Path planning method, system and equipment for complex terrain and storage medium

Abstract

The invention discloses a path planning method, system and device for a complex terrain and a storage medium. The method comprises the following steps: acquiring topographic data of a complex terrain through a laser radar, and constructing a map model based on an irregular triangular network; calculating the topographic cost of each triangle according to the topographic features of the map model, and converting the topographic cost of each triangle into the total topographic cost; and inputting the total terrain cost of each triangle into a Double DQN algorithm, and planning an optimal path. By combining the three-dimensional terrain modeling technology and the reinforcement learning algorithm, efficient, accurate and real-time path planning can be realized in a complex three-dimensional environment, the limitation of the prior art is overcome, and the development of the autonomous navigation technology of the robot is promoted.

Description

Technical Field

[0001] The present invention relates to the technical field of robot path planning, and in particular to a method, system, device and storage medium for path planning in complex terrain. Background Art

[0002] The goal of path planning is to enable the robot to accurately identify and avoid obstacles in complex three-dimensional terrain, ensure the accuracy of path planning, and enable it to accurately reach the target location in a changing terrain environment. Traditional robot navigation and path planning technologies mainly include methods based on graph search, sampling, and grid maps. These methods lay the theoretical foundation for path planning, propose key concepts such as heuristic search, and promote the application of grid maps in path planning.

[0003] The A* algorithm is a heuristic search algorithm based on graph search. It uses the cost function to estimate the optimal path from the starting point to the end point, and through reasonable heuristic function design, it can achieve a good balance between search efficiency and path quality.

[0004] Dijkstra's algorithm is another classic graph search algorithm that is specifically used to find the shortest path from a single source point to all other nodes. It gradually expands the nodes and eventually finds the shortest path to the target node. It can ensure that the global optimal path from the starting point to the end point is found, and it is applicable to various types of graphs.

[0005] RRT is a sampling-based path planning algorithm that generates a tree structure by randomly sampling points in space and gradually expands until a feasible path from the start point to the end point is found. It is usually used for path planning in high-dimensional and complex environments. It performs well in high-dimensional space and is suitable for handling path planning problems in dynamic obstacles or uncertain environments.

[0006] In addition to the above three path planning methods based on graph search and sampling, the grid map method discretizes the environment into regular grids and performs path search on the grids. Common search algorithms include A*, Dijkstra, etc. This method is usually used in relatively regular two-dimensional or three-dimensional space. This method is intuitive and easy to understand, and the calculation is simple. When combined with the graph search algorithm, it can find a feasible path quickly.

[0007] However, before path planning, the first problem to be solved is how to obtain complex terrain data as input data for the path planning algorithm. The mainstream three-dimensional terrain modeling technologies are mainly point cloud modeling and grid-based modeling. Point cloud modeling collects unordered point sets on the terrain surface through laser radar (LiDAR), structured light, time-of-flight (ToF) camera or stereo vision technology, and builds a three-dimensional model based on these points. Point cloud data usually has high density and high precision, can accurately reflect the slight changes in terrain, and is widely used in high-precision map production, architectural surveying and mapping, geological exploration, unmanned driving navigation and other fields. The grid model divides the three-dimensional space into a regular grid of cubes (voxels), and each voxel represents a part of the terrain surface. This method is simple and intuitive, easy to implement, and compatible with existing two-dimensional grid map methods. It is mainly used in geographic information systems, environmental modeling, drone navigation and other fields.

[0008] Traditional path planning technologies perform well in two-dimensional environments or simple three-dimensional environments, but they face many challenges and limitations in complex terrains. These limitations make traditional path planning technologies less than ideal when dealing with complex three-dimensional terrain environments. Therefore, it is necessary to combine more advanced three-dimensional terrain modeling technology and intelligent path planning algorithms (such as deep reinforcement learning) to improve the efficiency and effectiveness of planning.

[0009] (1) Computational complexity of high-dimensional space: In three-dimensional complex terrain, the dimension of the state space involved in path planning increases significantly, causing the computational complexity of the search algorithm to increase exponentially. In this case, traditional graph search algorithms such as A* and Dijkstra are difficult to complete path planning within an acceptable time.

[0010] (2) Inaccuracy in continuous space: Traditional path planning algorithms usually rely on discretized map models, such as grid maps or graph structures. However, in complex three-dimensional terrain, the discretization process may lead to the loss of terrain features or error accumulation, which in turn affects the accuracy and effectiveness of path planning.

[0011] (3) Real-time requirements and adaptability to dynamic environments: Complex terrain is often accompanied by dynamic changes, such as moving obstacles and sudden changes in terrain. In such an environment, path planning needs to have real-time and adaptive capabilities. Traditional algorithms are usually unable to respond quickly to changes in the environment, resulting in path planning results being invalid or unreliable in practical applications.

[0012] (4) Local optimal trap: In high-dimensional or complex terrain, path search algorithms are prone to fall into local optimal solutions and find it difficult to find the global optimal path. For example, the RRT algorithm may generate a large number of invalid nodes in some complex areas, resulting in poor path quality or waste of computing resources.

[0013] (5) Path smoothness and feasibility: In complex three-dimensional terrain, the generated path needs to have good smoothness and feasibility to ensure that robots or drones can execute safely and smoothly. However, traditional path planning algorithms often ignore the problem of path smoothness, and the generated path may have sharp turns or discontinuities, which is not conducive to actual operation.

[0014] In complex terrain, refining the grid can improve accuracy, but it will also lead to an explosive growth in data volume and computational complexity. This contradiction between accuracy and computing resources makes the raster model perform poorly in large-scale or detailed terrain modeling. In flat areas or areas with fewer features, the raster model will still generate a large amount of voxel data, which not only increases storage requirements but also reduces processing efficiency. In addition, the raster model uses fixed-size cubes for spatial division, and terrain boundaries are often not accurately expressed, especially when dealing with detailed areas such as slopes and cliffs, the model boundaries may appear discontinuous or inaccurate. In dynamic environments, especially when dealing with large-scale terrain or scenes that require high-frequency updates, the computational overhead of the raster model will increase significantly, affecting real-time performance. Summary of the invention

[0015] In view of the limitations of existing path planning technology in complex three-dimensional terrain, the present invention proposes a method, system, device and storage medium for path planning of an autonomous mobile robot under complex terrain; the present invention combines an irregular triangular mesh terrain modeling method and a robot real-time path planning method based on deep reinforcement learning, which can achieve efficient, accurate and real-time path planning effects in complex three-dimensional environments.

[0016] In order to achieve the above object, the present invention provides the following technical solutions: The present invention provides a path planning method for complex terrain, which is characterized by comprising: S1: Obtain terrain data of complex terrain and construct a map model based on irregular triangulated network; S2: Based on the map model, calculate the terrain features of each triangle, and then convert the terrain features of each triangle into a terrain cost; S3: Based on the terrain cost of each triangle and combined with the Double DQN algorithm, the best moving path is planned for the robot.

[0017] The method of obtaining terrain data of complex terrain and constructing a map model based on an irregular triangulated network includes: The terrain data of complex terrain is obtained by laser radar, and the noise in the terrain data is removed by statistical filter, and then down-sampled by voxel grid method to obtain point cloud data; the ground points and non-ground points in the point cloud data are separated by random sample consistency algorithm, and the point cloud data A except the ground is obtained after separation; Input point cloud data A into the point cloud set in space, select 4 points in the circumscribed cube of the input points to construct an initial tetrahedron, which covers all points; Select a new point to insert into the point cloud set in the space, and traverse all tetrahedrons adjacent to the new point; check whether each adjacent tetrahedron contains the new point. If it does, delete the tetrahedron containing the new point, and use the new point and adjacent points to re-form a new tetrahedron; New points are inserted repeatedly until all points in the point cloud data A are inserted into the point cloud set in space, thereby obtaining a map model based on an irregular triangulated network.

[0018] The terrain characteristics include unevenness, steepness and slope.

[0019] The calculation formula of the slope is:

[0020] Among them, Slope is the slope, Δz is the height difference between the two vertices of the triangle, and Δd is the horizontal distance between the two points; The calculation formula of the steepness is:

[0021] Among them, Steepness is steepness, Slope is slope; The calculation formula of the unevenness is:

[0022] Among them: Roughness is the unevenness, is the average value of the height of the area, and the calculation formula is:

[0023] Among them, z i There are i height values ​​in a certain area, and n is the total number of height values ​​in a certain area.

[0024] The terrain features of each triangle are converted into terrain costs, specifically: Weights are assigned to slope, steepness and unevenness respectively. , , , calculate the terrain cost of each triangle, the calculation formula is as follows:

[0025] Among them, Cost is the terrain cost; = 1; , , They represent the maximum slope, maximum steepness, and maximum roughness of the terrain that the robot can pass through.

[0026] The terrain cost of each triangle is combined with the Double DQN algorithm to plan the best moving path for the robot, including: The terrain cost of each triangle is input into the Double DQN algorithm, and the terrain cost attribute is added to the robot state after combining the terrain cost; Through the behavior network in the Double DQN algorithm, the next action is selected according to the current action; after each action is executed, the next action is entered from the current action and the reward given by the environment is obtained, which is obtained according to the terrain cost; The target network in the Double DQN algorithm evaluates the Q value of the next action to reduce the overestimation of the Q value. Get the action set with the highest cumulative reward value and plan the best movement path for the robot.

[0027] The Q value of the next action is evaluated by the target network in the Double DQN algorithm to reduce the over-estimation of the Q value, including: The overestimation of the behavior network and the target network is adjusted by the loss function, which is as follows:

[0028] Among them: s is the current state; is the chosen action; r is the reward obtained from the environment; s′ is the next state; is the discount factor; is the Q value of the behavior network, with parameter θ; is the Q value of the target network, with parameter θ′; is the next action selected from the behavior network.

[0029] The present invention also provides a complex terrain path planning system, comprising the following modules: The map model building module is used to obtain terrain data of complex terrain and build a map model based on irregular triangulated networks; A terrain cost calculation module, used to calculate the terrain features of each triangle based on the map model, and then convert the terrain features of each triangle into a terrain cost; The path planning module is used to plan the best moving path for the robot based on the terrain cost of each triangle and combined with the Double DQN algorithm.

[0030] The present invention also provides a computer-readable storage medium storing one or more programs, wherein the one or more programs include instructions, which, when executed by a computing device, enable the computing device to perform the steps of the path planning method for complex terrain as described in any one of claims 1 to 7.

[0031] The present invention also provides a terminal 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 configured to be executed by the one or more processors, and the one or more programs include steps for executing the path planning method for complex terrain according to any one of claims 1 to 7 Compared with the prior art, the present invention has the following beneficial effects: The present invention uses an irregular triangular network to establish a digital elevation map model of complex terrain. The size and shape of the triangle can be flexibly adjusted according to the complexity of the terrain. It can not only better adapt to terrain changes, but also use less storage space and reduce data redundancy. At the same time, TIN uses triangles to more accurately represent terrain features and provide higher terrain accuracy. Secondly, the terrain cost information of slope, steepness and unevenness is calculated based on the terrain features of the above map model, and this is used as the reward function of the deep reinforcement learning Double DQN algorithm network to plan the shortest path that the robot can pass, ensuring the feasibility and efficiency of the robot's path planning strategy.

[0032] The three-dimensional terrain is modeled using irregular triangulated network technology. Compared with the traditional grid model, TIN can more accurately represent the details of complex terrain, reduce data redundancy and computational burden, and improve the accuracy and efficiency of path planning. The Double DQN algorithm is applied to path planning in three-dimensional terrain. By separating the process of action selection and Q value evaluation, it effectively avoids the overestimation problem in the Q-learning algorithm, and can autonomously learn and find the global optimal path in complex terrain. The reinforcement learning Double DQN has adaptive capabilities and can update the path planning strategy in real time to adapt to dynamic changes in the environment, greatly improving the practicality and reliability of the system in dynamic or unknown environments. Combining efficient three-dimensional terrain modeling technology and deep reinforcement learning algorithms, the use of computing resources in the path planning process is optimized. While ensuring high-precision path planning, the system significantly reduces its dependence on computing resources and can still operate efficiently under resource-constrained conditions. It is suitable for the application scenario of efficient autonomous navigation of rescue robots. By combining TIN with Double DQN, the present invention can handle more complex terrain structures, accurately plan paths that meet actual needs, perform better when dealing with complex and detailed three-dimensional terrain, and effectively avoid the problem of insufficient accuracy of traditional methods in complex environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0034] Figure 1 A flow chart of a path planning method for complex terrain according to the present invention; Figure 2 A flow chart for constructing a terrain model of the present invention; Figure 3 A terrain model map of a small hill based on TIN according to an embodiment of the present invention; Figure 4 This is a schematic diagram of slope calculation in irregular triangulated network; Figure 5 This is a schematic diagram of steepness calculation in irregular triangulated network; Figure 6 It is a schematic diagram of the calculation of unevenness in irregular triangulated network; Figure 7 It is the algorithm block diagram of Double DQN in the present invention; Figure 8 A schematic diagram of a path planning system for complex terrain according to the present invention; Fig. 9 The present invention is a block diagram of a terminal device provided according to an embodiment. DETAILED DESCRIPTION

[0035] In order to enable those skilled in the art to understand the characteristics and effects of the present invention, the following is a general description and definition of the terms and expressions mentioned in the specification and claims. Unless otherwise specified, all technical and scientific terms used in the text are the common meanings understood by those skilled in the art for the present invention. In the event of a conflict, the definition in this specification shall prevail.

[0036] The theories or mechanisms described and disclosed in the present invention, whether correct or incorrect, should not limit the scope of the present invention in any way, that is, the present invention can be implemented without being limited by any specific theory or mechanism.

[0037] In the present invention, all features such as values, quantities, contents and concentrations defined in the form of numerical ranges or percentage ranges are for simplicity and convenience only. Accordingly, the description of numerical ranges or percentage ranges should be deemed to have included and specifically disclosed all possible secondary ranges and individual values ​​within the range (including integers and fractions).

[0038] In the present invention, unless otherwise specified, “comprising”, “including”, “containing”, “having” or similar terms cover the meanings of “consisting of” and “mainly consisting of”, for example, “A includes a” covers the meanings of “A includes a and other” and “A only includes a”.

[0039] In the present invention, in order to make the description concise, not all possible combinations of various technical features in various embodiments or examples are described. Therefore, as long as there is no contradiction in the combination of these technical features, the various technical features in various embodiments or examples can be combined arbitrarily, and all possible combinations should be considered to be within the scope of this specification.

[0040] In response to the limitations of existing path planning technology in complex three-dimensional terrain, this patent proposes a robot real-time path planning method that combines irregular triangular mesh terrain modeling method and deep reinforcement learning, which can achieve efficient, accurate and real-time path planning effects in complex three-dimensional environments.

[0041] The present invention utilizes an irregular triangulated network (ITN) for three-dimensional terrain modeling and a three-dimensional path planning method based on deep reinforcement learning (Double-Deep Q-Network, DDQN); it is applied to robot navigation and path optimization in complex environments, especially in three-dimensional spaces with irregular terrain features, to achieve efficient and accurate path planning.

[0042] A map of the complex ground environment is established, and an irregular triangular network digital elevation map model of the environment is established through laser radar. Then, the various costs of the terrain, including slope, step and unevenness costs, are calculated through the elevation value and entered into the DDQN algorithm network as input data, and a search strategy is designed and path planning and path realization are performed. In order to effectively represent the complex terrain model and perform visual analysis and subsequent path planning, the present invention obtains terrain data through laser radar and constructs an irregular triangular network model, including data acquisition, point cloud data preprocessing, triangulation to construct an irregular triangular network, and model optimization and visual analysis.

[0043] like Figure 1 As shown, the present invention provides a path planning method for complex terrain, comprising: S1: Obtain terrain data of complex terrain and construct a map model based on irregular triangulated network (TIN); S2: Based on the map model, calculate the terrain features of each triangle, and then convert the terrain features of each triangle into a terrain cost; S3: Based on the terrain cost of each triangle and combined with the Double DQN algorithm, the best path for the robot to move is planned.

[0044] The present invention introduces advanced three-dimensional terrain modeling technology and reinforcement learning algorithms. Reinforcement learning can autonomously learn and optimize path planning strategies in unknown or dynamic environments through continuous interaction with the environment, and is particularly suitable for dealing with complex and high-dimensional path planning problems. The present invention can achieve efficient, accurate, and real-time path planning in complex three-dimensional environments, overcome the limitations of existing technologies, and promote the development of robot autonomous navigation technology.

[0045] Among them, Figure 2 As shown, S1: Obtain terrain data of complex terrain and construct a map model based on irregular triangulated network (TIN), specifically: S11: Obtain terrain data of complex terrain through lidar, use statistical filters to remove noise in the terrain data, and then downsample through the voxel grid method to obtain point cloud data; use a random sample consistency algorithm to separate ground points from non-ground points in the point cloud data, and after separation, obtain point cloud data A except the ground.

[0046] S12: Construct a map model based on an irregular triangulated network according to the point cloud data A, specifically: Input point cloud data A into a point cloud set in space (three-dimensional space), select 4 points in the circumscribed cube of the input points to construct an initial tetrahedron, which covers all points; Select a new point to insert into the point cloud set in the space, and traverse all tetrahedrons adjacent to the new point; check whether each adjacent tetrahedron contains the new point. If it does, delete the tetrahedron containing the new point, and use the new point and adjacent points to re-form a new tetrahedron; New points are inserted repeatedly until all points in the point cloud data A are inserted into the point cloud set in space, thereby obtaining a map model based on an irregular triangulated network.

[0047] It should be noted that the principle of building a map model based on an irregular triangular network is based on Delaunay triangulation, the purpose of which is to generate a triangulated network (tetrahedron) on a given set of points so that the circumscribed circle (circumscribed sphere) of each triangle does not contain other points. Figure 3 As shown, an irregular triangular network model of a hillside is represented.

[0048] S2: Based on the map model, the terrain features of each triangle in the map model are calculated, and the terrain features of each triangle are converted into terrain costs, specifically: The terrain characteristics include unevenness, steepness and slope.

[0049] Figure 4 The diagram shows the slope calculation in irregular triangulated network. Slope refers to the degree of ground inclination, which is calculated by the height difference and base length of the triangle. The slope calculation formula is as follows:

[0050] Among them, Δz is the height difference between the two vertices of the triangle, and Δd is the horizontal distance between the two points. Figure 5 A schematic diagram showing the calculation of steepness in an irregular triangulated network. Steepness: Steepness usually refers to the intensity of the slope, which can be quantified by the absolute value of the slope. Areas with greater steepness may increase the difficulty of the robot's walking. The steepness is measured in degrees of slope angle. The calculation formula for steepness is as follows:

[0051] Where Slope is the slope.

[0052] Figure 6 Schematic diagram of roughness calculation in irregular triangulated network. Roughness: Roughness can be measured by calculating the height change rate of the triangle interval. The commonly used indicator is the standard deviation of the height change in the local area. Assuming that there are n height values ​​z1, z2, ..., zn in a certain area, the calculation formula of roughness is:

[0053] in: is the average value of the height of the area, and the calculation formula is:

[0054] In order to convert the above terrain features into terrain costs as the input data of the Double DQN reward function, weights are assigned to the slope, steepness, and unevenness, respectively. , , , and then calculate the terrain cost (Cost) of each triangle. The calculation formula is as follows:

[0055] in = 1; , , They represent the maximum slope, maximum step, and maximum roughness of the terrain that the robot can pass through.

[0056] like Figure 7 As shown, S3: Based on the terrain cost of each triangle, combined with the Double DQN algorithm, the best moving path is planned for the robot, specifically: The terrain cost of each triangle is input into the Double DQN algorithm, and the terrain cost attribute is added to the robot state after combining the terrain cost. For example, the state can be (x, y, z, cost). According to the state space, the robot's action space can be forward, backward, left, right, upper left, lower left, upper right, lower right, etc. Through the behavior network in the Double DQN algorithm, the next action is selected according to the current action; after each action is executed, the next action is entered from the current state and the reward given by the environment is obtained according to the terrain cost; The target network in the Double DQN algorithm evaluates the Q value of the next action to reduce the overestimation of the Q value. Get the action set with the highest cumulative reward value and plan the best movement path for the robot.

[0057] The overestimation of the behavior network and the target network is adjusted by the loss function, which is as follows:

[0058] Among them: s: current state; : the action chosen; r: the reward obtained from the environment; s′: the next state; : discount factor; : Q value of the behavior network, parameter is θ; : The Q value of the target network, the parameter is θ′; : The next action selected from the behavior network.

[0059] The present invention utilizes the terrain model constructed by irregular triangulated networks, combined with the deep reinforcement learning algorithm Double-DQN, to achieve efficient path planning. Based on the terrain model of irregular triangulated networks, the robot's position, posture and other states in three-dimensional space, as well as movement, turning and other dynamics are defined; combined with terrain features, a reward function is designed to encourage the robot to choose a safe and efficient path, avoid obstacles and dangerous areas, and the path planning strategy is trained and optimized through the Double DQN algorithm, so that the robot can autonomously learn in complex terrain and adapt to different environmental conditions. The present invention can accurately describe complex terrain and achieve rapid path planning through optimization algorithms; it is suitable for a variety of complex environments, including dynamically changing terrains such as mountainous areas, urban ruins, and forests; it can be applied to unmanned vehicles, search and rescue robots, agricultural automation equipment, and military reconnaissance robots, and has significant market value and technology promotion potential.

[0060] like Figure 8 As shown, in one embodiment of the present invention, a path planning system for complex terrain is provided, which specifically includes: The map model building module is used to obtain terrain data of complex terrain and build a map model based on irregular triangulated networks; A terrain total cost calculation module, used to calculate the terrain features of each triangle based on the map model, and then convert the terrain features of each triangle into a terrain cost; The path planning module is used to plan the best moving path for the robot based on the terrain cost of each triangle and combined with the Double DQN algorithm.

[0061] In another embodiment of the present invention, the present invention also provides a storage medium, specifically a computer-readable storage medium (Memory), which is a memory device in a terminal device for storing programs and data. It can be understood that the computer-readable storage medium here can include both the built-in storage medium in the terminal device and the extended storage medium supported by the terminal device. The computer-readable storage medium provides a storage space, which stores the operating system of the terminal. In addition, one or more instructions suitable for being loaded and executed by the processor are also stored in the storage space, and these instructions can be one or more computer programs (including program codes). It should be noted that the computer-readable storage medium here can be a high-speed RAM memory or a non-volatile memory (Non-Volatile Memory), such as at least one disk memory.

[0062] One or more instructions stored in a computer-readable storage medium may be loaded and executed by a processor to implement the corresponding steps of the above-mentioned calculation method; one or more instructions in the computer-readable storage medium are loaded and executed by a processor as follows: S1: Obtain terrain data of complex terrain and construct a map model based on irregular triangulated network; S2: Based on the map model, calculate the terrain features of each triangle, and then convert the terrain features of each triangle into a terrain cost; S3: Based on the terrain cost of each triangle and combined with the Double DQN algorithm, the best moving path is planned for the robot.

[0063] In another embodiment of the present invention, a terminal device is provided, the terminal device includes a processor and a memory, the memory is used to store a computer program, the computer program includes program instructions, and the processor is used to execute the program instructions stored in the computer storage medium. The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc., which are the computing core and control core of the terminal, which are suitable for implementing one or more instructions, and are specifically suitable for loading and executing one or more instructions to implement the corresponding method flow or corresponding functions; the processor described in the embodiment of the present invention can be used for the operation of the computing method, including: S1: Obtain terrain data of complex terrain and construct a map model based on irregular triangulated network; S2: Based on the map model, calculate the terrain features of each triangle, and then convert the terrain features of each triangle into a terrain cost; S3: Based on the terrain cost of each triangle and combined with the Double DQN algorithm, the best moving path is planned for the robot.

[0064] like Fig. 9 As shown, the terminal device 600 is an electronic device, and the electronic device is in the form of a general computing device. The components of the electronic device may include but are not limited to: at least one processing unit 610, at least one storage unit 620, a bus 630 connecting different platform components (including the storage unit 620 and the processing unit 610), a display unit 640, etc.

[0065] The storage unit stores program codes, which can be executed by the processing unit 610, so that the processing unit 610 executes the steps according to various exemplary embodiments of the present invention described in the above method part of this specification.

[0066] The storage unit 620 may include a readable medium in the form of a volatile storage unit, such as a random access memory unit (RAM) 6201 and / or a cache memory unit 6202 , and may further include a read-only memory unit (ROM) 6203 .

[0067] The storage unit 620 may also include a program / utility 6204 having a set (at least one) of program modules 6205, such program modules 6205 including but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment.

[0068] Bus 630 may represent one or more of several types of bus structures, including a memory unit bus or memory unit controller, a peripheral bus, an accelerated graphics port, a processing unit, or a local bus using any of a variety of bus architectures.

[0069] The electronic device 600 may also communicate with one or more external devices 700 (e.g., keyboards, pointing devices, Bluetooth devices, etc.), may also communicate with one or more devices that enable a user to interact with the electronic device 600, and / or communicate with any device that enables the electronic device 600 to communicate with one or more other computing devices (e.g., routers, modems, etc.). Such communication may be performed via an input / output (I / O) interface 650. Furthermore, the electronic device 600 may also communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) via a network adapter 660. The network adapter 660 may communicate with other modules of the electronic device 600 via a bus 630. It should be understood that, although not shown in the figure, other hardware and / or software modules may be used in conjunction with the electronic device 600, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage platforms, etc.

[0070] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0071] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0072] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0073] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0074] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A path planning method for complex terrain, characterized in that: include: S1: Obtain terrain data of complex terrain and construct a map model based on irregular triangulated network; S2: Based on the map model, calculate the terrain features of each triangle, and then convert the terrain features of each triangle into a terrain cost; S3: Based on the terrain cost of each triangle and combined with the Double DQN algorithm, the best moving path is planned for the robot.

2. A complex terrain path planning method according to claim 1, characterized in that: The method of obtaining terrain data of complex terrain and constructing a map model based on an irregular triangulated network includes: The terrain data of complex terrain is obtained by laser radar, and the noise in the terrain data is removed by statistical filter, and then down-sampled by voxel grid method to obtain point cloud data; the ground points and non-ground points in the point cloud data are separated by random sample consistency algorithm, and the point cloud data A except the ground is obtained after separation; Input point cloud data A into the point cloud set in space, select 4 points in the circumscribed cube of the input points to construct an initial tetrahedron, which covers all points; Select a new point to insert into the point cloud set in the space, and traverse all tetrahedrons adjacent to the new point; check whether each adjacent tetrahedron contains the new point. If it does, delete the tetrahedron containing the new point, and use the new point and adjacent points to re-form a new tetrahedron; New points are inserted repeatedly until all points in the point cloud data A are inserted into the point cloud set in space, thereby obtaining a map model based on an irregular triangulated network.

3. A complex terrain path planning method according to claim 1, characterized in that: The terrain characteristics include unevenness, steepness and slope.

4. A complex terrain path planning method according to claim 3, characterized in that: The calculation formula of the slope is: Among them, Slope is the slope, Δz is the height difference between the two vertices of the triangle, and Δd is the horizontal distance between the two points; The calculation formula of the steepness is: Among them, Steepness is steepness, Slope is slope; The calculation formula of the unevenness is: Among them: Roughness is the unevenness, is the average value of the height of the area, and the calculation formula is: Among them, z i There are i height values ​​in a certain area, and n is the total number of height values ​​in a certain area.

5. A complex terrain path planning method according to claim 4, characterized in that: The terrain features of each triangle are converted into terrain costs, specifically: Assign weights to the slope, steepness, and unevenness, respectively. , , , calculate the terrain cost of each triangle, the calculation formula is as follows: Among them, Cost is the terrain cost; = 1; , , They represent the maximum slope, maximum steepness, and maximum roughness of the terrain that the robot can pass through.

6. The complex terrain path planning method according to claim 1, characterized in that: The terrain cost of each triangle is combined with the Double DQN algorithm to plan the best moving path for the robot, including: Input the terrain cost of each triangle into the Double DQN algorithm, and add the terrain cost attribute to the robot state after combining the terrain cost; Through the behavior network in the Double DQN algorithm, the next action is selected according to the current action; after each action is executed, the next action is entered from the current action and the reward given by the environment is obtained, which is obtained according to the terrain cost; The target network in the Double DQN algorithm evaluates the Q value of the next action to reduce the overestimation of the Q value. Get the action set with the highest cumulative reward value and plan the best movement path for the robot.

7. A complex terrain path planning method according to claim 6, characterized in that: The Q value of the next action is evaluated by the target network in the Double DQN algorithm to reduce the over-estimation of the Q value, including: The overestimation of the behavior network and the target network is adjusted by the loss function, which is as follows: Among them: s is the current state; is the chosen action; r is the reward obtained from the environment; s′ is the next state; is the discount factor; is the Q value of the behavior network, with parameter θ; is the Q value of the target network, with parameter θ′; is the next action selected from the behavior network.

8. A path planning system for complex terrain, characterized in that: Includes the following modules: The map model building module is used to obtain terrain data of complex terrain and build a map model based on irregular triangulated networks; A terrain cost calculation module, used to calculate the terrain features of each triangle based on the map model, and then convert the terrain features of each triangle into a terrain cost; The path planning module is used to plan the best moving path for the robot based on the terrain cost of each triangle and combined with the Double DQN algorithm.

9. A computer-readable storage medium storing one or more programs, characterized in that: The one or more programs include instructions, which, when executed by a computing device, cause the computing device to perform the steps of the complex terrain path planning method according to any one of claims 1 to 7.

10. A terminal device, characterized in that: include: One or more processors, a memory and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, and the one or more programs include steps for executing the path planning method for complex terrain as described in any one of claims 1 to 7.

Images