Intelligent navigation planning method and system for mobile robot

Abstract

The invention relates to the technical field of robot autonomous navigation, in particular to an intelligent navigation planning method and system for a mobile robot, and the method comprises the following steps: generating a global path based on a global path planning algorithm, and optimizing and adjusting the global path by using a local path optimization algorithm in combination with real-time environment information, the global path is adjusted in real time to avoid dynamic obstacles, and an optimized path is obtained; the problems that in the prior art, a path planning scheme is poor in adaptability in a dynamic environment, the real-time requirement is difficult to meet, and a strategy cannot be adaptively adjusted to plan an optimal path are solved. The system comprises an environment sensing module, a global path planning module, a local path optimization module and a robot control module, information interaction among the modules is realized through a standardized interface, and the path planning efficiency and navigation performance in a dynamic environment are remarkably improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of robot autonomous navigation, and in particular relates to an intelligent navigation planning method and system for a mobile robot. Background Art

[0002] With the rapid development of artificial intelligence technology and robot autonomous navigation technology, mobile robots have been widely used in logistics, security, medical and other fields. At present, the mobile robot path planning technology in domestic and foreign industries is mainly divided into two categories: rule-based path planning method and machine learning-based path planning method.

[0003] Rule-based path planning methods usually use classic algorithms, such as the Dijkstra algorithm and the A* algorithm. The above algorithms aim at global optimization and perform path planning based on a pre-known static environment map, which can guarantee the feasibility and optimality of the planned path to a certain extent. However, such methods have poor adaptability in dynamic environments. When dynamic obstacles appear in the environment or the path changes, the path often needs to be replanned, resulting in a long calculation time. In addition, high-dimensional space cannot be processed. When the robot needs to navigate in a high-dimensional state space (such as a complex 3D scene), the computational complexity of the algorithm increases rapidly, making it difficult to meet real-time requirements. At the same time, such algorithms lack intelligence. Traditional algorithms rely on fixed rules and global information and cannot adaptively adjust strategies or optimize paths.

[0004] At the same time, machine learning-based methods also have limitations. With the development of reinforcement learning technology, deep reinforcement learning is widely used in robot path planning, such as deep Q learning (DQN) and proximal policy optimization (PPO) algorithms. Such methods enable robots to complete navigation tasks in unknown or dynamic environments by exploring and learning the environment. This method also has the following disadvantages:

[0005] 1. Slow convergence: Existing reinforcement learning algorithms often require a lot of training time to converge when dealing with complex environments, resulting in high development costs.

[0006] 2. Low exploration efficiency: Since reinforcement learning relies on random sampling, it may lead to repeated exploration of low-value areas and prolong training time.

[0007] 3. Lack of optimality of planned paths: Reinforcement learning algorithms are often difficult to match the global optimal capabilities of traditional algorithms and are prone to produce suboptimal paths.

[0008] Based on the above analysis, it is difficult for the existing technology to simultaneously meet the requirements of dynamic environment adaptability, real-time performance, high-dimensional environment planning capability and global path optimality. In view of this, the present invention is proposed. Summary of the invention

[0009] The purpose of the present invention is to provide an intelligent navigation planning method and system for a mobile robot, thereby solving the problems in the prior art that the path planning scheme has poor adaptability in a dynamic environment, is difficult to meet real-time requirements, and cannot adaptively adjust the strategy to plan the optimal path.

[0010] The objective of the present invention is achieved through the following technical solutions:

[0011] An intelligent navigation planning method for a mobile robot, comprising:

[0012] Collecting environmental information in real time, the environmental information includes static map information and real-time environmental information;

[0013] Planning a global path using a global path planning algorithm according to the static map information;

[0014] In combination with the real-time environmental information, a local path optimization algorithm is used to optimize and adjust the global path, and the global path is adjusted in real time to avoid dynamic obstacles to obtain an optimized path;

[0015] A control instruction is generated according to the optimized path, and the robot is driven to move along the optimized path by the control instruction.

[0016] The global path optimization adjustment adopts the soft actor critic algorithm, and the processing process includes:

[0017] Inputting the global path and the real-time environment information contained in the environment information;

[0018] Define a state space S and an action space A, and determine information contained in the state space S and the action space A according to the environmental information;

[0019] Based on the information contained in the state space S and the action space A, a dynamic reward function is determined; the dynamic reward function is used to update the relationship between the robot and the dynamic obstacle; wherein the dynamic reward function includes: rewarding the behavior close to the target point, punishing the behavior close to the dynamic obstacle and punishing the behavior deviating from the global path; updating the path planning strategy based on the above dynamic reward function combined with the soft actor critic algorithm; locally optimizing the global path based on the updated path planning strategy to obtain an optimized path.

[0020] The action space A includes: the robot's selectable movement direction and speed; the state space S includes: the robot's current position, speed and local environment characteristics.

[0021] The dynamic reward function is determined according to the environmental information. The specific dynamic reward function is calculated based on the following formula:

[0022] R(s,a)=α1·d target +α2·d obstacle +α3·d A*

[0023] Among them, d target is the distance from the robot to the target point, d obstacle is the distance from the robot to the nearest obstacle, d A* is the offset distance from the robot to the global path, α1, α2, and α3 are weight parameters; the weight parameters are set and adjusted according to the path planning requirements.

[0024] The updating of the path planning strategy includes strategy network optimization and value function optimization;

[0025] Strategy network optimization: maximize the objective function, which is expressed as: J(π) = E st，at [R(s, a)];

[0026] Among them, E is the expected function, st is the current state of the environment, at is the action selected according to the strategy π in the state st, and R(s, a) is the dynamic reward function;

[0027] Value function optimization: Minimize the Bellman residual and update the parameters of the policy network.

[0028] The real-time environmental information includes: road condition information, dynamic obstacle information and robot motion status.

[0029] The global path planning adopts the A-star algorithm.

[0030] The static map information includes: the current position of the robot, the target position and the global environment map; the control instructions include: navigation instructions, avoidance priority instructions and energy-saving mode instructions.

[0031] The intelligent navigation planning method further includes a navigation experience optimization strategy, and the navigation experience optimization strategy includes:

[0032] After completing each task, the robot stores the state, action and reward information of the path planning strategy process into the shared experience pool, and gradually optimizes the strategy network of the soft actor-critic algorithm.

[0033] An intelligent navigation planning system for a mobile robot, comprising:

[0034] Environmental perception module: used to collect environmental information in real time, including static map information and real-time environmental information;

[0035] Global path planning module: used to receive static map information sent by the environment perception module and generate a global path through a global path planning algorithm;

[0036] Local path optimization module: used to receive real-time environmental information sent by the environmental perception module, use the local path optimization algorithm to optimize and adjust the global path, adjust the global path in real time to avoid dynamic obstacles, and obtain an optimized path;

[0037] Robot control module: used to receive the optimized path sent by the local path optimization module and the real-time environmental information sent by the environmental perception module, generate control instructions according to the optimized path, and drive the robot to move along the optimized path through the control instructions.

[0038] The environment perception module includes: a laser radar and a depth camera, which are used to collect static obstacle information, dynamic obstacle information and the robot's own status information around the robot.

[0039] Compared with the prior art, the intelligent navigation planning method and system for a mobile robot provided by the present invention, in the specific application process, combines the A* algorithm and the SAC algorithm to realize global path planning and real-time optimization of dynamic environment, effectively improving the navigation accuracy and obstacle avoidance ability of the mobile robot in complex environments. It also adjusts the path planning through the reinforcement learning strategy, effectively overcoming the limitations of the traditional path planning method in dynamic environments, and improving the flexibility and navigation efficiency of the robot in practical applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0041] Figure 1 A schematic diagram of an intelligent navigation planning method provided by an embodiment of the present invention;

[0042] Figure 2 A schematic diagram of the architecture of an intelligent navigation planning system provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0043] The technical solution of the present invention will be clearly described below in conjunction with the accompanying drawings. Obviously, the described embodiments are not all embodiments of the present invention, and all other embodiments obtained by ordinary technicians in the field without making creative work are within the protection scope of the present invention.

[0044] It should be noted that, unless otherwise specifically stated, the relative arrangements of components and steps, and numerical expressions set forth in these embodiments should not be construed as limiting the scope of the present invention.

[0045] The following description of the exemplary embodiments is merely illustrative and is not intended to limit the present invention and its application or use in any sense. Techniques, methods and devices known to ordinary technicians in the relevant field may not be discussed in detail here, but where applicable, these techniques, methods and devices should be considered as part of this specification.

[0046] Explanation of professional terms:

[0047] A* algorithm: the full name is A-Star algorithm;

[0048] SAC algorithm: The full name is Soft Actor-Critic algorithm.

[0049] Embodiment 1

[0050] See also Figure 1 , a schematic diagram of an intelligent navigation planning method for a mobile robot proposed in an embodiment of the present invention, the specific implementation process of which may include the following steps:

[0051] A1. Real-time collection of environmental information

[0052] Environmental information is collected in real time through sensor equipment. The environmental information includes static map information and real-time environmental information. The static map information includes: the robot's current position, target position and global environmental map; the real-time environmental information includes: road condition information, dynamic obstacle information and robot motion status. The specific implementation steps for real-time collection of environmental information are as follows:

[0053] A11. Collect static map information and real-time environmental information of the surrounding environment through sensors in real time;

[0054] A12. Update the real-time environment information in real time by combining dynamic obstacle information and robot motion status to ensure the real-time and accuracy of the environment data when planning the path;

[0055] A13. Integrate static map information and real-time environmental information to generate multi-level environmental perception data to provide support for global planning and local optimization.

[0056] A2. Global Path Planning

[0057] According to the static map information, the A* algorithm is used to plan the global path. As a classic heuristic global path planning method, the A* algorithm can provide a global path based on the static map information. The A* algorithm is based on heuristic search and can quickly find the global optimal path. Through the planning of the A* algorithm, it can be ensured that the robot can reach the target point in the shortest path without considering dynamic obstacles. The specific operations are as follows:

[0058] A21. constructing a grid map according to the static map information, and defining the traversability and cost function of each unit included in the grid map;

[0059] A22, using a heuristic search strategy to calculate the cost value f(n)=g(n)+h(n) of each node in the grid map;

[0060] Among them, g(n) is the actual cost from the starting point to node n; h(n) is the estimated cost from node n to the target point;

[0061] A23, gradually expand from the starting point to the target point, generate a global optimal path, and mark the path nodes to form an initial path guide as the global path obtained by planning;

[0062] A24. Input the planned global path result into the local optimization module of the next stage as a reference.

[0063] A3. Local Path Optimization

[0064] By combining the environmental information, the global path is locally optimized and adjusted using a local path optimization algorithm, and the path is adjusted in real time to avoid dynamic obstacles to obtain an optimized path; the global path optimization adjustment may adopt a SAC algorithm; and the specific implementation steps are as follows:

[0065] A31, input the global path and real-time environment information planned by the A* algorithm;

[0066] A32. Define the state space S and action space A of reinforcement learning, where:

[0067] The state space S includes: the robot's current position, speed and local environment characteristics;

[0068] The action space A includes: the movement direction and speed that the robot can select;

[0069] The information included in the state space S and the action space A is determined according to the real-time environment information.

[0070] A33. Based on the information contained in the state space S and the action space A, determine a dynamic reward function; the dynamic reward function is used to update the relationship between the robot and the dynamic obstacle; wherein the dynamic reward function includes: rewarding the behavior of approaching the target point, punishing the behavior of approaching the dynamic obstacle, and punishing the behavior of deviating from the global path. The formula of the dynamic reward function is:

[0071] R(s,a)=α1·d target +α2·d obstacle +α3·d A*

[0072] Among them, d target is the distance from the robot to the target point, d obstacle is the distance from the robot to the nearest obstacle, d A* is the offset distance from the robot to the global path, and α1, α2, and α3 are weight parameters. The weight parameters can be set and adjusted according to the path planning requirements.

[0073] A34. Based on the above dynamic reward function combined with the SAC algorithm, the path planning strategy is updated; based on the updated path planning strategy, the global path is locally optimized to obtain the optimized path; the path planning strategy update includes strategy network optimization and value function optimization; strategy network optimization: maximize the objective function, which is expressed as: J(π) = E st，at [R(s, a)];

[0074] Among them, E is the expected function, st is the current state of the environment, at is the action selected according to the strategy π in the state st, and R(s, a) is the dynamic reward function;

[0075] The policy network optimization goal is to learn a policy π(a|s), that is, to output the probability distribution of the robot's choice of different actions a in any state s. Through this policy, the robot can choose the optimal action at each time step, thereby maximizing the cumulative reward of the entire path planning task. The optimization of the policy network directly determines the robot's real-time decision-making ability in local path planning, especially to achieve efficient path optimization in a dynamic environment.

[0076] Value function optimization: Minimize the Bellman residual, the parameters of the policy network;

[0077] The goal of value function optimization is to learn an action value function Q(s, a), that is, the expected value of the cumulative reward that may be obtained in the future after selecting action a in state s. The value function provides guidance for the training of the policy network, helping the policy network to more clearly evaluate the pros and cons of each action during the optimization process, avoiding blind exploration, and thus improving the learning efficiency and stability of the strategy.

[0078] Through the synergy of policy network optimization and value function optimization, it can dynamically adapt to complex environments and guide the robot to make efficient real-time decisions in local path planning.

[0079] A35. Generate new motion control instructions for the robot according to the optimized path, so as to guide the robot to effectively avoid dynamic obstacles during movement.

[0080] The local path optimization process provided in step A3 solves the problem of poor adaptability of the A* algorithm in dynamic environments through dynamic optimization, significantly improving the real-time and robustness of path planning. By combining the advantages of the A* algorithm and the SAC algorithm, the path can be adjusted in real time to adapt to complex and changing environments.

[0081] Furthermore, in the above processing, the corresponding A* algorithm is responsible for global path planning and quickly generates the initial path to provide global guidance; the corresponding SAC algorithm achieves local optimization and dynamic obstacle avoidance through reinforcement learning. In this way, the A* algorithm can be used to reduce the inefficiency of random exploration, while the SAC algorithm can also be used to perceive and optimize the real-time environmental information in real time, thereby further improving the response speed of path planning while ensuring the rationality of the global path. Therefore, the combination of the A* algorithm and the SAC algorithm effectively overcomes the limitation of the slow response of traditional methods in dynamic environments, thereby meeting the needs of mobile robots for real-time and adaptable navigation.

[0082] A4. The robot executes control instructions

[0083] Generate control instructions according to the optimized path, and drive the robot to move along the optimized path through the control instructions; also include receiving real-time environmental information, and generating emergency instructions according to the real-time environmental information; the real-time environmental information includes: road condition information, obstacle location and robot movement status, etc.; the road condition information includes: ground condition and ground inclination and other detailed information. The specific implementation steps are as follows:

[0084] A41, by receiving the key nodes in the optimized path and the robot state information, converting the path points into the target pose points;

[0085] A42. Based on the target position point and the robot kinematic model, calculate the control parameters of the robot's speed and direction, and then generate corresponding control instructions;

[0086] A43, converting the real-time environment information into a series of pose target points, and marking the pose target points as dynamic obstacles when they are repeated with the optimized path pose target points; when the robot cannot avoid the dynamic obstacles only by relying on the optimized path, generating emergency instructions according to the real-time environment information; the emergency instructions include: stop movement instructions, deceleration instructions and other operation instructions;

[0087] A44, sending the control instruction to the bottom driving module, driving the robot to move along the optimized path, and when receiving an emergency instruction, stopping the control instruction and executing the emergency instruction;

[0088] Furthermore, during the robot's movement, the robot's position and path deviation can be monitored in real time to ensure the execution accuracy of path planning.

[0089] A5. Navigation experience optimization strategy

[0090] After completing each task, the robot stores the status, action and reward information of the path planning strategy process into the shared experience pool, gradually optimizes the strategy network of the SAC algorithm, and improves the planning efficiency and dynamic response capabilities of future tasks.

[0091] The above processing process is applicable to various types of mobile robot tasks. The algorithm weight parameters α1, α2, and α3 can be adjusted to adapt to different task requirements, such as efficient navigation, obstacle avoidance priority, or energy-saving mode, thereby improving the versatility and practicality of the algorithm and making it suitable for applications in various industries.

[0092] In summary, the present invention can meet the real-time path planning requirements in complex dynamic environments. Whether it is efficient navigation, obstacle avoidance priority, or energy-saving mode requirements, they can all be achieved by adjusting the algorithm weight parameters. Through the adaptive optimization of reinforcement learning and the guidance of the global path, the efficiency, flexibility and safety of path planning are significantly improved, providing reliable protection for the application of mobile robots in dynamic environments.

[0093] Embodiment 2

[0094] See also Figure 2 , Figure 2 Schematic diagram of the structure of a mobile robot intelligent navigation system proposed in an embodiment of the present invention. The system adopts modular design and uses data interface to realize communication between modules, ensuring real-time information update and decision-making accuracy, and improving the stability and flexibility of the navigation system. Figure 2 As shown, the mobile robot intelligent navigation system may specifically include:

[0095] B1. Environmental perception module

[0096] The environment perception module is used to collect environment information in real time, including static map information and real-time environment information, and send the environment information to the global path planning module, local path optimization module and robot control module. This module collects environment data through a variety of sensors (such as lidar, depth camera, IMU sensor, etc.) to ensure the comprehensiveness and real-time nature of the data; the environment information includes static obstacle information, dynamic obstacle information and the robot's own status information around the robot. Specific functions include:

[0097] B11. Static obstacle information collection: Generate a global static map through environmental scanning, including obstacle location, size and shape;

[0098] B12. Dynamic obstacle information perception: Detect the position, speed, direction and other characteristics of dynamic obstacles to provide support for real-time path optimization;

[0099] B13. Self-status information monitoring: record the robot's current position, speed, direction, etc. to ensure that the path planning is consistent with the robot's actual status.

[0100] The static map information generated by the environmental perception module is sent to the global path planning module through the data interface for planning the global path; the real-time environmental information generated by it is sent to the local path optimization module and the robot control module for real-time optimization of the path and execution of navigation tasks.

[0101] B2. Global path planning module

[0102] The global path planning module is used to receive the environmental information sent by the environmental perception module and generate a global path through a global path planning algorithm; the global path planning module uses the A* algorithm to generate the global path; the specific functions include:

[0103] B21. Constructing a global static grid map based on the environmental information: gridding the environment to define traversable areas and obstacle areas;

[0104] B22. Use heuristic search strategy to plan the path: A* algorithm finds the optimal path from the starting point to the target point by evaluating the cost function f(n)=g(n)+h(n);

[0105] B23. Output global path: Send the global path as a reference path to the local path optimization module to provide a basis for subsequent dynamic adjustment.

[0106] The global path planning module has the characteristics of high computational efficiency and intuitive planning results. It provides global navigation direction guidance for the robot, but does not handle the interference of dynamic obstacles. It needs to be combined with the local path optimization module to complete dynamic obstacle avoidance.

[0107] B3. Local path optimization module

[0108] The local path optimization module is used to receive the environmental information sent by the environmental perception module, and to perform local optimization adjustment on the global path through the local path optimization algorithm, to adjust the global path in real time to avoid dynamic obstacles and obtain the optimized path; the local path optimization module uses the SAC algorithm to perform local optimization adjustment on the global path.

[0109] Specific features include:

[0110] B31, an environment information receiving unit: used to receive the global path and the real-time environment information included in the environment information;

[0111] B32, space definition unit: used to define the state space S and the action space A, and determine the information contained in the state space S and the action space A according to the environmental information;

[0112] B33, dynamic reward function generating unit: used to dynamically establish a dynamic reward function based on the information contained in the state space S and the action space A; the dynamic reward function is used to update the relationship between the robot and the dynamic obstacle; wherein the dynamic reward function includes: rewarding the behavior of approaching the target point, punishing the behavior of approaching the dynamic obstacle and punishing the behavior of deviating from the global path; the dynamic reward function is designed to consider the distance from the robot to the target point, the distance from the robot to the obstacle and the degree of deviation from the robot to the global path, and its formula is as follows:

[0113] R(s,a)=α1·d target +α2·d obstacle +α3·d A*

[0114] Among them, d target is the distance from the robot to the target point, d obstacle is the distance from the robot to the nearest obstacle, d A* is the offset distance from the robot to the global path, and α1, α2, and α3 are weight parameters. The weight parameters can be set and adjusted according to the path planning requirements.

[0115] B34, Policy Update Unit: Used to learn a policy network that can guide the robot to make real-time decisions in complex environments, helping the robot to complete path planning and local optimization in a dynamic environment; guiding the robot to select the optimal action at each time step to maximize the cumulative reward of the entire path planning task; defining the robot's real-time decision-making ability in local path planning, which is used for local path optimization in a dynamic environment.

[0116] It also includes value function optimization, which is used to learn an action value function. The action value function provides guidance to the policy network and is used to assist the training of the policy network. The value function provides guidance to the policy network and helps the policy network to more clearly evaluate the pros and cons of actions during the optimization process.

[0117] B35, optimized path generation unit: based on the updated path planning strategy, locally optimize the global path to obtain an optimized path.

[0118] By adjusting the weight parameters, it can adapt to the needs of different task scenarios and realize flexible optimization of strategies. This module realizes dynamic obstacle avoidance and optimization adjustment of the global path through reinforcement learning, which enhances the dynamic adaptability and environmental responsiveness of the system. At the same time, the introduction of the strategy update unit can more clearly evaluate the pros and cons of a certain action, avoid blind exploration, and improve the learning efficiency and stability of the strategy. More clearly evaluate the pros and cons of a certain action, avoid blind exploration, and improve the learning efficiency and stability of the strategy. More clearly evaluate the pros and cons of a certain action, avoid blind exploration, and improve the learning efficiency and stability of the strategy.

[0119] The above-mentioned global path planning module and local path optimization module realize information exchange through a standardized interface, and the standardized interface supports real-time data update and modular expansion. Among them, the global path planning module and the local path optimization module can run independently, and support the expansion and upgrade of the environmental perception module and the control module to meet different task requirements and scene changes.

[0120] B4. Robot control module

[0121] The robot control module is used to receive the optimized path sent by the local path optimization module and the real-time environmental information sent by the environmental perception module, generate control instructions according to the optimized path, and drive the robot to move along the optimized path through the control instructions; the real-time environmental information includes: road condition information, obstacle location and robot motion status; the road condition information includes: ground condition and ground inclination and other detailed information. Specific functions include:

[0122] B41, real-time environment information analysis unit: used to convert real-time environment information into position and posture target point information; when the position and posture target point is repeated with the optimized path position and posture target point, it is marked as a dynamic obstacle;

[0123] B42, control instruction generation unit: used to calculate the control parameters of the speed and direction of the robot according to the target posture point and the robot kinematic model, and then generate corresponding control instructions;

[0124] B43, emergency instruction generating unit: used to convert the real-time environment information into a series of posture target points, and when the posture target point is repeated with the optimized path posture target point, it is marked as a dynamic obstacle; when the robot cannot avoid the dynamic obstacle only by relying on the optimized path, an emergency instruction is generated according to the real-time environment information; the emergency instruction includes: stop movement instruction and deceleration instruction and other operation instructions;

[0125] B44, driving module execution unit: used to send the control instruction to the underlying driving module, drive the robot to move along the optimized path, and when receiving an emergency instruction, stop the control instruction and execute the emergency instruction.

[0126] B45, Navigation experience optimization strategy unit: After the robot completes each task, it stores the status, action and reward information of the path planning strategy process into the shared experience pool, gradually optimizes the strategy network of the SAC algorithm, and improves the planning efficiency and dynamic response capabilities of future tasks.

[0127] This system adopts modular design and realizes information interaction between modules through standardized interfaces, which has the following advantages:

[0128] 1. Combining the global and local: Combining the A* algorithm with the SAC algorithm to respond to dynamic environments in real time and make up for the limitations of a single algorithm;

[0129] 2. Independent modular design: Each module runs independently, which is convenient for development, debugging and later maintenance, and reduces the complexity of the system;

[0130] 3. Flexible expansion and upgrade: Through modular design and standardized interfaces, new functional modules can be easily added or existing modules can be optimized to meet system expansion needs;

[0131] 4. Rapid deployment capability: The reinforcement learning module reduces training time, significantly shortens the development cycle, and supports rapid deployment and application;

[0132] 5. Real-time and safety: Not only can the robot's behavior be adjusted quickly in complex environments, but it can also make autonomous adjustments even when the path planning is not fully covered or an emergency occurs on the planned path;

[0133] 6. Efficient data processing: The system realizes an efficient closed loop of environmental perception, path planning, strategy optimization and task execution, ensuring real-time information updating and accuracy of path planning.

[0134] Embodiment 3

[0135] This example uses robot path planning in a dynamic warehouse logistics simulation environment to demonstrate the advantages of combining the A* algorithm with the SAC algorithm to achieve efficient and practical path planning and dynamic obstacle avoidance. The following is the specific system construction, algorithm design, operation process and experimental results analysis.

[0136] C1. System construction and initialization

[0137] C11. Hardware equipment

[0138] C111, Mobile Robot: Autonomous navigation mobile robot with precise positioning and flexible movement capabilities. Its chassis integrates multiple degrees of freedom motion wheels to achieve all-round movement. The built-in processor supports real-time algorithm calculation and path optimization functions.

[0139] C112. Sensor equipment: As the environmental perception module, it can include a laser radar and a depth camera; the laser radar provides 360° environmental scanning capability, which is used to detect the distance and position of surrounding obstacles, dynamic obstacle information and the robot's own status information; the depth camera is used to capture the depth information of the environment, especially to provide support for the recognition and speed estimation of dynamic obstacles.

[0140] C113, Communication equipment: Equipped with high-speed wireless communication module, it can realize task sharing and collaborative path planning in multi-robot systems, and support remote control and real-time monitoring.

[0141] C12. Mission scenario

[0142] C121. Site setting: The simulation environment simulates a 100-square-meter dynamic warehouse logistics site, including fixed obstacles (such as shelves and pillars) and dynamic obstacles (such as other mobile robots or workers). The distribution points are randomly distributed, and the dynamic obstacles in the environment move at different speeds and directions to simulate the dynamic complexity in actual warehouse scenarios.

[0143] C122. Task requirements: The robot needs to start from the starting point, accurately avoid fixed obstacles and dynamic obstacles in the venue, and transport the goods safely to the target point. The task objectives and paths may change dynamically with the environment.

[0144] C13. System construction and initialization

[0145] C131, Global Path Planning Module: It is used to construct a static map of the site and grid it into several nodes. It generates a global path based on the heuristic search A* algorithm of the node cost value and provides a global path reference for the task.

[0146] C132, local path optimization module: used to receive the environmental information sent by the environmental perception module, use the SAC algorithm to optimize and adjust the global path, adjust the global path in real time to avoid dynamic obstacles, and obtain an optimized path;

[0147] C133, robot control module: used to receive the optimized path sent by the local path optimization module and the environmental information sent by the environmental perception module, generate control instructions according to the optimized path, and drive the robot to move along the optimized path through the control instructions.

[0148] C2. Operation process

[0149] C21. Task Allocation

[0150] After the system is started, the robot is assigned a target point according to the task requirements. The robot plans the global path through the A* algorithm, obtains the initial path from the starting point to the target point, and passes the path as a reference to the local path optimization module for optimization.

[0151] C22, Path Planning and Optimization

[0152] Planning the global path: The robot completes the initial navigation in the static environment according to the global path provided by the A* algorithm. The global path is passed to the local optimization module and is subsequently adjusted as one of the constraints.

[0153] Local path optimization: The SAC algorithm perceives the real-time environmental information in the environment and identifies the type, location and movement trend of obstacles.

[0154] Dynamically adjust weight parameters in the dynamic reward function: when obstacles are dense or the dynamic environment is complex, increase the obstacle avoidance weight to prioritize the success rate of obstacle avoidance; when the path is relatively smooth, increase the path length optimization weight to ensure driving efficiency.

[0155] SAC generates the optimal action instruction for the current time step through trial and error and updates the global path in real time.

[0156] C23. Task execution and dynamic adjustment

[0157] The robot performs navigation tasks according to the optimized path and uploads the surrounding environment data to the system sharing pool. The data in the shared pool supports real-time collaboration with other robots to avoid task conflicts.

[0158] When it is detected that the path is blocked due to obstacle location or environmental changes, the SAC algorithm re-optimizes the path according to the new environmental status to ensure that the task can be completed continuously and efficiently.

[0159] C24. Feedback and Learning

[0160] After completing each task, the robot stores the status, action and reward information of the path planning process into the shared experience pool, gradually optimizes the strategy network of the SAC algorithm, and improves the planning efficiency and dynamic response capabilities of future tasks.

[0161] C3. Experimental comparison and result analysis

[0162] C31. Experimental conditions

[0163] In a dynamic warehousing scenario, the performance of the traditional A* algorithm and the combination of the A* algorithm and the SAC algorithm provided by the present invention are compared. The experimental scenario is set to 50 simulation experiments, dynamic obstacles move randomly at different speeds, and the task target points and starting points are randomly distributed.

[0164] C32. Experimental indicators

[0165] Task completion time: The average time from task assignment to the robot reaching the target point.

[0166] Obstacle avoidance success rate: the ratio of the robot successfully avoiding obstacles in a dynamic environment.

[0167] Path length: The average length of the path that a robot takes to complete a task.

[0168] System stability: The consistency and reliability of the robot's planned paths across multiple tasks.

[0169] C33 Experimental Results

[0170] Task completion time: The task completion time of the technical solution provided by the present invention is significantly lower than that of the traditional A* algorithm, with an average reduction of about 30%, indicating that the path planning efficiency is higher.

[0171] Obstacle avoidance success rate: The obstacle avoidance success rate of the present invention reaches over 95%, which is significantly better than the obstacle avoidance performance of the traditional A* algorithm in dynamic scenes.

[0172] Path length: The path length generated by the present invention is shorter than that of the traditional A* algorithm, and the path is smoother, reducing unnecessary travel turns.

[0173] System stability: The task success rate of the present invention reached 100% in 50 experiments, while the traditional A* algorithm caused some tasks to fail due to the static nature of the path.

[0174] C34. Technical effects and application value

[0175] Through experimental verification, the present invention combines the global path planning capability of the A* algorithm with the local optimization and dynamic adaptability of the SAC algorithm, which not only significantly improves the efficiency of path planning, but also solves the problem of insufficient adaptability of traditional methods in dynamic environments. Its technical effect is better than that of using the A* algorithm alone, and it has a wide range of promotion and application value, especially suitable for dynamic warehousing, intelligent logistics, unmanned driving and other scenarios.

[0176] In summary, the present invention has the following advantages:

[0177] (1) Improving the real-time and adaptability of path planning: Combining the global planning capability of the A* algorithm and the local optimization capability of the SAC algorithm, this method solves the problems of slow response and limitations of traditional methods in dynamic environments, and meets the real-time and adaptability requirements.

[0178] (2) Improve the ability to plan in high-dimensional environments: Through the reinforcement learning mechanism of the SAC algorithm, it can effectively process high-dimensional state space, dynamically perceive multi-dimensional changes in complex environments, accurately plan paths, and improve the robot's navigation ability and adaptability in high-dimensional environments.

[0179] (3) Global path optimality planning capability: The A* algorithm is deeply integrated with the SAC algorithm to ensure the global rationality of the path while having dynamic adjustment capabilities, which makes up for the limitations of a single algorithm and makes path planning more intelligent and efficient.

[0180] (4) Reduce development and maintenance costs: Adopt modular design and standardized interfaces to achieve efficient interaction, reduce technical complexity and maintenance costs, facilitate system expansion and upgrade, and have broad application prospects and promotion value.

[0181] The above specific implementation methods are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to examples, a person skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the scope of the technical solution of the present invention, which should be included in the scope of the claims of the present invention.

Claims

1. An intelligent navigation planning method for a mobile robot, characterized in that: include: Collecting environmental information in real time, the environmental information includes static map information and real-time environmental information; Planning a global path using a global path planning algorithm according to the static map information; In combination with the real-time environmental information, a local path optimization algorithm is used to optimize and adjust the global path, and the global path is adjusted in real time to avoid dynamic obstacles to obtain an optimized path; A control instruction is generated according to the optimized path, and the robot is driven to move along the optimized path by the control instruction.

2. The intelligent navigation planning method according to claim 1, characterized in that: The global path optimization adjustment adopts the soft actor critic algorithm, and the processing process includes: Inputting the global path and the real-time environment information contained in the environment information; Define a state space S and an action space A, and determine information contained in the state space S and the action space A according to the environmental information; Based on the information contained in the state space S and the action space A, a dynamic reward function is determined; the dynamic reward function is used to update the relationship between the robot and the dynamic obstacle; wherein the dynamic reward function includes: rewarding the behavior close to the target point, punishing the behavior close to the dynamic obstacle and punishing the behavior deviating from the global path; updating the path planning strategy based on the above dynamic reward function combined with the soft actor critic algorithm; locally optimizing the global path based on the updated path planning strategy to obtain an optimized path.

3. The intelligent navigation planning method according to claim 2, characterized in that: The action space A includes: the robot's selectable movement direction and speed; the state space S includes: the robot's current position, speed and local environment characteristics.

4. To customers According to the intelligent navigation planning method according to claim 2, it is characterized in that: The dynamic reward function is determined according to the environmental information. The specific dynamic reward function is calculated based on the following formula: R(s,a)=α1·d target +α2·d obstacle +α3·d A* Among them, d target is the distance from the robot to the target point, d obsta2le is the distance from the robot to the nearest obstacle, d A* is the offset distance from the robot to the global path, α1, α2, and α3 are weight parameters; the weight parameters are set and adjusted according to the path planning requirements.

5. The intelligent navigation planning method according to claim 2, characterized in that: The updating of the path planning strategy includes strategy network optimization and value function optimization; Strategy network optimization: maximize the objective function, which is expressed as: J(π) = E st,at [R(s,a)]; Among them, E is the expected function, st is the current state of the environment, at is the action selected according to the strategy π in the state st, and R(s,a) is the dynamic reward function; Value function optimization: Minimize the Bellman residual and update the parameters of the policy network.

6. The intelligent navigation planning method according to claim 1, characterized in that: The real-time environmental information includes: road condition information, dynamic obstacle information and robot motion status; The static map information includes: the current position of the robot, the target position and the global environment map; the control instructions include: navigation instructions, avoidance priority instructions and energy-saving mode instructions.

7. The intelligent navigation planning method according to claim 1, characterized in that: The global path planning adopts the A-star algorithm.

8. The intelligent navigation planning method according to claim 1, characterized in that: The intelligent navigation planning method further includes a navigation experience optimization strategy, and the navigation experience optimization strategy includes: After completing each task, the robot stores the state, action and reward information of the path planning strategy process into the shared experience pool, and gradually optimizes the strategy network of the soft actor-critic algorithm.

9. An intelligent navigation planning system for a mobile robot, characterized in that: include: Environmental perception module: used to collect environmental information in real time, including static map information and real-time environmental information; Global path planning module: used to receive static map information sent by the environment perception module and generate a global path through a global path planning algorithm; Local path optimization module: used to receive real-time environmental information sent by the environmental perception module, use the local path optimization algorithm to optimize and adjust the global path, adjust the global path in real time to avoid dynamic obstacles, and obtain an optimized path; Robot control module: used to receive the optimized path sent by the local path optimization module, generate control instructions according to the optimized path, and drive the robot to move along the optimized path through the control instructions.

10. The intelligent navigation planning system according to claim 9, characterized in that: The environment perception module includes: a laser radar and a depth camera, which are used to collect static obstacle information, dynamic obstacle information and the robot's own status information around the robot.

Images