicted distance to the target, s(n) is the number of nearby obstacles, r(n) is the cost of rotating in place, and z(n) represents the cost of zero speed.

[0087] The local navigation target point is selected in the adaptive trajectory based on the local navigation evaluation function. Where point i For the i-th point on the adaptive trajectory, line 0，n For the line connecting the nth point and the 0th point, dis() is the distance function;

[0088] The second initial sampling point, which is furthest from the line connecting the points whose local navigation evaluation function is less than a preset threshold, is determined as the local target point;

[0089] The corresponding narrow-channel target path and the narrow-channel target speed are generated based on the local navigation target point.

[0090] The planned distance reflects the length of the path already traveled, helping to avoid excessive detours. The predicted distance to the target represents the estimated distance from the current position to the target location, helping to guide the moving object towards the target. The number of nearby obstacles reflects the number of obstacles in the surrounding environment, ensuring path safety. The cost of rotating in place reflects the cost of performing a rotation operation in place, which is usually related to the vehicle's mobility and energy consumption. The zero-speed cost represents the cost of keeping the vehicle or robot stationary at a certain position, which may be related to time efficiency or energy consumption. An optimal local navigation target point is selected from the adaptive trajectory based on a local navigation evaluation function. The local navigation evaluation function considers the distance from each point on the adaptive trajectory to the line connecting it to the starting point, as well as the position of these points relative to the entire trajectory. By combining the adaptive evaluation function and the local navigation evaluation function, optimal path and speed planning can be generated based on the specific conditions of narrow road sections and the performance characteristics of the vehicle. By fully considering factors such as the number of obstacles, the cost of rotating in place, and the zero-speed cost, it helps to ensure path safety and driving efficiency. The i-th point on the adaptive trajectory can represent a candidate point on the trajectory. The line connecting the n-th point and the 0th point can represent the endpoint (or the farthest point currently considered) and the starting point of the trajectory, used to evaluate the offset of the candidate point relative to the entire trajectory. By calculating the local navigation evaluation function value of each candidate point and finding the point that is less than a preset threshold and farthest from the connecting line, the local target point can be determined. This ensures the continuity of the path while minimizing unnecessary offsets and detours. Based on the determined local navigation target points, a specific narrow-road target path can be generated, and a corresponding narrow-road target speed can be assigned to it. Specifically, this may include path smoothing and speed optimization to ensure that the vehicle maintains a smooth and safe driving state when passing through narrow road sections.

[0091] Specifically, the main objectives of path planning in narrow-path mode are to generate safer planned trajectories and to generate more stable control commands, such as... Figure 5 As shown, Figure 5 A schematic diagram of path planning in a narrow passage mode according to an embodiment of the present disclosure is shown. It mainly includes the following steps:

[0092] Step 1: Path planning is performed using the improved A* algorithm for narrow passages to generate a trajectory that considers the robot's shape and kinematic constraints. The algorithm flow is as follows: Figure 2 As shown. The improved adaptive hybrid A* algorithm dynamically alternates between traditional A* and hybrid A* methods to fully utilize the inspection vehicle's in-situ rotation capability, enhancing trajectory planning flexibility. Simultaneously, it increases optimization accuracy and computational efficiency in constrained environments through multi-scale adaptive step sizes. The cost function of the improved adaptive hybrid A* algorithm is defined as:

[0093] f(n)=g(n)+h(n)+s(n)+r(n)+z(n) (3)

[0094] Where g(n) represents the planned distance, h(n) is the distance to the target predicted using traditional A*, s(n) represents the number of nearby obstacles, r(n) represents the cost of rotating in place, and z(n) represents the cost of zero speed.

[0095] Step 2: Prune the search nodes according to h(n) to improve computational efficiency;

[0096] Step 3: If a trajectory that meets the kinematic constraints cannot be found, it means that the narrow passage is impassable. In this case, set up obstacles to block the narrow passage, and then replan to find other feasible solutions until a suitable path is generated.

[0097] Step 4: After generating the path, start the trajectory with a fixed starting point and a distance of n as the execution trajectory, and begin executing the trajectory;

[0098] Step 5: Adaptively select a suitable local navigation target point to better plan control commands that conform to the global path during the local planning phase, thereby improving safety in narrow passage control. The strategy for selecting the local navigation target point involves calculating an evaluation function.

[0099]

[0100] Where, point i Let i be the i-th point on the trajectory, line 0，n The evaluation function represents the line connecting the nth point and the 0th point, and the line connecting the target point and the origin. 0，n The maximum distance to each point before the target point is the largest value of the evaluation function. The smaller the result of this function, the more suitable it is as a local target point. Points with evaluation values ​​less than the threshold and far distances are selected as local target points.

[0101] Step 6: Generate control commands using the DWA algorithm based on the rotated local navigation target point;

[0102] Step 7: Simultaneously continue planning subsequent trajectories, and return to Step 3 when nearing the end of the execution trajectory.

[0103] In step S340, a target path is obtained based on the non-narrow target path and the narrow target path, and the vehicle travels on the target path based on the non-narrow target speed data and the narrow target speed.

[0104] Once the start and end points of the non-narrow-path and narrow-path target paths are clearly defined, specific algorithms or strategies (such as path splicing and smooth transitions) can be used to seamlessly connect these two paths, forming a continuous and smooth target path. Additional smoothing can be applied at the path connection points to ensure that the vehicle does not experience unnecessary bumps or sharp turns due to sudden path changes. For example, parameters such as curvature and speed at the connection points can be adjusted to meet the vehicle's driving characteristics and safety requirements. While integrating the paths, the speed data of the non-narrow-path and narrow-path target paths can be combined. For example, speed data can be smoothly transitioned near path transition points to avoid the impact of sudden speed changes on vehicle stability. Furthermore, reasonable speed control strategies can be formulated based on the specific conditions of the target path and the vehicle's performance characteristics. For example, parameters such as maximum speed, acceleration, and deceleration on different road sections can be adjusted to ensure the vehicle can drive safely and efficiently.

[0105] Compared with existing technologies, the method according to embodiments of this disclosure improves environmental adaptability through intelligent switching between narrow-path and non-narrow-path modes. In narrow-path environments, it ensures the safety of path planning; in non-narrow-path environments, it optimizes inspection efficiency. By adaptively hybridizing the A* and DWA algorithms, it generates more robust and efficient control commands, ensuring stable inspections in various scenarios. Through hash table storage and adaptive resolution map generation, it reduces system storage requirements and achieves low-overhead generation of high-resolution maps, enabling the system to adapt to inspection needs in a wide range of scenarios.

[0106] In some embodiments, the point cloud map is divided into different sub-point clouds based on the point cloud coordinates in the point cloud map;

[0107] The sub-point cloud data is obtained by saving the sub-point cloud data into a hash table.

[0108] In response to receiving a map generation instruction for the current location, the sub-point cloud data corresponding to the sub-point cloud within a preset range of the current location coordinates is obtained from the hash table based on the current location coordinates.

[0109] Based on the resolution corresponding to the current location, a raster map is generated using the sub-point cloud data to obtain a map of the current location.

[0110] The process involves dividing the point cloud map into different sub-point clouds according to certain rules. This process can be based on point cloud coordinates, such as geographical location (longitude, latitude, altitude) or a spatial partitioning algorithm (such as octree, KD-tree, etc.) to divide the point cloud into smaller, more manageable regions or blocks. Each sub-point cloud contains point cloud data within a certain range, which can include 3D coordinates, color information, reflection intensity, etc. The data of each sub-point cloud is stored in a hash table. A hash table is an efficient data structure that allows us to quickly look up and access the corresponding value (sub-point cloud data) based on the key (in this scenario, it can be the identifier or range of the sub-point cloud). The advantage of this is that when point cloud data for a specific area is needed, we can quickly locate and retrieve it without traversing the entire point cloud map. When a map generation command is received for the current location, the system will look up and retrieve the sub-point cloud data within a preset range for that location from the hash table based on the coordinate information of the current location. The preset range can be adjusted according to application requirements, such as setting it to a circular area, rectangular area, or other shaped area with a fixed radius. After acquiring the sub-point cloud data within the current location area, we need to convert this data into a raster map according to the resolution requirements of the current location. A raster map is a map representation method that divides space into a series of uniform grids (or "grids"), each grid containing some statistical information about the area (such as height, obstacle density, etc.). The process of generating a raster map typically involves projecting the point cloud data onto a raster plane and calculating the attribute values ​​of each grid as needed.

[0111] Resolution determines the level of detail in a raster map. Higher resolution means smaller raster sizes and more raster cells, allowing for more accurate representation of terrain details; lower resolution means larger raster sizes and fewer raster cells, suitable for scenarios where detailed terrain requirements are less critical. Dynamically generating raster maps of varying detail based on the current location's resolution requirements satisfies the needs of different application scenarios and improves the flexibility of map generation.

[0112] Based on the sub-point cloud data, various attributes of each raster can be calculated. For example, for height information, the average height of all points within the raster can be calculated; for obstacle density, the number or density of points within the raster can be calculated, etc. The generated raster map is output as a map of the current location. This map can be used for various applications, such as navigation, route planning, and terrain analysis. Storing and quickly retrieving sub-point cloud data using a hash table significantly improves the speed and efficiency of map generation. It is also easily scalable to larger-scale point cloud map processing; simply adjust the hash table storage strategy and the raster map generation algorithm. The generated raster map is easy to understand and use, providing strong support for various map-based applications and improving the usability of map generation.

[0113] Specifically, such as Figure 4 As shown, the map management thread achieves efficient map management through point cloud data storage and raster map generation. First, the point cloud map is further segmented and stored in a hash table. Then, during the planning process, sub-point clouds are selected from the hash table as needed, and raster maps of corresponding resolutions are generated according to requirements. Point clouds in unvisited areas are not generated, enabling the map to generate high-resolution raster maps with relatively low storage overhead.

[0114] (1) Point cloud map storage

[0115] The system reads a prior point cloud map and divides it into sub-point clouds based on the coordinates of each point cloud. Each sub-point cloud is then placed into a hash table based on its coordinates. This method effectively manages map data across large-scale scenes.

[0116] (2) Raster map generation

[0117] Based on the current coordinates, select a sub-point cloud within a range of n from the hash table. Merge the extracted sub-point clouds into a complete map. Analyze the traversability of the point cloud map using a terrain analysis algorithm. First, place the point cloud into different voxels based on their x and y coordinates. Calculate the z-axis coordinate value of each point within the same voxel and sort them. The point with the lowest z-axis coordinate is considered a ground point, and points with z-axis coordinates within a certain threshold above the lowest point are also considered traversable points. Other point clouds are considered insurmountable obstacles. Furthermore, depending on whether the current planning state is a narrow-path mode, generate a raster map at different resolutions based on obstacle points to ensure map accuracy and computational efficiency in both narrow-path and non-narrow-path modes.

[0118] The safety and effectiveness of the improved algorithm in narrow-lane mode can be verified. By comparing A* algorithms with different safety radii and a hybrid A* algorithm that considers the vehicle's shape, the superior trajectory planning capability of the proposed adaptive hybrid A* algorithm in narrow-lane environments is verified. Specifically, the experimental roads include two types of narrow lanes with bends. Lane A has a lane width slightly greater than the vehicle width throughout, with the narrowest point being the vehicle width + 10cm, but lacks sufficient rotation space at the corners. Lane B has a lane width slightly greater than the vehicle width throughout, with the narrowest point being the vehicle width + 10cm, and the corner width being the vehicle width + 20cm, providing sufficient space for rotation and orientation adjustment. The experiments compare the performance of different algorithms in the two narrow-lane scenarios. The experimental results are shown in Table 1, demonstrating that the proposed adaptive hybrid A* algorithm performs better in narrow-lane environments.

[0119] Method Narrow A (actually not drivable) Narrow B (actually drivable) A* (inscribed circle expansion) Drivable Drivable A* (circumscribed circle expansion) Not drivable Not drivable Hybrid A* Not drivable Not drivable Adaptive Hybrid A* Not drivable Drivable

[0120] Table 1

[0121] In Table 1, when A* uses the radius of its inscribed circle as the safe distance, it will incorrectly plan a path in narrow passage A, leading to an unsafe path, such as... Figure 6 As shown, Figure 6 The diagram illustrates the path planning using the A* (inscribed circle expansion) method. When A* uses the radius of the inscribed circle as a safety distance, the overly strict safety distance judgment can lead to a situation where, even with a passable path, a narrow passage (B) cannot find a suitable trajectory, mistakenly concluding that the narrow passage is impassable. Figure 7 As shown, Figure 7 The diagram illustrates the path planning using the A* (external circle expansion) method. Hybrid A*, however, fails to consider the in-situ rotation of the inspection vehicle, making it impossible to find a suitable trajectory in narrow passages like B, where in-situ rotation for direction adjustment is required. Figure 8 As shown, Figure 8 A schematic diagram of the path planning using the hybrid A* method is shown. The proposed adaptive hybrid A* method can plan a passable path in narrow passage B while simultaneously detecting that narrow passage A is impassable, validating the algorithm's beneficial effects in narrow passage scenarios. Figure 9 As shown, Figure 9 A schematic diagram of path planning for an adaptive hybrid A* method according to an embodiment of the present disclosure is shown.

[0122] As can be seen, the method according to the embodiments of this disclosure, by introducing narrow-path scene recognition technology, prioritizes safety compared to non-narrow-path scenes, and achieves targeted path planning and control strategy optimization. Employing a multi-iteration path planning approach eliminates abnormal paths caused by observation errors, reducing their impact. Simultaneously, a local target point selection strategy based on trajectory shape effectively guides the robot to better conform to the trajectory for control, thereby reducing control errors. Furthermore, a hash table is used to store the original point cloud data to represent map information, effectively reducing storage requirements and avoiding space waste in large-scale scenes. Moreover, the precise coordinates of the original point cloud are further utilized to dynamically generate raster maps of different resolutions based on the characteristics of the narrow-path scene, ensuring planning accuracy while reducing the computational overhead of the algorithm.

[0123] According to the method of this disclosure, the proposed adaptive path planning algorithm exhibits more robust performance in narrow-path environments, while achieving efficient path planning in non-narrow-path environments. This allows the inspection system to ensure safety in narrow-path environments while maintaining high efficiency throughout the entire inspection process. Furthermore, through efficient map storage management, high-resolution map acquisition is achieved with minimal storage overhead, enabling the system to adapt to inspection needs in a wide range of scenarios and further improving its practicality and scalability.

[0124] It should be noted that the method of this disclosure embodiment can be executed by a single device, such as a computer or server. The method of this embodiment can also be applied to a distributed scenario, where multiple devices cooperate to complete the task. In such a distributed scenario, one of these devices may execute only one or more steps of the method of this disclosure embodiment, and the multiple devices will interact with each other to complete the method described.

[0125] It should be noted that the above description describes some embodiments of this disclosure. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in a different order than that shown in the above embodiments and still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0126] Based on the same technical concept, corresponding to the methods of any of the above embodiments, this disclosure also provides a path planning device for narrow scenarios, see [link to relevant documentation]. Figure 10 The path planning device for narrow scenarios includes:

[0127] The narrow road identification module is used to analyze the initial planned path based on point cloud map data and determine the narrow road segments and non-narrow road segments in the initial planned path;

[0128] The non-narrow road planning module is used to determine the non-narrow road target path and corresponding non-narrow road target speed data of the non-narrow road segment based on the first initial sampling point and preset speed data of the non-narrow road segment; and

[0129] The narrow road planning module is used to determine the target path and corresponding target speed of the narrow road section that meet preset conditions based on the second initial sampling point of the narrow road section;

[0130] The target path module is used to obtain a target path based on the non-narrow target path and the narrow target path;

[0131] A control module is used to drive on the target path based on the non-narrow-path target speed data and the narrow-path target speed.

[0132] For ease of description, the above apparatus is described in terms of its functions, divided into various modules. Of course, in implementing this disclosure, the functions of each module can be implemented in one or more software and / or hardware.

[0133] The apparatus described above is used to implement the corresponding path planning method for narrow scenarios in any of the foregoing embodiments, and has the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0134] Based on the same technical concept, corresponding to the methods of any of the above embodiments, this disclosure also provides a non-transitory computer-readable storage medium storing computer instructions for causing the computer to execute the path planning method for narrow scenarios as described in any of the above embodiments.

[0135] The computer-readable medium of this embodiment includes permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device.

[0136] The computer instructions stored in the storage medium of the above embodiments are used to cause the computer to execute the path planning method for narrow scenarios as described in any of the above embodiments, and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0137] Those skilled in the art should understand that the discussion of any of the above embodiments is merely exemplary and is not intended to imply that the scope of this disclosure (including the claims) is limited to these examples; within the framework of this disclosure, the technical features of the above embodiments or different embodiments can also be combined, the steps can be implemented in any order, and there are many other variations of different aspects of the embodiments of this disclosure as described above, which are not provided in detail for the sake of brevity.

[0138] Additionally, to simplify the description and discussion, and to avoid obscuring the embodiments of this disclosure, the provided drawings may or may not show well-known power / ground connections to integrated circuit (IC) chips and other components. Furthermore, the apparatus may be shown in block diagram form to avoid obscuring the embodiments of this disclosure, and this also takes into account the fact that the details of implementation of these block diagram apparatuses are highly dependent on the platform on which the embodiments of this disclosure will be implemented (i.e., these details should be fully understood by those skilled in the art). While specific details (e.g., circuitry) have been set forth to describe exemplary embodiments of this disclosure, it will be apparent to those skilled in the art that the embodiments of this disclosure may be implemented without these specific details or with variations thereof. Therefore, these descriptions should be considered illustrative rather than restrictive.

[0139] Although this disclosure has been described in conjunction with specific embodiments thereof, many substitutions, modifications, and variations of these embodiments will be apparent to those skilled in the art from the foregoing description. For example, other memory architectures (e.g., dynamic RAM (DRAM)) may be used with the embodiments discussed.

[0140] This disclosure is intended to cover all such substitutions, modifications, and variations that fall within the broad scope of the appended claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A path planning method for narrow scenarios, comprising: Based on point cloud map data, the initial planned path is analyzed to determine narrow road segments and non-narrow road segments within the initial planned path, including: The initial planned path is sampled uniformly to obtain initial sampling points; Determine the nearest valuable raster points on both sides of the path direction for the initial sampling point in the point cloud map; In response to the fact that the distance between the valuable grid points is less than a preset distance, the initial sampling point is determined as the first candidate narrow channel sampling point; Traverse all the initial sampling points and determine a preset number of sampling points adjacent to the first candidate narrow channel sampling point as the second candidate narrow channel sampling point; The narrow road segment is obtained based on the first candidate narrow road sampling point and the second candidate narrow road sampling point, and the non-narrow road segment in the initial planned path is determined as the non-narrow road segment; Based on the first initial sampling point and preset speed data of the non-narrow road segment, the non-narrow road target path and corresponding non-narrow road target speed data of the non-narrow road segment are determined, including: The first initial sampling point is subjected to a first smoothing process to obtain a first smooth non-narrow road segment; The first smoothed non-narrow road segment is subjected to a second smoothing process based on the smoothing cost function of the B-spline optimization algorithm to obtain the second smoothed non-narrow road segment. A sampling trajectory is generated for the preset speed data and the second smooth non-narrow road segment, and the trajectory cost function of the sampling trajectory is obtained; The sampled trajectory corresponding to the minimum value of the trajectory cost function is determined to be the non-narrow-path target path, and the preset speed data corresponding to the minimum value of the trajectory cost function is determined to be the non-narrow-path target speed data; and Based on the second initial sampling point of the narrow road section, determine the target path and corresponding target speed of the narrow road that meet the preset conditions, including: An adaptive trajectory is generated for the narrow road segment based on a multi-scale adaptive step size and an adaptive evaluation function; wherein, the adaptive evaluation function f(n) = g(n) + h(n) + s(n) + r(n) + z(n), where g(n) is the planned distance, h(n) is the predicted distance to the target, s(n) is the number of nearby obstacles, r(n) is the cost of rotating in place, and z(n) represents the cost of zero speed. The local navigation target point is selected in the adaptive trajectory based on the local navigation evaluation function. , where point i For the i-th point on the adaptive trajectory, line 0，n Let dis() be the distance function connecting the nth point and the 0th point. The second initial sampling point, which is furthest from the line connecting the points whose local navigation evaluation function is less than a preset threshold, is determined as the local target point; Based on the local navigation target point, generate the corresponding narrow channel target path and the narrow channel target speed; A target path is obtained based on the non-narrow target path and the narrow target path, and the vehicle travels on the target path based on the non-narrow target speed data and the narrow target speed.

2. The method according to claim 1, wherein, The smoothing cost function includes: f=λ1f s +λ2f c +λ3(f v +f w ), where f s For the cost of smoothness, f c For the cost of collision, f v and f w λ1 represents the soft constraints on linear velocity and angular velocity, λ2 represents the weight of smoothness cost, λ3 represents the weight of collision cost, and λ3 represents the weight of the soft constraints on linear velocity and angular velocity. The trajectory cost function includes: G(v,w)=α·head(v,w)+β·dist(v,w)+γ·velocity(v,w)+δ·path(v,w), Where head(v, w) is the azimuth evaluation function, dist(v, w) is the safety evaluation function, velocity(v, w) is the trajectory velocity evaluation function, path(v, w) is the overlap evaluation function between the sampled trajectory and the subsequent direction of the planned trajectory, α is the weight of the azimuth evaluation function, β is the weight of the safety evaluation function, γ is the weight of the trajectory velocity evaluation function, and δ is the weight of the overlap evaluation function.

3. The method according to claim 1, wherein, The first initial sampling point is subjected to a first smoothing process to obtain a first smooth non-narrow road segment, including: Starting from the first starting point of the first initial sampling point, determine whether there is a straight line among the first and last sampling points of the three adjacent first initial sampling points; If a straight line exists between the first and last sampling points of three adjacent first initial sampling points, the middle sampling point among the three adjacent first initial sampling points is removed. By traversing the first initial sampling points, the first smooth non-narrow road segment is obtained.

4. The method according to claim 1, further comprising: The point cloud map is divided into different sub-point clouds based on the point cloud coordinates in the point cloud map; The sub-point cloud data is obtained by saving the sub-point cloud data into a hash table. In response to receiving a map generation instruction for the current location, the sub-point cloud data corresponding to the sub-point cloud within a preset range of the current location coordinates is obtained from the hash table based on the current location coordinates. Based on the resolution corresponding to the current location, a raster map is generated using the sub-point cloud data to obtain a map of the current location.

5. A path planning device for narrow scenarios, comprising: The narrow road identification module is used to analyze the initial planned path based on point cloud map data and determine the narrow road segments and non-narrow road segments in the initial planned path; The non-narrow road planning module is used to determine the non-narrow road target path and corresponding non-narrow road target speed data of the non-narrow road segment based on the first initial sampling point and preset speed data of the non-narrow road segment, including: The first initial sampling point is subjected to a first smoothing process to obtain a first smooth non-narrow road segment; The first smoothed non-narrow road segment is subjected to a second smoothing process based on the smoothing cost function of the B-spline optimization algorithm to obtain the second smoothed non-narrow road segment. A sampling trajectory is generated for the preset speed data and the second smooth non-narrow road segment, and the trajectory cost function of the sampling trajectory is obtained; The sampled trajectory corresponding to the minimum value of the trajectory cost function is determined to be the non-narrow-path target path, and the preset speed data corresponding to the minimum value of the trajectory cost function is determined to be the non-narrow-path target speed data; and The narrow-path planning module is used to determine the target path and corresponding target speed of the narrow-path that meet preset conditions based on the second initial sampling point of the narrow-path segment, including: An adaptive trajectory is generated for the narrow road segment based on a multi-scale adaptive step size and an adaptive evaluation function; wherein, the adaptive evaluation function f(n) = g(n) + h(n) + s(n) + r(n) + z(n), where g(n) is the planned distance, h(n) is the predicted distance to the target, s(n) is the number of nearby obstacles, r(n) is the cost of rotating in place, and z(n) represents the cost of zero speed. The local navigation target point is selected in the adaptive trajectory based on the local navigation evaluation function. , where point i For the i-th point on the adaptive trajectory, line 0，n Let dis() be the distance function connecting the nth point and the 0th point. The second initial sampling point, which is furthest from the line connecting the points whose local navigation evaluation function is less than a preset threshold, is determined as the local target point; Based on the local navigation target point, generate the corresponding narrow channel target path and the narrow channel target speed; The target path module is used to obtain a target path based on the non-narrow target path and the narrow target path; The control module is used to drive on the target path based on the non-narrow-path target speed data and the narrow-path target speed.

6. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, implements the method as claimed in any one of claims 1 to 4.

7. A non-transitory computer-readable storage medium storing computer instructions for causing a computer to perform the method of any one of claims 1 to 4.

Images