Path planning method and device, related equipment and computer program product

Abstract

The invention discloses a path planning method and device, related equipment and a computer program product. According to the path planning method and device, multi-modal environment data is obtained based on a sensor, and a state space is created. And analyzing environment complexity of different areas in the state space, dynamically adjusting sampling strategies matched with the different areas according to the environment complexity, and sampling nodes in the corresponding areas according to the sampling strategies. The higher the environment complexity is, the higher the sampling density is, otherwise, the lower the environment complexity is, the lower the sampling density is. And searching a neighbor node set of each node in the state space, establishing a global route map, and searching a path from a starting point to an ending point in the global route. According to the scheme, in a dynamic and complex scene, different sampling strategies can be adopted in areas with different environmental complexity through the self-adaptive sampling strategy, the processing efficiency of the algorithm is improved, and the constructed global route map is smoother and higher in safety.

Description

Technical Field

[0001] The present application relates to the technical field of path planning, and more specifically, to a path planning method, device, related equipment, and computer program product. Background Art

[0002] With the development of fields such as autonomous driving and drone navigation, how to perform efficient path planning in complex environments has become a key issue. The traditional PRM (Probabilistic Roadmap Method) algorithm generates nodes and connection paths by randomly sampling in the state space to construct a roadmap, and realizes path search based on the roadmap. It is often inefficient in dynamic and complex scenarios, and the path is not smooth and lacks adaptability. Summary of the Invention

[0003] In view of the above problems, the present application is proposed to provide a path planning method, device, related equipment, and computer program product to at least overcome some or all of the defects existing in the traditional PRM algorithm. The specific solutions are as follows:

[0004] In a first aspect, a path planning method is provided, including:

[0005] Obtain multi-modal environmental data collected by sensors, and create a state space based on the environmental data;

[0006] Combined with the environmental data, analyze the environmental complexity of different regions in the state space, and dynamically adjust the sampling strategies matched by different regions according to the environmental complexity. Sample nodes in the corresponding regions according to the sampling strategies, where the higher the environmental complexity, the higher the sampling density in the corresponding sampling strategy;

[0007] Search for a set of neighbor nodes of each node in the state space, and connect the node with the neighbor nodes in the set of neighbor nodes to obtain a global roadmap;

[0008] Search for a path from the starting point to the ending point in the global roadmap.

[0009] In a possible design, in another implementation manner of the first aspect of the embodiments of the present application, it further includes:

[0010] Based on the multi-modal environmental data collected by sensors in real time, when it is detected that the environment has changed, analyze the local region where the affected path segment is located for the generated path;

[0011] Give priority to re-searching and updating the local path segment in the local region, and keep the path outside the local region unchanged.

[0012] In a possible design, in another implementation manner of the first aspect of the embodiments of the present application, the process of re-searching and updating local path segments within the local area includes:

[0013] Determine the first starting point and the first ending point of the affected path segment;

[0014] Resample nodes within the local area and construct a local roadmap based on the resampled nodes;

[0015] Search for a local path segment from the first starting point to the first ending point in the local roadmap.

[0016] In a possible design, in another implementation manner of the first aspect of the embodiments of the present application, the process of detecting whether the environment has changed includes:

[0017] Based on the multi-modal environmental data collected by the sensor in real time, determine the movement trend of the obstacle;

[0018] Based on the movement trend of the obstacle, determine that the environment has changed when the obstacle moves to a set safety distance around the path.

[0019] In a possible design, in another implementation manner of the first aspect of the embodiments of the present application, after updating the local path segment, it further includes:

[0020] Evaluate whether the environmental change range reaches a set proportional threshold, or whether the path quality of the complete path after updating the local path segment meets the set quality requirements;

[0021] If the environmental change range reaches the set proportional threshold, or the path quality does not meet the set quality requirements, continue with the global path update.

[0022] In a possible design, in another implementation manner of the first aspect of the embodiments of the present application, the process of analyzing the environmental complexity of different regions in the state space in combination with the environmental data includes:

[0023] In combination with the obstacle information in the environmental data, analyze the obstacle distribution density in different regions of the state space;

[0024] And / or,

[0025] In combination with the terrain information in the environmental data, analyze the terrain complexity of different regions in the state space;

[0026] Based on the obstacle distribution density and / or the terrain complexity, determine the environmental complexity of the region.

[0027] In a possible design, in another implementation manner of the first aspect of the embodiments of the present application, the sampling strategy further includes:

[0028] Avoiding a set range near obstacles during the sampling process;

[0029] And / or

[0030] Increasing the sampling density in a path candidate region, where the path candidate region is a set region around a connection line between the starting point and the ending point and estimated based on the environmental data.

[0031] In a possible design, in another implementation manner of the first aspect of the embodiments of the present application, the process of searching for a set of neighboring nodes of each node in the state space includes:

[0032] For each sampled node, using an approximate nearest neighbor search algorithm to search for a set of neighboring nodes in the state space, and dynamically adjusting the search radius according to the distribution density of neighboring nodes around the node during the search process, and adding each neighboring node within the search radius to the set of neighboring nodes of the node, where the higher the distribution density of neighboring nodes around, the smaller the search radius.

[0033] In a possible design, in another implementation manner of the first aspect of the embodiments of the present application, after searching for a set of neighboring nodes of each node, it further includes:

[0034] For each neighboring node in the set of neighboring nodes of each node, checking the terrain complexity and the distance to dynamic obstacles of the neighboring node one by one;

[0035] Deleting neighboring nodes whose terrain complexity exceeds a set complexity threshold or whose distance to dynamic obstacles is less than a set safety distance threshold from the set of neighboring nodes.

[0036] In a possible design, in another implementation manner of the first aspect of the embodiments of the present application, it further includes:

[0037] For the path searched from the starting point to the ending point, optimizing the path from at least one dimension of path smoothness, path length minimization, and path safety.

[0038] In a possible design, in another implementation manner of the first aspect of the embodiments of the present application, the process of obtaining multi-modal environmental data collected by a sensor includes:

[0039] Preprocessing the original multi-modal environmental data collected by the sensor to obtain preprocessed multi-modal environmental data;

[0040] Among them, the preprocessing process includes at least one of the following:

[0041] Noise filtering and anomaly detection, data standardization and normalization processing, data alignment and fusion, data compression and feature extraction, where the data alignment and fusion is used to align and fuse multi-modal data in time and space.

[0042] In a second aspect, a path planning device is provided, including:

[0043] A data acquisition unit, configured to acquire multi-modal environmental data collected by a sensor and create a state space based on the environmental data;

[0044] A node sampling unit, configured to analyze the environmental complexity of different regions in the state space in combination with the environmental data, dynamically adjust the sampling strategy matched by different regions according to the environmental complexity, and sample nodes in the corresponding regions according to the sampling strategy, where the higher the environmental complexity, the higher the sampling density in the corresponding sampling strategy;

[0045] A mapping unit, configured to search for a set of neighbor nodes of each node in the state space and connect the node with the neighbor nodes in the set of neighbor nodes to obtain a global roadmap;

[0046] A path search unit, configured to search for a path from a starting point to an ending point in the global roadmap.

[0047] In a third aspect, a movable device is provided, including: a sensor, a memory, and a processor;

[0048] The sensor is configured to collect multi-modal environmental data;

[0049] The memory is used to store a program;

[0050] The processor is configured to execute the program to implement each step of the path planning method described in any one of the first aspects of the present application.

[0051] In a fourth aspect, a readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, each step of the path planning method described in any one of the first aspects of the present application is implemented.

[0052] In a fifth aspect, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, each step of the path planning method described in any one of the first aspects of the present application is implemented.

[0053] With the above technical solutions, this application acquires multi-modal environmental data based on sensors, creates a state space based on this, can accurately model the surrounding environment, and provides accurate environmental information for path planning. When sampling in the state space, this application does not adopt a static random sampling strategy. Instead, it analyzes the environmental complexity of different regions in the state space, dynamically adjusts the sampling strategies matched by different regions according to the environmental complexity, and samples nodes in the corresponding regions according to the sampling strategies. The higher the environmental complexity, the higher the corresponding sampling density. Conversely, the lower the environmental complexity, the lower the corresponding sampling density. In this way, the sampling density can be increased in complex environments to achieve more refined path planning, which can improve the smoothness and safety of the path. In simple environments, the sampling density can be reduced to avoid waste of resources and ineffective computational volume caused by excessive sampling points, and the algorithm efficiency is improved. Further, the set of neighboring nodes of each node can be searched in the state space, and the connections between the node and each neighboring node in the set of neighboring nodes can be established to obtain a global roadmap, so that the path from the starting point to the ending point can be searched in the global roadmap. The solution of this application can adopt different sampling strategies in regions with different environmental complexities through an adaptive sampling strategy in dynamic and complex scenarios, improve the processing efficiency of the algorithm, and the constructed global roadmap is smoother and has higher safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] By reading the following detailed description of the preferred embodiments, various other advantages and benefits will become clear to those of ordinary skill in the art. The drawings are only for the purpose of showing the preferred embodiments and are not considered to be a limitation of this application. Moreover, throughout the drawings, the same reference numerals are used to represent the same components. In the drawings:

[0055] Figure 1 It is a schematic diagram of an implementation system architecture of the path planning method provided by an embodiment of this application;

[0056] Figure 2 It is a schematic diagram of the flow of a path planning method provided by an embodiment of this application;

[0057] Figure 3 It is a schematic diagram of the structure of a path planning device provided by an embodiment of this application;

[0058] Figure 4 It is a schematic diagram of the structure of a movable device provided by an embodiment of this application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0059] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0060] Through the multi-modal perception capabilities of sensors (including lidar, vision, and infrared data, etc.) in the present application, an accurate model of the surrounding environment is built. The collected data is processed through fusion, denoising, and standardization, providing accurate environmental information for path planning. By improving the traditional PRM algorithm, the solution introduces approximate nearest neighbor search (such as KD-tree and LSH, etc.) to optimize the node search efficiency; combined with an adaptive sampling strategy, the node distribution is dynamically adjusted, especially for areas with dense obstacles or frequent changes, increasing the sampling density and reducing invalid nodes. In addition, the solution proposes a dynamic path update mechanism based on local adjustment, enabling the system to respond to environmental changes in real time and maintaining the feasibility and safety of the path. Combining path optimization techniques (such as smoothing, minimizing path length, path safety, etc.), the present application ensures the global optimality of the path while ensuring the efficiency and safety of the system.

[0061] The path planning solution provided by the present application can be applied to movable devices such as autonomous driving vehicles, robots, and drones, providing path planning capabilities for such devices.

[0062] The present application provides a path planning method that can be applied to a system architecture as Figure 1 shown. The system may include a movable device 100 and a server 200. The server 200 may include one or more servers ( Figure 1 illustrated by including one server as an example). The movable device 100 may be various types of devices with mobility capabilities ( Figure 1 illustrated by drones, robots, and autonomous driving vehicles as examples).

[0063] The movable device 100 can be used alone to execute the path planning method provided in the embodiments of the present application. In addition, the movable device 100 and the server 200 can also be used in cooperation to execute the path planning method provided in the embodiments of the present application.

[0064] Embodiments of the present application provide a path planning method. Taking the application of this method to a computer device as an example, the computer device may specifically be Figure 1 the movable device 100 in Figure 2 or a system composed of the movable device 100 and the server 200. Referring to

[0065] Step S100: Obtain the multi-modal environmental data collected by the sensor, and create a state space based on the environmental data.

[0066] Among them, the multi-modal environmental data includes, but is not limited to, lidar data, visual data, infrared data, etc.

[0067] The sensor in this embodiment can adopt one or more types of sensors. In one example, a sensor with multi-modal perception ability can be adopted. For example, the CP-15-A4 type sensor has multi-modal perception ability and can collect lidar data, visual data, infrared data, etc. at the same time. In another example, the sensor in this embodiment can include two or more sensors with single-modal data collection ability. For example, it includes a lidar sensor, a visual sensor, and an infrared sensor, which collect lidar data, visual data, and infrared data respectively.

[0068] In an optional example, considering that the raw data collected by the sensor may contain noise and irregular information, directly using these data will affect the accuracy of path planning. Therefore, in this step, the raw multi-modal environmental data collected by the sensor can be further preprocessed to ensure the accuracy and efficiency of subsequent algorithm execution.

[0069] The process of preprocessing the multi-modal environmental data can include various preprocessing methods. In one possible implementation, the preprocessing process can include at least one of the following:

[0070] Noise filtering and anomaly detection, data standardization and normalization processing, data alignment and fusion, data compression and feature extraction.

[0071] 1. For the noise filtering and anomaly detection process:

[0072] The raw data collected by the sensor usually contains noise and outliers (such as lidar outliers, random noise in images, etc.). If not filtered first, subsequent steps may be affected by these interferences, resulting in inaccurate results or introducing errors. Therefore, noise filtering and anomaly detection are the first step in data preprocessing, aiming to clean the raw data and provide a reliable basis for subsequent processing.

[0073] For the collected lidar data: The abnormal point cloud data can be removed through statistical filtering or the RANSAC (Random Sample Consensus) algorithm to ensure the accuracy of the obstacle contour.

[0074] For the collected visual data: Image denoising techniques based on Gaussian filters can be used to remove random noise in the images captured by the camera. Further optionally, an image edge detection algorithm can also be used to identify the key obstacle contours and shapes.

[0075] For the collected infrared data: The invalid temperature information can be filtered out by the threshold segmentation method to ensure effective obstacle detection in low-light or complex environments.

[0076] Specifically, infrared data is usually generated by detecting the thermal radiation on the surface of an object through an infrared sensor (such as a thermal imager) to obtain temperature distribution information, which is usually output in the form of a two-dimensional thermal map or a numerical matrix. The temperature value reflects the thermal characteristics of an object or area in the environment.

[0077] The threshold segmentation method belongs to a signal processing method. By setting one or more temperature thresholds, the data is divided into valid and invalid parts. Temperatures below or above a certain threshold are considered "invalid", depending on the application target. In the obstacle detection scenario of path planning in this embodiment, "invalid temperature information" refers to those temperature ranges that do not contribute to obstacle recognition or may interfere with the judgment. One example is that the definition of the invalid temperature range is as follows:

[0078] Too low temperature range:

[0079] Range: Areas close to the ambient background temperature (such as the natural temperature of the outdoor ground surface, such as 0°C to 30°C, depending on the scenario).

[0080] In an infrared image, the ambient background (such as the ground, air) usually shows a relatively low temperature, and the temperature difference from obstacles (such as people, vehicles) is relatively obvious. If these low-temperature data are retained, it will increase the background noise and blur the obstacle contours.

[0081] Too high temperature range:

[0082] Range: Extreme values far exceeding the normal obstacle temperature (such as higher than 100°C or higher, depending on the sensor range).

[0083] These may be the results of sensor noise, reflection interference, or non-target objects (such as a heat source), which have nothing to do with typical obstacles (such as pedestrians, vehicles) in path planning.

[0084] The above only gives an exemplary description of the noise filtering and anomaly detection processes for lidar data, visual data, and infrared data. Corresponding processing methods can also be adopted for other modalities of environmental data, which will not be listed one by one in this embodiment.

[0085] 2. For the data standardization and normalization process:

[0086] Sensor data of different modalities (such as LiDAR point clouds, RGB images, and infrared data) vary greatly in format, dimension, and range. If data fusion is carried out directly, the integration effect may be affected due to inconsistent scales. Therefore, after noise filtering, it is necessary to standardize and normalize the single-modal data to convert it into a unified scale for subsequent fusion and calculation.

[0087] During the processing, the resolution and precision of the data can be adjusted according to the sampling resolution of the sensor and the scale of the environment, so as to ensure the consistency of data input. Exemplarily, the ranges and formats of the LiDAR point clouds, RGB images, and infrared data after noise filtering can be unified. For example, all data can be normalized to the interval [0, 1], and the resolution can be adjusted according to the scale of the environment.

[0088] 3. For the data alignment and fusion process:

[0089] To better represent the complex environment, the collected multi-modal environmental data can be effectively fused. First, the environmental data of different modalities can be spatially and temporally aligned and fused into a multi-modal environmental model to generate complete obstacle and terrain information. Through clock synchronization and spatial calibration, it is ensured that the environmental data of different modalities have the same timestamp and coordinate system. Through this multi-source fusion technology, a more complete environmental model can be generated, reducing the perception blind area and improving the accuracy of the data.

[0090] 4. For the data compression and feature extraction process:

[0091] To improve the real-time performance and processing efficiency of the algorithm, data compression and feature extraction can also be carried out in data preprocessing. For the point cloud data of LiDAR, the density of the point cloud can be reduced by voxel grid method, but the key feature points are retained; for image data, a convolutional neural network (CNN) or a neural network of other structures can be used for feature extraction to convert the high-dimensional original image data into low-dimensional feature vectors and retain the key obstacle information.

[0092] For the four preprocessing methods in the above example, a possible processing order is shown as: 1. Noise filtering and anomaly detection, 2. Data standardization and normalization processing, 3. Data alignment and fusion, 4. Data compression and feature extraction.

[0093] Through the above data preprocessing process, the multi-modal environmental data collected by the sensor has been greatly optimized, laying a solid foundation for the subsequent path planning algorithm processing.

[0094] Furthermore, the multi-modal environmental data can be used to create a state space, which contains obstacles and free space for the device to move.

[0095] Step S110: Analyze the environmental complexity of different regions in the state space in combination with the environmental data, dynamically adjust the sampling strategies matched by different regions according to the environmental complexity, and sample nodes in the corresponding regions according to the sampling strategies.

[0096] Among them, the higher the environmental complexity, the higher the sampling density in the corresponding sampling strategy; conversely, the lower the environmental complexity, the lower the sampling density in the corresponding sampling strategy.

[0097] The environmental complexity describes the complexity of environmental features such as obstacles and terrain in the region. The more dense the obstacles and the more uneven the terrain structure, the higher the corresponding environmental complexity. Since the sensor can obtain real-time environmental data, the sampling strategy matched by the region can be dynamically adjusted according to the real-time environmental complexity of each region in the state space, realizing an adaptive sampling strategy.

[0098] In a possible implementation, the obstacle distribution density of different regions in the state space can be analyzed in combination with the obstacle information in the environmental data, and / or the terrain complexity of different regions in the state space can be analyzed in combination with the terrain information in the environmental data. Based on the obstacle distribution density and / or terrain complexity, the environmental complexity of the region is determined.

[0099] The traditional PRM algorithm randomly samples in the state space. Although this method is simple, it does not consider the specific characteristics of the environment when sampling nodes, resulting in uneven node density. The sampling density is often insufficient in areas with dense obstacles and complex terrain, which in turn leads to difficulties in subsequent path planning, the generated path is not smooth enough, and the safety is not high; in open areas, there are easily too many sampling points, causing resource waste and increasing the computational amount.

[0100] The adaptive sampling strategy based on environmental data provided in this embodiment can analyze the environmental complexity of different regions based on environmental data, and then dynamically adjust the sampling strategy, improve the processing efficiency of the algorithm, and the constructed global roadmap is smoother and has higher safety.

[0101] Step S120: Search for the set of neighbor nodes of each node in the state space, and connect the node with the neighbor nodes in the set of neighbor nodes to obtain a global roadmap.

[0102] By searching for the set of neighbor nodes of each sampling node, several nearby neighbor nodes connected to each sampling node can be obtained, so that a global roadmap can be obtained as the basis for subsequent path search.

[0103] Considering that the nearest neighbor search adopted by the traditional PRM algorithm lock is performed in a high-dimensional space, its computational complexity is relatively large. Especially when path planning involves complex three-dimensional scenarios, the computational overhead of each search will increase significantly. To improve efficiency, an approximate nearest neighbor search algorithm can be introduced in this embodiment, such as approximate nearest neighbor search algorithms like KD tree (K-Dimensional Tree) or LSH (Locality-Sensitive Hashing). These methods can significantly reduce the computational complexity while maintaining a relatively high search accuracy by partitioning the space or hash mapping, and are particularly suitable for scenarios of long-range path planning.

[0104] Taking the KD tree algorithm as an example, the KD tree is a tree structure that recursively divides a multi-dimensional space into several sub-spaces. When using the KD tree for approximate nearest neighbor search, the spatial data collected by the sensor is first dimensionally reduced. Through a clustering algorithm or PCA (Principal Component Analysis), the high-dimensional spatial point cloud data is compressed to a dimension suitable for KD tree processing. Then, the KD tree will perform spatial partitioning according to the dimensionally reduced data, and each node represents a region. When searching for the nearest neighbor node, the KD tree can quickly exclude impossible regions through pruning operations, reducing the search space, thereby accelerating the neighbor search process.

[0105] In a possible implementation, for each node obtained by sampling in this embodiment, an approximate nearest neighbor search algorithm is used to search for a set of nearest neighbor nodes in the state space.

[0106] Optionally, to further improve the search efficiency, a dynamic search radius adjustment mechanism is also provided in this embodiment. As the spatial scale increases and the environmental complexity increases, the search radius will be dynamically adjusted according to the distribution density of the nodes. Specifically, during the process of searching for a set of nearest neighbor nodes for any sampling node, the search radius is dynamically adjusted according to the distribution density of the nearest neighbor nodes around the node, and the nearest neighbor nodes within the search radius are added to the set of nearest neighbor nodes of the node. Among them, the higher the distribution density of the surrounding nearest neighbor nodes, the smaller the search radius; conversely, the lower the distribution density of the surrounding nearest neighbor nodes, the larger the search radius.

[0107] In this embodiment, in areas with more obstacles or more complex environments, the search radius will be correspondingly reduced to ensure the accuracy and safety of the connection; while in open areas, the search radius can be appropriately enlarged to reduce meaningless sampling and connection, thereby balancing the computational burden.

[0108] In a possible implementation, after obtaining the set of neighboring nodes for each node in this step, it is also possible to check the terrain complexity and the distance to dynamic obstacles of each neighboring node in the set of neighboring nodes of each node one by one. Remove the neighboring nodes whose terrain complexity exceeds the set complexity threshold or the distance to dynamic obstacles is less than the set safety distance threshold from the set of neighboring nodes.

[0109] By further screening the set of neighboring nodes, it is possible to eliminate the neighboring nodes in complex terrains and the neighboring nodes that are too close to dynamic obstacles, reduce the ineffective connections between nodes, and improve the safety of the path.

[0110] Step S130: Search for a path from the starting point to the ending point in the global roadmap.

[0111] Specifically, based on the requirements of path planning, the positions of the starting point and the ending point can be determined. Furthermore, the starting node closest to the starting point and the ending node closest to the ending point can be determined in the global roadmap. Then, a path search algorithm can be used to search for a feasible path from the starting node to the ending node.

[0112] In this embodiment, in order to improve the calculation efficiency, a heuristic search method can be adopted, which can quickly find the global optimal path. At the same time, the real-time data of the sensor can be combined, and the heuristic function can be dynamically adjusted to avoid falling into local optimal solutions during the search process. The introduction of heuristic search not only speeds up the planning speed of the global path but also ensures the rationality of the path in complex environments.

[0113] The path planning method provided by the present application based on the embodiment is based on the sensor to obtain multi-modal environmental data. Based on this, a state space can be created to accurately model the surrounding environment and provide accurate environmental information for path planning. When sampling in the state space, the present application does not adopt a static random sampling strategy. Instead, it analyzes the environmental complexity of different regions in the state space, dynamically adjusts the sampling strategies matched by different regions according to the environmental complexity, and samples nodes in the corresponding regions according to the sampling strategies. The higher the environmental complexity, the higher the corresponding sampling density. Conversely, the lower the environmental complexity, the lower the corresponding sampling density. In this way, the sampling density can be increased in complex environmental regions to achieve more refined path planning, which can improve the smoothness and safety of the path. In simple environmental regions, the sampling density can be reduced to avoid waste of resources and ineffective calculation amount caused by too many sampling points, and the algorithm efficiency is improved. The solution of the present application can adopt different sampling strategies in regions with different environmental complexities through an adaptive sampling strategy in dynamic and complex scenarios, improve the processing efficiency of the algorithm, and the constructed global roadmap is smoother and has higher safety.

[0114] In some possible implementations, the sampling strategy in the foregoing steps of the present application may further include:

[0115] The sampling process avoids the set range near obstacles.

[0116] Specifically, to further optimize the sampling efficiency, this application can combine the precise obstacle detection ability of the sensor to avoid sampling within the set range (including inside the obstacle) near the obstacle. Among them, the sensor can provide real-time contour information of the obstacle. Based on this, sampling points inside the obstacle and sampling points within a set distance around the obstacle contour can be automatically filtered out during sampling, reducing ineffective calculations, ensuring that all sampling nodes are located in the feasible area, and avoiding collisions between the subsequently generated path and the obstacle or being too close to the obstacle, which may affect traffic safety.

[0117] In some other possible implementations, the sampling strategy in the foregoing steps of this application may further include:

[0118] Increase the sampling density in the path candidate area (the area where the path may exist). Among them, the path candidate area can be a set area around the connection line estimated based on the connection line between the starting point and the ending point and the environmental data.

[0119] As an example, determine the target area around the connection line from the starting point to the ending point, and based on the environmental data, eliminate the area where the obstacle is located and the area where the terrain does not meet the requirements (such as areas with uneven terrain) in the target area. The remaining area forms the path candidate area.

[0120] By setting the above sampling strategy, more sampling points are preferentially generated in the area where the path may exist, improving the rationality and feasibility of the overall path.

[0121] The core of path planning lies not only in the generation of the initial path but also in the dynamic adjustment ability in the face of environmental changes. The traditional PRM (Probabilistic Roadmap Method) algorithm often uses static planning in path search, that is, it searches along a fixed path after generating nodes, lacking adaptability to the dynamic environment. This application introduces a dynamic adjustment mechanism and combines the real-time environment perception function of the sensor to achieve flexible response to environmental changes during the path planning process, especially having significant advantages in remote and complex scenarios.

[0122] Specifically, some embodiments of this application further propose a dynamic path update mechanism based on local adjustment. Through local update, the algorithm can significantly reduce the cost of recalculation and improve the real-time response ability.

[0123] After generating the planned path in the foregoing embodiments, based on the multi-modal environmental data collected by the sensor in real time, when it is detected that the environment has changed, the local area where the affected path segment is located in the generated path can be analyzed. Priority is given to re-searching and updating the local path segment within this local area, while keeping the path outside the local area unchanged.

[0124] The sensor in this application has the ability of multi-modal real-time perception and can continuously monitor the obstacles and terrain changes in the environment. Through the fusion of lidar, visual information, infrared data, etc., the environmental model can be updated in real time, accurately identifying the dynamic changes in the environment, such as moving vehicles, pedestrians, or suddenly appearing obstacles. These real-time data enable the algorithm to dynamically adjust the path search strategy, no longer relying on static environmental assumptions, thus avoiding path failure caused by environmental changes.

[0125] In a dynamic scenario, the obstacles, terrain, etc. in the environment may change. In this embodiment, the detection and judgment of environmental changes can be performed based on the environmental data collected by the sensor in real time. When it is determined that the environmental change meets the set conditions, the path local update process is triggered to execute the above-mentioned local path update process.

[0126] Among them, whether the environment has changed can be judged based on predefined conditions. An optional example is to determine the movement trend of the obstacle based on the multi-modal environmental data collected by the sensor in real time; based on the movement trend of the obstacle, it is determined that the environment has changed when the obstacle moves to a set safety distance around the path.

[0127] In a dynamic environment, the safety of the path is crucial. To ensure the safety of the path during dynamic adjustment, a set safety distance can be introduced in this embodiment to generate a buffer zone around the obstacle. When it is detected based on the movement trend of the obstacle that the obstacle is approaching the path, in order to avoid the path colliding with the obstacle or entering the buffer zone of the obstacle, it can be determined that the environment has changed when it is determined that the obstacle moves to a set safety distance around the path, that is, the path local update process is triggered. On the basis of ensuring the safety distance, the local path is replanned to prevent the path from passing through the obstacle or falling into a dangerous area.

[0128] In this embodiment, based on the real-time environmental detection ability of the sensor, the movement trend of the obstacle can be predicted in advance, and the local path adjustment can be completed before the obstacle reaches the path, further improving the safety and stability of the path.

[0129] In some other possible implementations, the predefined judgment conditions for environmental changes may further include one or more of the following:

[0130] An obstacle appears or disappears; the real-time traffic condition changes, such as traffic congestion, accidents, road construction, etc.; the meteorological condition changes, such as heavy rain, strong wind, haze, etc. are detected.

[0131] Since the sensor can obtain multi-modal environmental data in real time, it supports users to flexibly configure conditions according to the actual usage scenario to trigger the local path update process when the conditions are met.

[0132] In a possible implementation, the process of re-searching and updating the local path segment in the above embodiment in the local area can be implemented according to the following steps:

[0133] S1. Determine the first starting point and the first ending point of the affected path segment.

[0134] S2. Resample nodes in the local area and construct a local roadmap based on the resampled nodes.

[0135] Specifically, nodes can be resampled in the local area according to the environmental information after the local area changes. For example, when an obstacle moves or disappears, new nodes can be resampled near the original obstacle position.

[0136] S3. Search for the local path segment from the first starting point to the first ending point in the local roadmap.

[0137] By resampling nodes in the local area where the path is affected and constructing a local roadmap, it is possible to quickly search for the local path segment from the first starting point to the first ending point based on the local roadmap, realizing the rapid update of the affected local path segment.

[0138] In some embodiments of the present application, a node dynamic sampling rule is provided, which can adjust the sampling strategy according to the requirements of different path planning stages. Specifically, in the path initialization stage, nodes can be sampled in the global scope (that is, in the global scope of the state space) to generate a relatively rough path framework. In the subsequent path optimization and update processes, resampling can be performed in the local area, thereby improving the accuracy and flexibility of the path. This dynamic sampling strategy can effectively reduce the redundancy of the initial sampling while ensuring that the generated path is smooth enough.

[0139] Based on the local path update mechanism disclosed in the above embodiment, in this embodiment, a balance scheme between global update and local update under dynamic adjustment is further provided.

[0140] In some cases, such as when there are significant changes in the environment or the existing path is too complex, this application can also comprehensively consider the balance between local updates and global replanning. Through the global perception ability of the sensor, it can be determined whether global path updates need to be continued after local updates to ensure the integrity and optimality of the final path.

[0141] In one possible implementation, after completing the local path update, it is possible to evaluate whether the range of environmental changes reaches a set proportional threshold, or whether the path quality of the complete path after the local path segment update meets the set quality requirements.

[0142] If the range of environmental changes reaches the set proportional threshold, or the path quality does not meet the set quality requirements, then global path updates continue.

[0143] Among them, the range of environmental changes can be the ratio of the area with environmental changes in the calculated state space to the overall area.

[0144] The path quality can be evaluated from dimensions such as path smoothness, path length, and path safety. The path smoothness represents the continuity and degree of curvature change of the path, the path length represents the total moving distance from the starting point to the ending point, and path safety refers to the degree to which the path is far from obstacles or dangerous areas, usually measured by the safety distance or collision probability.

[0145] It can be understood that the smoother the path, the shorter the length, and the higher the safety, the higher the corresponding path quality.

[0146] This balance strategy between global updates and local updates provided by this embodiment can effectively avoid the high computational cost brought by global replanning, and at the same time, on the basis of ensuring path quality, maintain the efficiency and stability of the path planning result. By introducing a path dynamic adjustment mechanism, this application improves the path search strategy of the traditional PRM algorithm, enabling it to flexibly respond to changes in the dynamic environment. Combining with the real-time environmental perception function of the sensor, it can achieve efficient and reliable path planning in complex remote scenarios through local path updates. The dynamic adjustment mechanism not only improves the adaptability of the system but also significantly reduces the computational cost of global replanning, realizing the real-time and robustness of path planning.

[0147] In some embodiments of this application, for the searched path, it is also possible to further optimize the path to meet requirements in multiple aspects such as efficiency, smoothness, and safety.

[0148] The generated initial path is usually based on node sampling and nearest neighbor connection. Although it can provide a feasible solution, it often has the following deficiencies: Path non - smoothness: The initial path is usually composed of multiple discrete nodes, resulting in the path may have corners or be non - smooth in some local areas. Especially in complex terrains or areas with dense obstacles, the path may appear tortuous, affecting the navigation efficiency. Path length non - optimality: Due to the initial path may be affected by sampling, the generated path is not necessarily the shortest path. Especially in a multi - obstacle environment, there is room for further optimizing the path. Insufficient safety: The initial path may have potential risks in a dynamic environment and does not fully consider the movement trends and spatial changes of dynamic obstacles. To solve these problems, path optimization has become a key step to ensure the path is efficient, smooth, and safe.

[0149] Therefore, for the path optimization solution provided in this embodiment, for the path searched from the starting point to the ending point, the path can be optimized from at least one dimension among path smoothness, path length minimization, and path safety.

[0150] Among them, path smoothness represents the continuity and curvature change degree of the path, path length minimization means the total moving distance from the starting point to the ending point is the shortest, and path safety refers to the degree that the path is far from obstacles or dangerous areas, usually measured by the safety distance or collision probability.

[0151] For path smoothness, smoothing techniques such as Bezier curves or B - Spline curves can be used to smooth the corners or discontinuous segments in the path. By interpolating between the initial nodes to generate a smooth curve, the optimized path can more naturally adapt to turns and narrow channels in complex environments.

[0152] For path length minimization, based on dynamic programming or heuristic methods (such as the A* algorithm etc.), the global length of the path can be optimized. Using the obstacle and terrain information sensed by the sensor, re - evaluate the local and global lengths of the path, and shorten the overall driving distance by searching for the shortest path or the optimal path segment, thereby improving the efficiency of the path.

[0153] For path safety, in order to ensure the safety of the path in complex and dynamic environments, during the optimization process, the driving route of the path can be adjusted according to the dynamic obstacle data provided by the sensor in real - time to avoid approaching moving obstacles or potential dangerous areas. By introducing a prediction model of dynamic obstacles, it is possible to predict the moving direction and speed of the obstacles, and reserve enough safety distance during path optimization.

[0154] In summary, the improvements of the solution of this application include but are not limited to the following aspects: 1. Multi-modal sensor integration: The CP-15-A4 sensor combines lidar, vision, and infrared data to achieve a comprehensive perception of the environment. The multi-modal data fusion technology can generate a more accurate environmental model to ensure the accuracy of path planning. 2. Approximate nearest neighbor search: Efficient data structures such as KD-trees or locality-sensitive hashing (LSH) are introduced for node search, significantly reducing the computational complexity in high-dimensional spaces. This method not only improves the node generation efficiency but also ensures the accuracy of path connection. During the search process, the search radius can be dynamically adjusted according to the node distribution density, reducing the computational complexity while ensuring the search accuracy. 3. Adaptive node sampling strategy: Combining real-time environmental information, the sampling strategy is dynamically adjusted according to the obstacle distribution in different regions. The sampling density is increased in areas with dense obstacles, while invalid nodes are reduced in open areas to improve the feasibility of the path. 4. Dynamic path adjustment mechanism: By monitoring environmental changes in real time, the system can flexibly perform local updates on the generated path, avoiding global replanning, enhancing the robustness and real-time response ability of path planning. At the same time, a smoothing scheme between global and local updates is further provided. 5. Path optimization and safety guarantee: Through multi-objective optimization strategies, the path is ensured to be in an optimal state in terms of length, smoothness, and safety. In addition, the detection and prediction model of dynamic obstacles can further enhance the safety of the path.

[0155] The path planning solution of this application can be widely applied to multiple fields such as unmanned driving, robot navigation, and intelligent transportation.

[0156] The path planning device provided by the embodiments of this application will be described below. The path planning device described below can be correspondingly referred to the path planning method described above.

[0157] See Figure 3 , Figure 3 which is a schematic structural diagram of a path planning device disclosed in the embodiments of this application.

[0158] As Figure 3 shown, the device may include:

[0159] A data acquisition unit 11, configured to acquire multi-modal environmental data collected by sensors and create a state space based on the environmental data;

[0160] A node sampling unit 12, configured to analyze the environmental complexity of different regions in the state space in combination with the environmental data, and dynamically adjust the sampling strategy matched by different regions according to the environmental complexity, and sample nodes in the corresponding regions according to the sampling strategy, where the higher the environmental complexity, the higher the sampling density in the corresponding sampling strategy;

[0161] A map building unit 13, configured to search for a set of neighbor nodes of each of the nodes in the state space, and connect the node with the neighbor nodes in the set of neighbor nodes to obtain a global roadmap;

[0162] A path search unit 14, configured to search for a path from a starting point to an ending point in the global roadmap.

[0163] In a possible implementation, the device of the present application may further include:

[0164] A local path update unit, configured to, based on multi-modal environmental data collected in real time by a sensor, when it is detected that the environment has changed, analyze a local area where an affected path segment is located for the generated path; preferentially re-search and update a local path segment within the local area, and keep the path outside the local area unchanged.

[0165] In a possible implementation, the process of the local path update unit re-searching and updating a local path segment within the local area includes:

[0166] Determine a first starting point and a first ending point of the affected path segment;

[0167] Re-sample nodes within the local area, and construct a local roadmap based on the re-sampled nodes;

[0168] Search for a local path segment from the first starting point to the first ending point in the local roadmap.

[0169] In a possible implementation, the process of the local path update unit detecting whether the environment has changed includes:

[0170] Based on multi-modal environmental data collected in real time by a sensor, determine the movement trend of an obstacle;

[0171] Based on the movement trend of the obstacle, determine that the environment has changed when the obstacle moves to a set safety distance around the path.

[0172] In a possible implementation, the device of the present application may further include:

[0173] A global and local update balance unit, configured to, after the local path update unit updates the local path segment, evaluate whether the range of environmental change reaches a set proportional threshold, or whether the path quality of the complete path after the local path segment is updated meets a set quality requirement;

[0174] If the range of environmental change reaches the set proportional threshold, or the path quality does not meet the set quality requirement, continue to perform global path update.

[0175] In one possible implementation, the process of the node sampling unit analyzing the environmental complexity of different regions in the state space by combining the environmental data includes:

[0176] Combining the obstacle information in the environmental data to analyze the obstacle distribution density in different regions of the state space;

[0177] And / or,

[0178] Combining the terrain information in the environmental data to analyze the terrain complexity in different regions of the state space;

[0179] Based on the obstacle distribution density and / or the terrain complexity, determining the environmental complexity of the region.

[0180] In one possible implementation, the sampling strategy further includes:

[0181] Avoiding a set range near obstacles during the sampling process;

[0182] And / or,

[0183] Increasing the sampling density in the path candidate region, where the path candidate region is a set region around the connection line between the start point and the end point estimated based on the environmental data.

[0184] In one possible implementation, the process of the mapping unit searching for the set of neighboring nodes of each node in the state space includes:

[0185] For each sampled node, using the approximate nearest neighbor search algorithm to search for the set of neighboring nodes in the state space. During the search process, dynamically adjust the search radius according to the distribution density of the neighboring nodes around the node, and add each neighboring node within the search radius to the set of neighboring nodes of the node. Among them, the higher the distribution density of the surrounding neighboring nodes, the smaller the search radius.

[0186] In one possible implementation, after the mapping unit searches for the set of neighboring nodes of each node, it is further used for:

[0187] For each neighboring node in the set of neighboring nodes of each node, check the terrain complexity and the distance to the dynamic obstacle of the neighboring node one by one;

[0188] Delete the neighboring nodes whose terrain complexity exceeds the set complexity threshold or the distance to the dynamic obstacle is less than the set safety distance threshold from the set of neighboring nodes.

[0189] In one possible implementation, the device of the present application may further include:

[0190] A path optimization unit is configured to optimize the searched path from the starting point to the ending point in at least one dimension among path smoothness, path length minimization, and path safety.

[0191] Among them, the path smoothness represents the continuity and curvature change degree of the path, the path length minimization represents the shortest total movement distance from the starting point to the ending point, and the path safety refers to the degree of the path being far from obstacles or dangerous areas, usually measured by a safety distance or a collision probability.

[0192] In a possible implementation, the process of the data acquisition unit acquiring the multi-modal environmental data collected by the sensor includes:

[0193] Preprocessing the original multi-modal environmental data collected by the sensor to obtain the preprocessed multi-modal environmental data;

[0194] Among them, the preprocessing process includes at least one of the following:

[0195] Noise filtering and anomaly detection, data standardization and normalization processing, data alignment and fusion, data compression and feature extraction, where the data alignment and fusion is used to align and fuse multi-modal data in terms of time and space.

[0196] An embodiment of the present application also provides a movable device. Refer to Figure 4 As shown, it shows a schematic structural diagram of a movable device suitable for implementing the movable device in the embodiment of the present application. The movable device in the embodiment of the present application may include, but is not limited to, movable devices such as unmanned aerial vehicles, robots, automobiles, and the like. Figure 4 The shown movable device is only an example and should not bring any limitation to the functions and usage scope of the embodiment of the present application.

[0197] As Figure 4 As shown, the movable device may include a processing device (such as a central processing unit, a graphics processing unit, etc.) 601, which may perform various appropriate actions and processes according to the program stored in the read-only memory (ROM) 602 or the program loaded from the storage device 608 into the random access memory (RAM) 603 to implement the path planning method of the foregoing embodiment of the present application. When the movable device is powered on, various programs and data required for the operation of the movable device are also stored in the RAM 603. The processing device 601, the ROM 602, and the RAM 603 are connected to each other through a bus 604. The input / output (I / O) interface 605 is also connected to the bus 604.

[0198] Typically, the following devices can be connected to the I / O interface 605: an input device 606, which includes sensors for collecting multi-modal environmental data, as well as an accelerometer, a gyroscope, etc.; an output device 607, which includes, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage device 608, which includes, for example, a memory card, a hard disk, etc.; and a communication device 609. The communication device 609 can allow the mobile device to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 4 a mobile device with various devices is shown, it should be understood that it is not required to implement or have all the shown devices. More or fewer devices can be alternatively implemented or had.

[0199] An embodiment of the present application also provides a computer program product, including computer-readable instructions, which, when running on an electronic device, enable the electronic device to implement any path planning method provided by the embodiment of the present application.

[0200] An embodiment of the present application also provides a computer-readable storage medium, which carries one or more computer programs, and when the one or more computer programs are executed by an electronic device, can enable the electronic device to implement any path planning method provided by the embodiment of the present application.

[0201] In addition, it should be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place, or may be distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. In addition, in the drawings of the device embodiments provided in the present application, the connection relationship between the modules indicates that they have a communication connection, which can be specifically implemented as one or more communication buses or signal lines.

[0202] Through the description of the above embodiments, those skilled in the art can clearly understand that the present application can be implemented by means of software plus necessary general hardware. Of course, it can also be implemented by dedicated hardware including application-specific integrated circuits, dedicated CPUs, dedicated memories, dedicated components, etc. Generally, functions accomplished by computer programs can be easily implemented by corresponding hardware, and the specific hardware structures for implementing the same function can also be diverse, such as analog circuits, digital circuits, or dedicated circuits. However, for the present application, software program implementation is a better embodiment in more cases. Based on such an understanding, the technical solution of the present application, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product is stored in a readable storage medium, such as a floppy disk, USB flash drive, mobile hard disk, ROM, RAM, magnetic disk, or optical disc of a computer, and includes several instructions for causing a computer device (which can be a personal computer, training device, or network device, etc.) to execute the methods described in various embodiments of the present application.

[0203] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product.

[0204] The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the processes or functions described in the embodiments of the present application are generated in whole or in part. The computer can be a general-purpose computer, a dedicated computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from a website, computer, training device, or data center to another website, computer, training device, or data center in a wired manner (such as coaxial cable, optical fiber, digital subscriber line (DSL)) or a wireless manner (such as infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that a computer can store or a data storage device such as a training device or data center that includes one or more integrated available media. The available medium can be a magnetic medium (such as a floppy disk, hard disk, magnetic tape), an optical medium (such as a DVD), or a semiconductor medium (such as a solid state disk (SSD)).

[0205] The various embodiments in this specification are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. The various embodiments can be combined as needed, and the same or similar parts can be referred to each other.

Claims

1. A path planning method, characterized in that: include: Acquire multimodal environmental data collected by sensors, and create a state space based on the environmental data; In combination with the environmental data, the environmental complexity of different areas in the state space is analyzed, and the sampling strategies matched to the different areas are dynamically adjusted according to the environmental complexity, and nodes are sampled in the corresponding areas according to the sampling strategies, wherein the higher the environmental complexity, the higher the sampling density in the corresponding sampling strategy; Searching for a neighbor node set of each of the nodes in the state space, and connecting the node with neighbor nodes in the neighbor node set to obtain a global route map; A path from a starting point to an end point is searched in the global roadmap.

2. The method according to claim 1, characterized in that Also includes: Based on the multi-modal environmental data collected by sensors in real time, when a change in the environment is detected, the local area where the affected path segment is located is analyzed for the generated path; The local path segments are preferentially re-searched and updated within the local area, and the path outside the local area is kept unchanged.

3. The method according to claim 2, characterized in that The process of re-searching and updating the local path segment in the local area includes: determining a first start point and a first end point of the affected path segment; resampling nodes in the local area and constructing a local roadmap based on the resampled nodes; A local path segment from the first starting point to the first end point is searched in the local route map.

4. The method according to claim 2, characterized in that: The process of detecting whether the environment has changed includes: Determine the movement trend of obstacles based on multi-modal environmental data collected by sensors in real time; Based on the movement trend of the obstacle, it is determined that the environment has changed when the obstacle moves to a set safety distance around the path.

5. The method according to claim 2, characterized in that: After updating the local path segments, also include: Evaluate whether the scope of environmental change reaches a set ratio threshold, or whether the path quality of the complete path after the local path segment is updated meets the set quality requirements; If the environmental change range reaches the set ratio threshold, or the path quality does not meet the set quality requirements, the global path update will continue.

6. The method according to claim 1, characterized in that The process of analyzing the environmental complexity of different regions in the state space in combination with the environmental data includes: Analyzing the obstacle distribution density in different areas of the state space in combination with the obstacle information in the environmental data; and / or, Analyzing the terrain complexity of different areas in the state space in combination with the terrain information in the environmental data; The environmental complexity of the area is determined based on the obstacle distribution density and / or the terrain complexity.

7. The method according to claim 1, characterized in that The sampling strategy also includes: The sampling process avoids the set range near obstacles; and / or, The sampling density in the path candidate area is improved, and the path candidate area is a set area around the line between the starting point and the end point and the estimated environment data.

8. The method according to claim 1, characterized in that: The process of searching for a set of neighboring nodes of each of the nodes in the state space comprises: For each node obtained by sampling, an approximate nearest neighbor search algorithm is used to search for a set of neighbor nodes in the state space. During the search process, the search radius is dynamically adjusted according to the distribution density of the neighbor nodes around the node, and each neighbor node within the search radius is added to the neighbor node set of the node. The higher the distribution density of the surrounding neighbor nodes, the smaller the search radius.

9. The method according to claim 1, characterized in that: After searching and obtaining a set of neighboring nodes of each of the nodes, the method further includes: For each neighboring node in the neighboring node set of each node, check the terrain complexity and dynamic obstacle distance of the neighboring node one by one; The neighboring nodes whose terrain complexity exceeds a set complexity threshold or whose dynamic obstacle distance is less than a set safety distance threshold are deleted from the neighboring node set.

10. The method according to claim 1, characterized in that Also includes: For the searched path from the starting point to the end point, the path is optimized from at least one dimension of path smoothness, path length minimization, and path safety.

11. The method according to any one of claims 1 to 10, characterized in that: The process of obtaining multi-modal environmental data collected by sensors includes: Preprocessing the original multimodal environmental data collected by the sensor to obtain preprocessed multimodal environmental data; Wherein, the pretreatment process includes at least one of the following: Noise filtering and anomaly detection, data standardization and normalization processing, data alignment and fusion, data compression and feature extraction, wherein the data alignment and fusion is used to align and fuse multimodal data in time and space.

12. A path planning device, characterized in that: include: A data acquisition unit, used to acquire multi-modal environmental data collected by sensors, and create a state space based on the environmental data; A node sampling unit, used to analyze the environmental complexity of different areas in the state space in combination with the environmental data, and dynamically adjust the sampling strategies matched to different areas according to the environmental complexity, and sample nodes in corresponding areas according to the sampling strategies, wherein the higher the environmental complexity, the higher the sampling density in the corresponding sampling strategy; A mapping unit, configured to search for a set of neighboring nodes of each of the nodes in the state space, and connect the node with the neighboring nodes in the set of neighboring nodes to obtain a global route map; A path search unit is used to search for a path from a starting point to an end point in the global route map.

13. A movable device, characterized in that: include: Sensors, memory and processors; The sensor is used to collect multi-modal environmental data; The memory is used to store programs; The processor is used to execute the program to implement each step of the path planning method as described in any one of claims 1 to 11.

14. A readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, each step of the path planning method according to any one of claims 1 to 11 is implemented.

15. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, each step of the path planning method as described in any one of claims 1 to 11 is implemented.

Images