A path planning method for a mobile charging robot in an unstructured environment

Abstract

The application relates to a path planning method for a mobile charging robot in an unstructured environment, which combines global path planning and local dynamic obstacle avoidance. The method uses an improved A* algorithm for global path planning and a dynamic window method for real-time local obstacle avoidance. The global and local paths are fused through dynamic weight adjustment to ensure that the robot avoids obstacles while maintaining the continuity of the global path. During path planning, factors such as terrain complexity and robot speed are considered, and an energy optimization algorithm is used to minimize the energy consumption of the path, thereby prolonging the working time of the robot and improving the execution efficiency. The path planning method of the application is suitable for complex dynamic environments such as warehouses, industries and field inspections, and has strong adaptability, real-time performance and robustness, providing an effective solution for the efficient application of mobile charging robots in unstructured environments.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of robot path planning, in particular to a path planning method for a mobile charging robot in an unstructured environment. BACKGROUND

[0002] With the continuous development of automation technology, mobile robots are increasingly used in various fields, especially in complex unstructured environments such as warehouses, industries, and field inspections. Mobile charging robots, as a kind of automation equipment specially used for charging tasks, are responsible for providing charging services for devices and other robots in these environments. In order to improve efficiency and avoid collisions, path planning technology for mobile charging robots is particularly important.

[0003] However, existing path planning techniques still have some deficiencies and cannot fully meet the needs of unstructured environments, mainly in the following aspects:

[0004] Most existing path planning algorithms, such as A* algorithm and Dijkstra algorithm, are mainly applied in structured environments, i.e., the obstacles and target points in the environment are static and known. These algorithms can calculate the optimal path by constructing a pre-defined map (such as a grid map or a topological map), and usually can quickly provide a global path from the starting point to the target point.

[0005] However, these algorithms have significant defects in unstructured environments. The obstacle distribution in unstructured environments is usually irregular and dynamically changing, such as complex terrain in the field, shelf layout in the warehouse environment, and dynamic obstacles such as pedestrians and vehicles that may appear at any time in the city. Existing path planning methods cannot effectively handle dynamic changes in these environments, resulting in poor real-time performance and adaptability of path planning.

[0006] Although there are some solutions for dynamic obstacle avoidance, such as dynamic window method and artificial potential field method, these algorithms can perceive the surrounding environment in real time and react to avoid collisions between robots and obstacles. However, existing local obstacle avoidance methods have the following shortcomings:

[0007] Local optimal problem: Many dynamic obstacle avoidance algorithms are prone to local optimal solutions, causing the robot to deviate from the global optimal path during obstacle avoidance, ultimately resulting in reduced path efficiency.

[0008] High computational complexity: Real-time environmental perception and decision-making calculations are required during dynamic obstacle avoidance, especially in situations where dynamic obstacles are dense or the environment changes significantly. Existing algorithms often have difficulty making decisions within a short period of time.

[0009] Path discontinuity or non-smoothness: Existing obstacle avoidance algorithms lack in handling path smoothness, especially when multiple obstacles exist or the robot needs to frequently change its direction of travel, leading to discontinuous or excessively steep paths, increasing energy consumption.

[0010] In the path planning of mobile charging robots, energy efficiency is an important consideration. Existing path planning algorithms usually focus on the shortest distance or minimum time of the path, ignoring the energy efficiency optimization of the path. When a mobile robot performs a charging task, the battery life is directly affected by energy consumption. Existing technologies often fail to comprehensively optimize the energy consumption of the path in combination with terrain and motion parameters, resulting in the inability of the robot to minimize energy consumption when performing tasks.

[0011] Insufficient consideration of terrain energy consumption: The slope and roughness of the terrain have a significant impact on the energy consumption of the robot's movement, but existing path planning methods often lack accurate energy consumption models for different terrains.

[0012] Insufficient dynamic energy consumption adjustment: Existing algorithms often lack the ability to adjust energy consumption in real time, and cannot automatically optimize the energy efficiency of the path according to environmental changes during task execution, leading to energy waste.

[0013] Current path planning methods usually separate global path planning and local obstacle avoidance planning. Global path planning focuses on the calculation of the globally optimal path, while local obstacle avoidance planning handles real-time dynamic obstacle avoidance. However, the coordination between the two has the following problems:

[0014] Difficulty in merging global and local paths: Existing path merging algorithms are mostly based on path correction under static conditions. When encountering dynamic obstacles, they cannot effectively combine the globally optimal path with the local obstacle avoidance path, affecting the smoothness and efficiency of the path.

[0015] Path deviation problem: During local obstacle avoidance, the robot may deviate from the predetermined trajectory of the global path, causing the path to be lengthy or inefficient, and this deviation is difficult to adjust effectively.

[0016] Poor adaptability: Most existing path planning algorithms perform poorly in dynamic environments, especially when real-time responses to environmental changes are required. Existing technologies often fail to meet the real-time requirements of robots.

[0017] Poor robustness: Factors such as the type, distribution, and speed of change of obstacles in the environment can affect the robustness of path planning. Existing technologies may fail or be inefficient when faced with complex and uncertain environments.

[0018] Therefore, we urgently need to design a path planning method for mobile charging robots in unstructured environments to solve the above problems. SUMMARY

[0019] The object of the present application is to solve the technical problems raised in the background art, and to provide a path planning method for mobile charging robots in unstructured environments.

[0020] The scheme for achieving the above object of the present application is as follows:

[0021] A path planning method for mobile charging robots in unstructured environments, the method comprising the following steps:

[0022] 1) Environment modeling: Collect static and dynamic data in the unstructured environment through laser radar and camera, generate a two-dimensional grid map or a three-dimensional point cloud map, and mark the obstacle distribution and terrain features;

[0023] 2) Global path planning: Calculate the global optimal path from the starting point to the target point based on the improved A* algorithm, and the cost function of the path is:

[0024]

[0025] Wherein, represents the total cost from the starting point to the target node; represents the actual cost of the path from the starting point to the current node; represents the heuristic estimated cost from the current node to the target point; represents the complexity weight of the terrain where the current node is located;

[0026] 3) Local path planning: Based on the dynamic window method, combine real-time sensor data to generate a dynamic obstacle avoidance path, and the local path cost function is:

[0027]

[0028] Wherein: velocity; angular velocity;

[0029] represents the degree of the path towards the global target point

[0030] : represents the obstacle avoidance clarity, which is the minimum distance between the obstacle detected by the sensor and the path;

[0031] : represents the local path cost, combined with the terrain complexity weight of the current node ;

[0032] Fixed weight coefficient, satisfying ;

[0033] 4) Path fusion: the global path and the local path are combined by the path fusion algorithm, and the path fusion model is:

[0034]

[0035] wherein, represents the comprehensive cost of the final path; represents the global path weight;

[0036] 5) Path optimization: on the basis of path planning, the final path is optimized according to the energy consumption model, and the optimization function is:

[0037]

[0038] wherein, : motion energy consumption; : terrain energy consumption; T: total time of path planning; represents the total energy consumption of the path.

[0039] As a preferred technical solution of the present application, the terrain complexity weight of the global path planning is calculated by the following formula:

[0040]

[0041] wherein, represents the complexity weight of the terrain where the current node is located; represents the slope value of the current node; is the slope critical value; represents the obstacle density in the node neighborhood; is the terrain complexity adjustment coefficient.

[0042] As a preferred technical solution of the present application, the terrain cost function of the local path is calculated by the following formula:

[0043]

[0044] wherein, represents the terrain cost of the local path; is the linear velocity of the current node; is the angular velocity of the current node; is the terrain complexity weight of the current node, which is consistent with the global path cost; is the weight coefficient of the cost function.

[0045] As a preferred technical solution of the present application, the dynamic weight in the path fusion model is The calculation formula is:

[0046] ​​ ;

[0047] wherein, represents a global path weight, used to combine the global path cost and the local cost function to generate the final path cost; represents the distance from the current position of the robot to the target point; represents the distance threshold of path adjustment; represents the smoothing parameter of weight adjustment.

[0048] As a preferred technical solution of the present application, the motion energy consumption The calculation formula is:

[0049] ;

[0050] wherein, represents the motion energy consumption; represents the comprehensive cost of the final path.

[0051] As a preferred technical solution of the present application, the terrain energy consumption is calculated by the following formula:

[0052] ;

[0053] wherein, represents the slope value of the current node represents the terrain roughness; is a fixed coefficient for energy consumption calculation.

[0054] As a preferred technical solution of the present application, the path planning method is optimized by reinforcement learning, and the reward function of reinforcement learning is:

[0055] ;

[0056] wherein, , wherein is the obstacle density, is the collision penalty coefficient; , wherein is the target direction cost, is the progress reward coefficient, E.

[0057] As a preferred technical solution of the present application, the final path is processed by a path smoothing optimization algorithm, and the objective function of the path smoothing optimization is:

[0058] ;

[0059] wherein, Smooth cost of the path, the smaller the value, the smoother the path; Second-order derivatives of the path in and directions, used to measure the curvature change of the path; Total length of the path; by minimizing , ensure the smoothness of the path in the sparse obstacle area, while retaining the local obstacle avoidance ability.

[0060] As a preferred technical solution of the application, the path planning balances energy consumption, time and obstacle avoidance safety through a multi-objective optimization model, and the objective function of the multi-objective optimization model is:

[0061] ;

[0062] Wherein, represents the total optimization target, which comprehensively considers the energy consumption, time and safety of the path; represents the total energy consumption of the path; represents the path travel time, calculated by the following formula:

[0063] ;

[0064] Wherein is the linear velocity of the current node;

[0065] represents the obstacle avoidance safety of the path, calculated by the following formula:

[0066] ;

[0067] Wherein is the local path obstacle density corresponding to the current node; is the weight coefficient of the objective function, satisfying , which is dynamically adjusted according to the specific scene requirements.

[0068] As a preferred technical solution of the application, the path planning method is suitable for the following scenes:

[0069] Unstructured warehouse environment: mobile charging robot dynamically avoids obstacles between shelves and completes charging tasks;

[0070] Complex field environment: robot plans efficient energy consumption path in different slope and rough terrain, such as charging task in wind power plant equipment inspection;

[0071] Urban dynamic environment: robot plans safe path and completes charging task in urban public places with more pedestrians and vehicles.

[0072] Compared with the prior art, the application has the following beneficial effects:

[0073] The present application solves the problems of poor adaptability and real-time performance of traditional path planning methods in unstructured environments by combining global path planning with local dynamic obstacle avoidance. By using a local path planning algorithm (such as the dynamic window method) for real-time obstacle avoidance based on global path planning, the robot can dynamically respond to environmental changes and ensure that it can quickly and efficiently complete tasks in complex and dynamic environments.

[0074] The present application adds an energy consumption optimization module to the path planning, taking into account factors such as terrain complexity and robot travel speed, to minimize path energy consumption and extend robot working time. In addition, through the path smoothing optimization algorithm, sudden turns and violent movements of the robot path are reduced, making the path smoother and avoiding additional energy consumption due to path unevenness, improving the comfort and stability of robot movement.

[0075] The present application uses a dynamic fusion algorithm of global and local paths to keep the global path as coherent as possible while avoiding obstacles, avoiding the problem of path length or inefficiency caused by local obstacle avoidance. This global and local path fusion method not only improves the robustness of path planning, but also ensures that the robot can efficiently and safely complete tasks in dynamic obstacle-dense environments. BRIEF DESCRIPTION OF DRAWINGS

[0076] In order to more clearly illustrate the technical solutions in the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings described below are some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0077] Figure 1 is a flow chart of a path planning method for a mobile charging robot in an unstructured environment. DETAILED DESCRIPTION

[0078] In order to make the objectives, technical solutions and advantages of the present application clearer, the following will describe the technical solutions in the present application clearly and completely with reference to the drawings in the present application. Obviously, the described embodiments are 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 skilled in the art without creative labor are within the scope of protection of the present application.

[0079] The specific embodiments of the present application will be described in detail below with reference to the drawings and embodiments. Figure 1

[0080] Embodiment 1:​

[0081] Environment modeling and global path planning:

[0082] In unstructured warehouse environments, mobile charging robots collect environmental data through onboard lidar and cameras, generating two-dimensional grid maps and three-dimensional point cloud maps.

[0083] Environment modeling: Use lidar scanning to obtain obstacle positions and surrounding space depth information, generate point cloud data;

[0084] Use cameras to capture terrain images and extract ground features (such as slope, roughness), combine with lidar data to generate grid maps with terrain feature markers.

[0085] Global path planning: Based on improved A* algorithm to calculate the global optimal path from the starting point to the target point.

[0086] The specific implementation is as follows: , Where:

[0087] According to the cumulative travel distance from the current position of the robot to the node; According to the Euclidean distance to predict the cost of the node to the target point; Combined with the terrain slope And obstacle density , According to the formula:

[0088] ;

[0089] Result output: global path After generation, it is stored as a reference path for local planning to call.

[0090] Implementation 2:

[0091] Local path dynamic obstacle avoidance and fusion:

[0092] During the execution of the charging task by the robot, real-time sensor collection obtains the position information of dynamic obstacles (such as pedestrians, moving goods), and generates a dynamic obstacle avoidance path in combination with local path planning.

[0093] Use dynamic window method to calculate local path: define search range according to speed And angular velocity ;

[0094] Comprehensive obstacle avoidance and terrain cost calculation: ;

[0095] Where: Adjust the local direction according to the global path direction cost ; : Calculate the distance between the path and the nearest obstacle by sensor detection; : Combine the terrain complexity weight Calculate the local path terrain cost; Parameters Determined according to the application scenario.

[0096] Path fusion: fuse the global path and the local path to calculate the final path cost:

[0097] ;

[0098] Where, Dynamic adjustment, specific formula:

[0099] ;

[0100] The distance from the current position to the target point; The adjustment threshold.

[0101] Embodiment 3: Path optimization and energy minimization, after the robot performs path planning, further optimize the path to reduce energy consumption.

[0102] Motion energy optimization: according to the formula:

[0103] ;

[0104] Combine the path comprehensive cost , dynamically adjust the robot speed and angular velocity , without affecting the safety of obstacle avoidance, try to reduce the motion energy consumption.

[0105] Terrain energy optimization: combine the path terrain complexity weight , calculate the terrain energy consumption:

[0106] ;

[0107] Smooth optimization: adjust the path planning according to the slope and roughness , select the area with lower energy consumption as the priority path.

[0108] Use quadratic curve fitting to smooth the path to ensure the continuity and feasibility of robot motion:

[0109] ;

[0110] Minimize to ensure the smoothness of the path.

[0111] The following will be described in detail in combination with several embodiments:

[0112] Example 1: In a typical warehouse environment, a robot needs to charge multiple devices parked between shelves. The environment contains dynamic obstacles (such as workers and forklifts) and static obstacles (such as shelves and walls). The robot needs to avoid these obstacles while finding the shortest path to complete the charging task.

[0113] Step 1: Environment Modeling. Use LiDAR to scan the environment and generate a 2D grid map representing the locations of shelves, walls, and other static obstacles. Use cameras and depth sensors to detect dynamic obstacles (such as moving forklifts or workers) and update the map in real time. In this environment, the robot starts from a starting point (charging station) and aims to reach the charging location of a specific device.

[0114] Global path planning: A modified A* algorithm is used for global path planning, and path costs are calculated. ;

[0115] Based on the actual cost from the starting point to the current node, taking into account factors such as ground resistance and obstacles;

[0116] Predicting the cost of reaching the target point using Euclidean distance or Manhattan distance;

[0117] Calculate the terrain complexity weight for each node. If a node is near an obstacle, As the size increases, the path cost also increases.

[0118] The path generated after global path planning follows the shortest path principle while avoiding all static obstacles.

[0119] Step 2: Local Path Planning: Local path planning comes into play when the robot begins to move. The robot detects dynamic obstacles (such as the trajectory of a forklift) using real-time sensor data (such as LiDAR, ultrasonic sensors, etc.).

[0120] Local obstacle avoidance paths are generated using the Dynamic Window Method (DWA), and the cost of these local paths is... include: The cost of adjusting the direction of the current path, based on the global path direction. Adjustment; Obstacle avoidance clarity is determined by detecting the minimum distance between the robot and obstacles based on sensor data. The terrain cost of the path is based on the current terrain complexity weight. .

[0121] Local path fusion: Local paths and global paths are combined using a fusion algorithm to form a smooth path. Path fusion cost function. Balancing global and local paths through dynamically adjusted weights

[0122]

[0123] where Dynamic adjustment according to the distance between the robot's current position and the target point ensures rapid adjustment of the path when encountering dynamic obstacles.

[0124] Step 3: Path optimization: the robot optimizes its path according to the total energy consumption The total energy consumption of the path is composed of motion energy and terrain energy :

[0125]

[0126] Motion energy Consider the speed and angular velocity of the robot on the path, optimize the speed to reduce energy consumption;

[0127] Terrain energy Optimize the path according to the terrain slope and roughness to avoid passing through high slope or complex terrain areas to reduce energy consumption.

[0128] Path smoothing optimization: use B-spline curves for path smoothing to reduce the robot's sharp motion when turning, further optimize the continuity and comfort of the path.

[0129] Example 2: The robot performs a charging task in a wind farm while conducting equipment inspection. The environment is complex, containing undulating terrain, stones, grass obstacles, and dynamic obstacles such as moving maintenance personnel and vehicles.

[0130] Step 1: Environment modeling and path planning:

[0131] Environment modeling: generate a three-dimensional point cloud map using laser radar and vision sensors, and identify slope and terrain roughness through terrain analysis models to provide terrain information for path planning. Global path planning: use an improved A* algorithm for global path planning to calculate path cost .

[0132] When planning the path, focus on avoiding high slope areas to reduce energy consumption:

[0133]

[0134] where represents the slope, is the obstacle density.​​​​

[0135] Step 2 Local Path Planning: During travel, the robot detects obstacles in real-time (such as a maintenance vehicle on the road) through dynamic sensor data. A temporary obstacle-avoiding path is generated using the Dynamic Window Approach (DWA), and its motion trajectory is optimized.

[0136] Step 3 Path Optimization: The robot optimizes the path in combination with the energy consumption model, minimizing the total energy consumption of the path . The motion energy consumption is calculated according to the current speed and angular velocity of the robot, and the terrain energy consumption is calculated according to the terrain.

[0137] Example 3: In a busy city square, a service robot needs to charge an electric scooter. There are many dynamic obstacles (such as pedestrians, vehicles) in the environment, and path planning needs to consider complex obstacle avoidance situations.

[0138] Step 1 Environment Modeling: Real-time environment modeling is performed using laser radar and vision sensors to identify pedestrians, vehicles, and other dynamic obstacles. Global Path Planning: Global path planning is performed using an improved A algorithm to calculate the path cost . Optimization is performed in combination with obstacle density and terrain complexity to ensure that the global path avoids busy areas as much as possible.

[0139] Step 2 Local Path Planning: When the robot travels and encounters dynamic obstacles (such as pedestrians and moving vehicles), the local path is adjusted through real-time sensor data.

[0140] Local Path Fusion: The cost function of the local path combines obstacle avoidance clarity and target direction adjustment to generate a dynamic obstacle avoidance path.

[0141] The local path is combined with the global path to form the final path, and the dynamic weight is adjusted.

[0142] Step 3 Path Optimization: Optimize the total energy consumption of the path , calculate the optimal path in combination with motion energy consumption and terrain energy consumption .

[0143] Path Smoothing Optimization: Use path smoothing algorithms to process the path to ensure the smoothness of the robot's motion, avoiding sudden turns and energy waste.

[0144] It should be pointed out finally that the above embodiments are only used to illustrate the technical solutions of the present application, but not to limit the same; and although the present application has been described in detail with reference to the foregoing embodiments, it should be appreciated by those skilled in the art that the technical solutions recorded in the foregoing embodiments can be modified, or some technical features thereof can be replaced equivalently; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for path planning of a mobile charging robot in an unstructured environment, characterized in that, The method comprises the following steps: 1) Environment modeling: Collect static and dynamic data in unstructured environment through laser radar and camera, generate two-dimensional grid map or three-dimensional point cloud map, mark obstacle distribution and terrain features; 2) Global path planning: the global optimal path from the start point to the target point is calculated based on the improved A* algorithm, and the cost function of the path is: ; wherein, represents the total cost from the start node to the goal node; represents the actual cost of the path from the start node to the current node; represents the heuristic estimated cost from the current node to the goal node; represents the complexity weight of the terrain where the current node is located; the terrain complexity weight of the global path planning is calculated by the following formula: ; wherein, represents a complexity weight of the terrain in which the current node is located; represents a slope value of the current node; is a slope critical value; represents the density of obstacles within the neighborhood of the node; is a terrain complexity adjustment coefficient; 3) Local path planning: Based on the dynamic window method, combined with real-time sensor data, a dynamic obstacle avoidance path is generated, and the local path cost function is: ; wherein: speed; angular velocity; represents the degree to which the path is heading towards the global goal point; : represents the obstacle avoidance clarity, the minimum distance between the obstacle and the path detected by the sensor; : represents the terrain cost of the local path, combined with the terrain complexity weight of the current node ; the terrain cost of the local path is calculated by the following equation: ; wherein, represents a terrain cost of a local path; is a linear velocity of the current node; is an angular velocity of the current node; is a terrain complexity weight of the current node, consistent with the global path cost; is a weight coefficient of the cost function; The fixed weight coefficient satisfies ; 4) Path fusion: the global path and the local path are combined by the path fusion algorithm, and the path fusion model is: ; wherein, represents the overall cost of the final path; represents the global path weight; the global path weight The calculation formula is: ; wherein, represents a global path weight, used in combination with the total cost from the start node to the goal node and a local cost function to generate a final path cost; represents the distance from the robot's current position to the goal point; represents a distance threshold for path adjustment; represents a smoothing parameter for weight adjustment; 5) Path optimization: On the basis of path planning, the final path is optimized according to the energy consumption model, and the optimization function is: ; wherein, : motion energy consumption; : terrain energy consumption; T: total time of path planning; represents the total energy consumption of the path; motion energy consumption The calculation formula is: ; wherein, represents the energy cost of motion; represents the overall cost of the final path; Terrain energy consumption Calculated from the following equation: ; wherein, represents a slope value of the current node; represents a terrain roughness; is a fixed coefficient for the energy consumption calculation.

2. The path planning method for a mobile charging robot in an unstructured environment according to claim 1, wherein, The path planning method is optimized by reinforcement learning, and a reward function of the reinforcement learning is: ; wherein wherein is an obstacle density, is a collision penalty coefficient; wherein is a target direction cost, is a progress reward coefficient, E.

3. The path planning method for a mobile charging robot in an unstructured environment according to claim 1, wherein, The final path is processed by a path smoothing optimization algorithm, and an objective function of the path smoothing optimization is: ; wherein, represents the smoothness cost of a path, the smaller the value, the smoother the path; are the second-order derivatives of the path in and directions, respectively, used to measure the change of path curvature; is the total length of the path; by minimizing , the smoothness of the path in the sparse obstacle area is ensured, while the local obstacle avoidance ability is retained.

4. The path planning method for a mobile charging robot in an unstructured environment according to claim 1, wherein, The path planning balances energy consumption, time and obstacle avoidance safety through a multi-objective optimization model, and objective functions of the multi-objective optimization model are: ; wherein, represents the total optimization target, which comprehensively considers the energy consumption, time and safety of the path; represents the total energy consumption of the path; represents the path travel time, which is calculated by the following formula: ; wherein is the linear velocity of the current node; represents the obstacle avoidance safety of the path, calculated by the following equation: ; wherein is the corresponding local path obstacle density; is the weight coefficient of the objective function, satisfying , which is dynamically adjusted according to the specific scene requirements.

5. The path planning method for a mobile charging robot in an unstructured environment according to claim 1, wherein, The path planning method is suitable for the following scenarios: Unstructured warehouse environment: mobile charging robot dynamically avoids obstacles between shelves and completes charging task; Complex outdoor environment: robot plans high energy consumption path in different slope and rough terrain; Urban dynamic environment: robot plans safe path and completes charging task in urban public places with many pedestrians and vehicles.

Images