A hybrid path planning method and system capable of autonomous obstacle avoidance in a dynamic environment

Abstract

This invention discloses a hybrid path planning method and system for autonomous obstacle avoidance in dynamic environments. The method specifically includes: using state changes to coordinate global and local path planning; improving the artificial potential field method; proposing a dynamic obstacle avoidance mechanism with a collision penalty coefficient and obstacle avoidance compensation angle; and combining fuzzy control concepts with a state-change-based hybrid path planner. This invention solves the problem that traditional artificial potential field methods cannot effectively handle dynamic obstacles. By introducing fuzzy control concepts, adaptive changes in the obstacle avoidance compensation angle are achieved. The state-change-based hybrid path planning method coordinates the globally optimal path and local obstacle avoidance paths, achieving timely obstacle avoidance when needed and rapid return and movement along the optimal path when obstacle avoidance is not required. This results in path planning with lower time consumption, more stable control, less oscillation, and higher safety.

Description

Technical Field

[0001] This invention relates to a hybrid path planning method and system that enables autonomous obstacle avoidance in dynamic environments, belonging to the technical field of path planning methods. Background Technology

[0002] Mobile robots move towards a target point according to a planned path using a global path planning algorithm such as the A* algorithm. However, the A* algorithm can only plan paths when global map information is known, and it cannot perceive changes in the external environment, leading to collisions with obstacles and path planning failure. Therefore, a local path planning algorithm capable of sensing environmental changes is needed in conjunction with the global path planning algorithm to complete obstacle avoidance.

[0003] Among the mainstream local path planning algorithms for mobile robots, the dynamic window method has a complex model and insufficient foresight, resulting in a long planned path and generally poor real-time obstacle avoidance. The time elastic band method requires the distance difference, angle difference, and time difference between two states to calculate the velocity and angular velocity within the control cycle, resulting in high computational complexity and unstable control. The artificial potential field method has a simple principle, good real-time performance, and low computational complexity, and has been widely used. However, the traditional artificial potential field method cannot be applied to dynamic obstacle avoidance.

[0004] Currently, dynamic obstacle avoidance is typically achieved by fusing global and local path planning algorithms or by combining path planning algorithms with reinforcement learning. However, fusion algorithms of global and local path planning, such as the fusion of A* and dynamic window methods, still suffer from path redundancy and a large number of inflection points. While combining path planning with reinforcement learning offers strong environmental adaptability and good real-time obstacle avoidance, it requires extensive training starting from the simplest environments. Summary of the Invention

[0005] To address the problems existing in the prior art, this invention provides a hybrid path planning method and system that enables autonomous obstacle avoidance in dynamic environments. This method and system can achieve autonomous and rapid obstacle avoidance in dynamic environments and quickly return to the globally optimal path, thereby improving the intelligence and work efficiency of the robot.

[0006] To achieve the above objectives, the technical solution adopted by this invention is: a hybrid path planning method capable of autonomous obstacle avoidance in dynamic environments, comprising the following steps:

[0007] Step S1: Add the repulsive potential field generated by the obstacle velocity and acceleration on the robot to the traditional repulsive potential field function. At the same time, introduce the concept of collision penalty coefficient and divide the penalty coefficient into multiple risk levels, representing whether the mobile robot needs to avoid a certain dynamic obstacle, or how to select the dynamic obstacle that needs to be avoided when facing multiple dynamic obstacles. The improved repulsive potential field function is shown in Equation (1).

[0008]

[0009] In the formula, U rep (p, v, a) represents the repulsive potential field applicable to dynamic environments; U rep (v) represents the repulsive potential field generated by the velocity of the obstacle; U rep (a) is the repulsive potential field generated by the acceleration of the obstacle; K warn Let K be the collision penalty coefficient. warn A value greater than or equal to 1 indicates that the obstacle is close to the mobile robot, while a value less than 1 indicates that the obstacle is far away from the mobile robot.

[0010] Step S2: Set the relative distance d between the robot and the dynamic obstacle. ro Angle γ and relative velocity V rel and the set threshold distance d safe Threshold angle γ safe and threshold velocity V th The collision penalty coefficient K is obtained by comparison. warn As shown in equation (2);

[0011]

[0012] In the formula, τ is the threshold distance d safe Scale factor;

[0013] Step S3: The obstacle avoidance compensation angle fuzzy controller uses the relative distance d between the robot and the dynamic obstacle. ro Angle γ and relative velocity V rel The input is the magnitude ε of the dynamic compensation angle θ; after defuzzifying the output θ of the obstacle avoidance compensation angle fuzzy controller, the corresponding compensation angle is obtained, as shown in equation (3);

[0014]

[0015] In the formula, the coefficients This represents the proportional relationship between the compensation angle and the included angle, where a is less than b;

[0016] Step S4: Perform obstacle avoidance angle compensation for dynamic obstacles with a collision penalty coefficient greater than or equal to 1, and calculate the new heading angle after compensation. The calculation method is shown in equation (4);

[0017]

[0018] In the formula, θ is the obstacle avoidance compensation angle, which is given by the obstacle avoidance compensation angle fuzzy controller in step S3; m is the positive and negative coefficient of the angle γ between the robot and the obstacle;

[0019] Step S5: After the above steps realize the local dynamic obstacle avoidance planning of the mobile robot, in order to minimize the path length caused by local obstacle avoidance, the global path planning and local dynamic obstacle avoidance planning are completed by state change to achieve hybrid path planning, under the premise of ensuring successful obstacle avoidance.

[0020] Step S6: After the robot completes localization, the path planning module begins to plan global and local paths. Then, it decides whether to change the state based on the external environment and feeds back the planning results. Based on the feedback information, control commands are formulated and transmitted to the lower layer. After the lower layer performs inverse kinematics analysis, the control signals are input to the motor drive to achieve control.

[0021] Furthermore, in step S2, if the relative distance or angle between the robot and the dynamic obstacle is greater than or equal to a set threshold, or the relative speed is less than or equal to 0, a coefficient of -1 is set to provide a negative reward, indicating that obstacle avoidance is not required; when the relative distance or angle between the robot and the obstacle is less than the set threshold and the relative speed is greater than 0, a coefficient of 1 is set to provide a positive reward, indicating that there is a risk of collision between the robot and the dynamic obstacle, and obstacle avoidance is required; at this time, if the relative speed between the two is greater than the set threshold, a coefficient of 2 is set to prioritize obstacle avoidance of that obstacle; when the relative distance is less than the threshold... When the collision penalty coefficient is set to 3, the robot stops moving. After obtaining the collision penalty coefficient of the dynamic obstacles around the robot, the obstacle with the largest coefficient is selected and the obstacle avoidance angle compensation is continuously performed until the obstacle avoidance is completed. When multiple obstacles are at the same penalty level, the dynamic obstacle closest to the robot is compensated first. During the obstacle avoidance process, the penalty coefficient of the obstacle will be updated to adjust the obstacle avoidance order.

[0022] Furthermore, in step S4, the direction of the resultant force of the artificial potential field method is used as the reference, with clockwise as the positive direction and counterclockwise as the negative direction. When the robot forms a negative angle with the obstacle, m is set to 0. At this time, taking the absolute value of the negative angle, the angle will be adjusted to the clockwise direction. After adding the obstacle avoidance compensation angle, the robot avoids obstacles along the heading angle ||γ|+θ|. When the robot forms a positive angle with the obstacle, m is set to 1. At this time, taking the absolute value of the positive angle, the angle is still in the clockwise direction. After adding the obstacle avoidance compensation angle, ||γ|+θ| is reversed, and the heading angle is adjusted to the counterclockwise direction. The robot avoids obstacles along the reversed heading angle.

[0023] Furthermore, in step S6, the state change consists of an initial state, a global planning state, a judgment state, and a local obstacle avoidance state, and the specific change method is as follows:

[0024] Initial state: Perform one location, and after the matching is completed, enter the global planning state; otherwise, maintain the initial state.

[0025] Global planning state: Perform a global optimal path and local obstacle avoidance path planning. In this state, the global path has higher priority than the local path, and the robot moves along the global path. After the robot moves a certain distance, it enters the judgment state; otherwise, it remains in the global planning state.

[0026] Determine the state: If the local planner perceives that dynamic obstacles are affecting the global path or that the path safety is low, it enters the local obstacle avoidance state; otherwise, it returns to the global planning state.

[0027] Local obstacle avoidance state: The priority is adjusted to local path over global path, and global planning is re-performed starting from the current position; a point in the forward direction of the global path is selected as a temporary target point, and the robot moves towards the point along the local obstacle avoidance path; if obstacle avoidance is completed, return to the global planning state; otherwise, maintain the local obstacle avoidance state until the temporary target point is reached; if there is a sudden obstacle near the selected temporary target point, planning is re-performed and a point is selected as a new temporary target point; if obstacle avoidance path planning fails, return to the initial state.

[0028] Furthermore, in step S6, the path planning module includes a global path planning unit, a local path planning unit, and a dynamic obstacle avoidance judgment unit. The global path planning unit and the local path planning unit plan the globally optimal path and the local obstacle avoidance path, respectively. The dynamic obstacle avoidance judgment unit is used to detect whether the robot has identified a dynamic obstacle. If no obstacle is detected, the global path has a higher priority than the local path, and the robot moves along the global path. The planning result is sent to the control command generation unit, which generates a speed command and transmits it to the lower-level STM32 controller via a serial port. If a dynamic obstacle is detected, the collision penalty coefficient is calculated in real time by using the distance, angle, and relative speed between the robot and the obstacle. If the penalty level is low, obstacle avoidance is not required, and the global path has a higher priority than the local path, and a control command is generated. If the penalty level is high, the distance, angle, and relative speed between the robot and the obstacle are fuzzified, and the obstacle avoidance compensation angle is output by the obstacle avoidance compensation angle fuzzy control unit. After receiving the obstacle avoidance compensation angle, the path priority correction unit adjusts the path priority to prioritize the local path over the global path and generates a control command, which is transmitted to the lower-level controller to complete the dynamic obstacle avoidance.

[0029] A hybrid path planning system capable of autonomous obstacle avoidance in dynamic environments includes a lidar, a Raspberry Pi, an STM32 controller, and a motor. The lidar is used to collect environmental information and communicates with the Raspberry Pi via a USB port to transmit lidar data to a map building unit. The STM32 controller is used to control the actuators and transmits information with the Raspberry Pi via a serial port to obtain control commands from a host computer. The Raspberry Pi is used for global path planning and local path planning.

[0030] Furthermore, the lidar is used to transmit the current laser data to the Raspberry Pi, obtain map information from the positioning and mapping service unit, and complete positioning through initial pose data and laser data; after receiving the target point information, the global path planning unit and the local path planning unit plan the global optimal path and the local obstacle avoidance path, respectively.

[0031] The beneficial effects of this invention are as follows: By improving the repulsive function of the traditional artificial potential field method and proposing a dynamic obstacle avoidance mechanism based on the collision penalty coefficient and obstacle avoidance compensation angle, the problem that the traditional artificial potential field method cannot effectively cope with dynamic obstacles is solved. By introducing fuzzy control concepts (consistent with the characteristic of the precise model that does not require a process in this invention), adaptive changes in the obstacle avoidance compensation angle are achieved. By coordinating the globally optimal path and the local obstacle avoidance path through a hybrid path planning method based on state changes, the invention achieves the effect of timely obstacle avoidance when obstacle avoidance is needed and rapid return and movement along the optimal path when obstacle avoidance is not needed, completing path planning with less time consumption, more stable control, less oscillation, and higher safety. Attached Figure Description

[0032] Figure 1 This is a schematic diagram of the obstacle avoidance compensation angle according to an embodiment of the present invention;

[0033] Figure 2 This is a schematic diagram of the planner structure according to an embodiment of the present invention;

[0034] Figure 3 This is a schematic diagram illustrating four state change processes according to an embodiment of the present invention;

[0035] Figure 4 This is a schematic diagram of a hybrid path planning system according to an embodiment of the present invention;

[0036] Figure 5 This is the implementation environment for the embodiments of the present invention;

[0037] Figure 6 This is a dynamic obstacle avoidance verification under experimental conditions in an embodiment of the present invention; where (a) is the global planning state, (b) is the obstacle avoidance angle compensation, and (c) is the return to the global planning state after completing dynamic obstacle avoidance. Detailed Implementation

[0038] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. However, it should be understood that the specific embodiments described herein are merely illustrative and are not intended to limit the scope of the invention.

[0039] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of this invention is for the purpose of describing particular embodiments only and is not intended to limit the invention.

[0040] This invention provides a hybrid path planning method for autonomous obstacle avoidance of mobile robots in dynamic environments; the specific steps are as follows:

[0041] (1) The repulsive potential field generated by the obstacle velocity and acceleration on the robot is added to the traditional repulsive potential field function. At the same time, the concept of collision penalty coefficient is introduced and the penalty coefficient is divided into multiple risk levels, representing whether the mobile robot needs to avoid a certain dynamic obstacle, or how to select the dynamic obstacle that needs to be avoided when facing multiple dynamic obstacles. The improved repulsive potential field function is shown in Equation (1).

[0042]

[0043] In the formula, U rep (p, v, a) represents the repulsive potential field applicable to dynamic environments; U rep (v) represents the repulsive potential field generated by the velocity of the obstacle; U rep (a) is the repulsive potential field generated by the acceleration of the obstacle; K warn Let K be the collision penalty coefficient. warn A value greater than or equal to 1 indicates that the obstacle is close to the mobile robot, while a value less than 1 indicates that the obstacle is far away from the mobile robot.

[0044] (2) The relative distance d between the robot and the dynamic obstacle ro Angle γ and relative velocity V rel and the set threshold distance d safe Threshold angle γ safe and threshold velocity V th The collision penalty coefficient K is obtained by comparison. warn As shown in equation (2).

[0045]

[0046] In the formula, τ is the threshold distance d safeThe scaling factor is set as follows: If the relative distance or angle between the robot and a dynamic obstacle is greater than or equal to a set threshold, or the relative speed is less than or equal to 0, a coefficient of -1 is set to provide a negative reward, indicating that obstacle avoidance is not required; when the relative distance or angle between the robot and an obstacle is less than the set threshold and the relative speed is greater than 0, a coefficient of 1 is set to provide a positive reward, indicating that there is a risk of collision between the robot and the dynamic obstacle, and obstacle avoidance is required; at this point, if the relative speed between the two is greater than the set threshold, a coefficient of 2 is set to prioritize obstacle avoidance; when the relative distance is less than one-fifth of the threshold, a coefficient of 3 is set to indicate that the robot stops moving. After obtaining the collision penalty coefficients of the dynamic obstacles around the robot, the obstacle with the largest coefficient is selected for continuous obstacle avoidance angle compensation until obstacle avoidance is completed; if multiple obstacles are at the same penalty level, the dynamic obstacle closest to the robot is compensated first; during the obstacle avoidance process, the penalty coefficients of the obstacles are updated to adjust the obstacle avoidance order.

[0047] (3) The obstacle avoidance compensation angle fuzzy controller uses the relative distance d between the robot and the dynamic obstacle. ro Angle γ and relative velocity V rel With the input being the magnitude ε of the dynamic compensation angle θ, this embodiment of the invention establishes 125 fuzzy rules in a 5×5×5 format, as shown in Table 1.

[0048] Table 1. Semantic Comparison Table of Obstacle Avoidance Compensation Angle

[0049] After defuzzifying the output θ of the obstacle avoidance compensation angle fuzzy controller, the corresponding compensation angle is obtained, as shown in equation (3).

[0050]

[0051] In the formula, the coefficients This represents the proportional relationship between the compensation angle and the included angle, where a is greater than b;

[0052] (4) Perform obstacle avoidance angle compensation for dynamic obstacles with a collision penalty coefficient greater than or equal to 1, and the new heading angle after compensation. The calculation method is shown in equation (4).

[0053]

[0054] In the formula, θ is the obstacle avoidance compensation angle, given by the obstacle avoidance compensation angle fuzzy controller in step (3); m is the positive and negative coefficient of the angle γ between the robot and the obstacle. In this invention, the direction of the resultant force of the artificial potential field method is defined as the reference, with the clockwise direction as the positive direction and the counterclockwise direction as the negative direction. When the robot and the obstacle form a negative angle, m is set to 0. At this time, the absolute value of the negative angle is taken, and the angle will be adjusted to the clockwise direction. After adding the obstacle avoidance compensation angle, the robot avoids obstacles along the heading angle ||γ|+θ|. When the robot and the obstacle form a positive angle, m is set to 1. At this time, the absolute value of the positive angle is taken, and the angle is still in the clockwise direction. After adding the obstacle avoidance compensation angle, the ||γ|+θ| is reversed, and the heading angle is adjusted to the counterclockwise direction. The robot avoids obstacles along the reversed heading angle. The obstacle avoidance angle compensation process is as follows: Figure 1 As shown.

[0055] (4) The previous four steps achieved local dynamic obstacle avoidance planning for the mobile robot. To minimize the path length caused by local obstacle avoidance while ensuring successful obstacle avoidance, this invention integrates global path planning and local path planning through state changes to complete the design of a hybrid path planner. The planner structure is as follows: Figure 2 As shown.

[0056] After the robot completes localization, the path planning module begins planning global and local paths. It then decides whether to change the state based on the external environment and feeds back the planning results. Control commands are generated based on the feedback and transmitted to the lower layer. The lower layer performs inverse kinematics analysis and inputs the control signals to the motor drive to achieve control. State change is one of the core functions of the planner, consisting of an initial state, a global planning state, a decision state, and a local obstacle avoidance state. The change process is as follows: Figure 3 As shown.

[0057] Initial state: Perform one location and match. If the match is successful, enter the global planning state. Otherwise, maintain the initial state.

[0058] Global planning state: Performs a global optimal path and local obstacle avoidance path planning. In this state, the global path has higher priority than the local path, and the robot moves along the global path. After moving a certain distance, the robot enters the decision state; otherwise, it remains in the global planning state.

[0059] Determine the state: If the local planner perceives that dynamic obstacles affect the global path or the path safety is low, it enters the local obstacle avoidance state; otherwise, it returns to the global planning state.

[0060] Local obstacle avoidance state: Prioritize the local path over the global path, and re-plan the global path from the current position; select a point in the forward direction of the global path as a temporary target point, and move the robot towards that point along the local obstacle avoidance path; if obstacle avoidance is successful, return to the global planning state; otherwise, maintain the local obstacle avoidance state until the temporary target point is reached; if there is a sudden obstacle near the selected temporary target point, re-plan and select a new temporary target point. When obstacle avoidance path planning fails, return to the initial state.

[0061] This invention also provides a hybrid path planning system suitable for mobile robots, such as... Figure 4 As shown, the system consists of a lidar, a Raspberry Pi, an STM32 controller, motors, etc.

[0062] The LiDAR is used to collect environmental information and communicates with the Raspberry Pi via USB to transmit LiDAR data to the map building unit.

[0063] The STM32 controller is used to control the actuators, transmit information with the Raspberry Pi via serial port, and obtain control commands from the host computer.

[0064] Raspberry Pi is used for global path planning and local path planning. After the LiDAR transmits the current laser data to the Raspberry Pi, it obtains map information from the positioning and mapping service unit and completes positioning using initial pose data and laser data. Upon receiving target point information, the global path planning unit and the local path planning unit plan the globally optimal path and the local obstacle avoidance path, respectively. The dynamic obstacle avoidance judgment unit detects whether the robot has identified a dynamic obstacle. If no obstacle is detected, the global path has higher priority than the local path, and the robot moves along the global path. The planning result is sent to the control command generation unit, which generates a speed command and transmits it to the lower-level STM32 controller via serial port. If a dynamic obstacle is detected, the collision penalty coefficient real-time calculation unit calculates the collision penalty coefficient based on the distance, angle, and relative speed between the robot and the obstacle. If the penalty level is low, obstacle avoidance is not required, the global path has higher priority than the local path, and control commands are generated. If the penalty level is high, the distance, angle, and relative speed between the robot and the obstacle are fuzzified, and the obstacle avoidance compensation angle is output by the obstacle avoidance compensation angle fuzzy control unit. After receiving the obstacle avoidance compensation angle, the path priority correction unit adjusts the path priority to local path higher than global path and generates control commands, which are transmitted to the lower-level controller to complete dynamic obstacle avoidance.

[0065] In this embodiment of the invention, a mobile chassis is constructed using Raspberry Pi, STM32, LiDAR, omnidirectional wheels, etc. Figure 5 The method described in this paper is implemented on the omnidirectional mobile platform shown.

[0066] This invention utilizes RViz to observe the robot's motion state. After receiving target point information, the robot performs global and local path planning. At this point, no dynamic obstacles are detected, and the global path has a higher priority than the local path. The robot moves along the global path, such as... Figure 6 As shown in Figure a; after the robot detects a dynamic obstacle, it calculates the collision penalty coefficient for the obstacle. Obstacle 1 is closer to the robot and has a smaller angle, resulting in a larger penalty coefficient. Obstacle 2, on the other hand, has a larger angle with the robot and a lower penalty coefficient. Therefore, obstacle 1 is compensated for with an obstacle avoidance angle, as shown in Figure a. Figure 6 As shown in b; after completing dynamic obstacle avoidance, the robot quickly returns to the global path planning state and continues to move along the globally optimal path, as shown in b. Figure 6 As shown in c. Experimental results show that after obstacle avoidance angle compensation, the robot can avoid dynamic obstacles with high safety and a smooth obstacle avoidance path; after completing dynamic obstacle avoidance, continuing to move along the globally optimal path greatly shortens navigation time and path length.

[0067] Comparative Implementation Examples: The local planner is applied using the virtual force method, dynamic window method, and time elastic band method in dynamic environments with varying obstacle speeds.

[0068] The performance of the proposed method was compared with that of the method in this paper. Under the same starting and target points, the navigation time is shown in Table 2. The results show that the proposed method achieves autonomous obstacle avoidance in dynamic environments and reaches the target point in the shortest possible time. In environments with slow-moving dynamic obstacles, it improves performance by 19% compared to the dynamic window method and 13% compared to the time-elastic band method; in environments with fast-moving dynamic obstacles, it improves performance by 12% compared to the dynamic window method and 9% compared to the time-elastic band method.

[0069] Table 2 Comparison of Navigation Time

[0070]

[0071] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions or improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A hybrid path planning method capable of autonomous obstacle avoidance in dynamic environments, characterized in that, Includes the following steps: Step S1: Add the repulsive potential field generated by the obstacle velocity and acceleration on the robot to the traditional repulsive potential field function. At the same time, introduce the concept of collision penalty coefficient and divide the penalty coefficient into multiple risk levels, representing whether the mobile robot needs to avoid a certain dynamic obstacle, or how to select the dynamic obstacle that needs to be avoided when facing multiple dynamic obstacles. The improved repulsive potential field function is shown in Equation (1). In the formula, U rep (p, v, a) represents the repulsive potential field applicable to dynamic environments; U rep (v) represents the repulsive potential field generated by the velocity of the obstacle; U rep (a) is the repulsive potential field generated by the acceleration of the obstacle; K warn Let K be the collision penalty coefficient. warn A value greater than or equal to 1 indicates that the obstacle is close to the mobile robot, while a value less than 1 indicates that the obstacle is far away from the mobile robot. Step S2: Set the relative distance d between the robot and the dynamic obstacle. ro Angle γ and relative velocity V rel and the set threshold distance d safe Threshold angle γ safe and threshold velocity V th The collision penalty coefficient K is obtained by comparison. warn As shown in equation (2); In the formula, τ is the threshold distance d safe Scale factor; Step S3: The obstacle avoidance compensation angle fuzzy controller uses the relative distance d between the robot and the dynamic obstacle. ro Angle γ and relative velocity V rel The input is the magnitude ε of the dynamic compensation angle θ; after defuzzifying the output θ of the obstacle avoidance compensation angle fuzzy controller, the corresponding compensation angle is obtained, as shown in equation (3); In the formula, the coefficients This represents the proportional relationship between the compensation angle and the included angle, where a is less than b; Where PB, MD, PS, and ZE are fuzzy variables represented by fuzzy functions, corresponding to semantic large, medium, small, and zero, respectively; Step S4: Perform obstacle avoidance angle compensation for dynamic obstacles with a collision penalty coefficient greater than or equal to 1, and calculate the new heading angle after compensation. The calculation method is shown in equation (4); In the formula, θ is the obstacle avoidance compensation angle, which is given by the obstacle avoidance compensation angle fuzzy controller in step S3; m is the positive and negative coefficient of the angle γ between the robot and the obstacle; Step S5: After the above steps realize the local dynamic obstacle avoidance planning of the mobile robot, in order to minimize the path length caused by local obstacle avoidance, the global path planning and local dynamic obstacle avoidance planning are completed by state change to achieve hybrid path planning, under the premise of ensuring successful obstacle avoidance. Step S6: After the robot completes localization, the path planning module begins to plan global and local paths. Then, it decides whether to change the state based on the external environment and feeds back the planning results. Based on the feedback information, control commands are formulated and transmitted to the lower layer. After the lower layer performs inverse kinematics analysis, the control signals are input to the motor drive to achieve control.

2. The hybrid path planning method for autonomous obstacle avoidance in dynamic environments according to claim 1, characterized in that, In step S2, if the relative distance or angle between the robot and the dynamic obstacle is greater than or equal to a set threshold or the relative speed is less than or equal to 0, a coefficient of -1 is set to give a negative reward, indicating that obstacle avoidance is not required; when the relative distance or angle between the robot and the obstacle is less than the set threshold and the relative speed is greater than 0, a coefficient of 1 is set to give a positive reward, indicating that there is a risk of collision between the robot and the dynamic obstacle and obstacle avoidance is required. At this moment, when the relative speed between the two objects exceeds a set threshold, a coefficient of 2 is set to prioritize obstacle avoidance; when the relative distance is less than the threshold... When the collision penalty coefficient is set to 3, the robot stops moving. After obtaining the collision penalty coefficient of the dynamic obstacles around the robot, the obstacle with the largest coefficient is selected and the obstacle avoidance angle compensation is continuously performed until the obstacle avoidance is completed. When multiple obstacles are at the same penalty level, the dynamic obstacle closest to the robot is compensated first. During the obstacle avoidance process, the penalty coefficient of the obstacle will be updated to adjust the obstacle avoidance order.

3. The hybrid path planning method for autonomous obstacle avoidance in dynamic environments according to claim 1, characterized in that, In step S4, the direction of the resultant force of the artificial potential field method is used as the reference, with clockwise as the positive direction and counterclockwise as the negative direction. When the robot forms a negative angle with the obstacle, m is set to 0. At this time, taking the absolute value of the negative angle, the angle will be adjusted to the clockwise direction. After adding the obstacle avoidance compensation angle, the robot avoids obstacles along the heading angle ||γ|+θ|. When the robot forms a positive angle with the obstacle, m is set to 1. At this time, taking the absolute value of the positive angle, the angle is still in the clockwise direction. After adding the obstacle avoidance compensation angle, ||γ|+θ| is reversed, and the heading angle is adjusted to the counterclockwise direction. The robot avoids obstacles along the reversed heading angle.

4. The hybrid path planning method for autonomous obstacle avoidance in dynamic environments according to claim 1, characterized in that, In step S6, the state change consists of the initial state, the global planning state, the judgment state, and the local obstacle avoidance state. The specific change method is as follows: Initial state: Perform one location, and after the matching is completed, enter the global planning state; otherwise, maintain the initial state. Global planning state: Perform a global optimal path and local obstacle avoidance path planning. In this state, the global path has higher priority than the local path, and the robot moves along the global path. After the robot moves a certain distance, it enters the judgment state; otherwise, it remains in the global planning state. Determine the state: If the local planner perceives that dynamic obstacles are affecting the global path or that the path safety is low, it enters the local obstacle avoidance state; otherwise, it returns to the global planning state. Local obstacle avoidance state: Adjust the priority to local path over global path, and re-perform global planning starting from the current position; Select a point in the global path forward direction as a temporary target point, and move the robot towards that point along the local obstacle avoidance path; if obstacle avoidance is completed, return to the global planning state; otherwise, maintain the local obstacle avoidance state until the temporary target point is reached; if there is a sudden obstacle near the selected temporary target point, re-plan and select a point as a new temporary target point; if obstacle avoidance path planning fails, return to the initial state.

5. The hybrid path planning method for autonomous obstacle avoidance in a dynamic environment according to claim 1, characterized in that, In step S6, the path planning module includes a global path planning unit, a local path planning unit, and a dynamic obstacle avoidance judgment unit. The global path planning unit and the local path planning unit plan the global optimal path and the local obstacle avoidance path, respectively. The dynamic obstacle avoidance judgment unit is used to detect whether the robot has identified a dynamic obstacle. If no obstacle is detected, the global path has a higher priority than the local path, and the robot moves along the global path. The planning result is sent to the control command generation unit, which generates a speed command and transmits it to the lower-level STM32 controller via a serial port. If a dynamic obstacle is detected, the collision penalty coefficient is calculated in real time by using the distance, angle, and relative speed between the robot and the obstacle. If the penalty level is low, obstacle avoidance is not required, and the global path has a higher priority than the local path, and a control command is generated. If the penalty level is high, the distance, angle, and relative speed between the robot and the obstacle are fuzzified, and the obstacle avoidance compensation angle is output by the obstacle avoidance compensation angle fuzzy control unit. After receiving the obstacle avoidance compensation angle, the path priority correction unit adjusts the path priority to prioritize the local path over the global path and generates a control command, which is transmitted to the lower-level controller to complete the dynamic obstacle avoidance.

6. A hybrid path planning system capable of autonomous obstacle avoidance in dynamic environments, comprising employing the hybrid path planning method for autonomous obstacle avoidance in dynamic environments as described in claim 1, characterized in that, The system includes a LiDAR, a Raspberry Pi, an STM32 controller, and a motor. The LiDAR is used to collect environmental information and communicates with the Raspberry Pi via a USB port to transmit LiDAR data to the map building unit. The STM32 controller is used to control the actuators and transmits information with the Raspberry Pi via a serial port to obtain control commands from the host computer. The Raspberry Pi is used for global path planning and local path planning.

7. A hybrid path planning system capable of autonomous obstacle avoidance in dynamic environments according to claim 6, characterized in that, The lidar is used to transmit the current laser data to the Raspberry Pi, obtain map information from the positioning and mapping service unit, and complete positioning through initial pose data and laser data; after receiving the target point information, the global path planning unit and the local path planning unit plan the global optimal path and the local obstacle avoidance path, respectively.

Images