Improved giant mouse algorithm-based path planning method and system for inspection robot

Abstract

The application discloses a kind of factory inspection robot path planning methods based on improved big cane mouse algorithm and dynamic window method fusion, it is related to the field of path planning, this method first constructs factory workshop three-dimensional environment model;Second, the triple improvement is carried out to basic big cane mouse algorithm:design adaptive weight linearly decreasing with iteration number, dynamically balance global exploration and local development capability;Introduce sine-cosine strategy, through specific parameter control and sine-cosine direction optimization, search dimension is enriched to avoid local optimum;Adaptive factor is embedded to core calculation formula, introduce nonlinear function, improve global path planning precision and convergence speed, simultaneously, fusion dynamic window method constructs "global-local" double-layer path planning framework.Experiments show that, in complex dynamic environment of factory, global path planning error is reduced by 15%-20%, dynamic obstacle avoidance response time is shortened to within 0.3s, can satisfy robot all-weather inspection demand, guarantee inspection efficiency and operation safety.

Description

Technical Field

[0001] This invention relates to the field of path planning, and in particular to a path planning method and system for inspection robots based on an improved cane rat algorithm. Background Technology

[0002] With the development of intelligent manufacturing technology, factory inspection modes are gradually shifting from manual inspection to autonomous robot inspection. Path planning, as the core technology of autonomous robot inspection, directly determines inspection efficiency and operational safety. Currently, factory inspection robot path planning methods are mainly divided into two categories: global path planning and local path planning, but significant technical shortcomings still exist in practical applications.

[0003] In terms of global path planning, traditional algorithms such as A While algorithms like Dijkstra's algorithm can generate feasible paths in static environments, they are poorly adapted to the complex spaces of factories with multiple devices and channels, and struggle to balance path optimality and planning efficiency. Metaheuristic algorithms such as particle swarm optimization and genetic algorithms, while improving global search capabilities, are prone to slow convergence and getting stuck in local optima. The basic Big Cane Mouse algorithm, employing fixed weights and linear position update logic, is prone to missing optimal paths in the early stages of factory scenarios with multiple obstacles due to insufficient exploration range, and its fixed step size in the later stages leads to low solution accuracy, failing to meet the requirements for high-precision inspection paths.

[0004] In terms of local path planning, dynamic window methods and vector field histogram methods are commonly used for dynamic obstacle avoidance. However, single local algorithms suffer from a "short-sighted" problem: they can only adjust the movement state based on the current obstacle information, lacking global path guidance, and are prone to detours and failure to reach the target point. In factory workshops, the trajectories of dynamic obstacles such as pedestrians and mobile devices are random. Although the single dynamic window method can avoid obstacles in real time, it may deviate from the globally optimal route due to local adjustments, resulting in longer inspection paths and increased task time.

[0005] Furthermore, existing technologies suffer from low integration of global and local path planning. Most solutions simply superimpose two types of algorithms without establishing an effective data interaction and collaboration mechanism. The global path cannot guide local obstacle avoidance decisions in real time, and local obstacle avoidance adjustments are not fed back to the global path for dynamic correction. This makes it easy for robots to encounter problems such as "global path is feasible but local obstacle avoidance fails" or "local obstacle avoidance is successful but deviates from the inspection target" during all-weather inspections. Especially during peak hours in the workshop, when pedestrians and equipment flow is dense, the risk of robot collisions increases significantly, making it difficult to ensure stable operation around the clock. Summary of the Invention

[0006] Based on the above problems, there is an urgent need for a fusion algorithm that combines high-precision global path planning capabilities with rapid dynamic obstacle avoidance capabilities, adapts to the complex and dynamic environment of factories, meets the inspection needs of robots in all weather conditions, with high safety and high efficiency, and promotes the improvement of the automation level of inspection in smart manufacturing factories.

[0007] The first aspect of this invention provides a path planning method for an inspection robot based on an improved Big Cane Mouse algorithm. This method includes: acquiring data within a workshop; determining an optimal inspection path based on the improved Big Cane Mouse algorithm, wherein the search range of the improved Big Cane Mouse algorithm decreases with the number of iterations; determining a velocity window and an angular velocity window based on robot motion constraints and safety braking constraints; sampling multiple candidate velocities and multiple candidate angular velocities within the velocity and angular velocity windows; generating multiple predicted paths based on the candidate velocities and candidate angular velocities, the predicted paths being used to move the robot to a local target point; scoring each predicted path based on approachability, safety, and speed smoothness; and determining the predicted path with the highest score as the motion path, wherein approachability indicates the directional consistency between the target point of the predicted path and the local target point of the optimal inspection path; safety indicates the ability to ensure a safe distance between the trajectory and obstacles; and speed smoothness indicates the ability to maintain motion continuity; moving the robot along the motion path to the local target point; recalculating the velocity and angular velocity windows and re-determining the motion path; and continuing until the robot moves to the global target point.

[0008] In some implementations, moving the robot along the motion path to a local target point, recalculating the velocity window and angular velocity window, and redetermining the motion path until the robot reaches the global target point includes:

[0009] If the distance between the local target point and the optimal inspection path exceeds a distance threshold, the improved Big Cane Mouse algorithm is invoked to recalculate the optimal inspection path, with the local target point as the starting point and the global target point as the ending point.

[0010] In some implementations, the step of moving the robot along the motion path to a local target point, recalculating the velocity window and angular velocity window, and redetermining the motion path until the robot moves to the global target point includes: acquiring the current position, speed, and direction of movement of dynamic obstacles in the workshop in real time using lidar and vision sensors, recalculating the minimum distance between the predicted trajectory and the obstacle, and recalculating the motion path.

[0011] In some implementations, the formula for adjusting population position based on leader guidance in the improved cane rat algorithm is as follows:

[0012]

[0013] in, Let be the position of the i-th individual in the (t+1)-th search. Let be the position of the i-th individual in the t-th search. Influence coefficient of optimal solution; For adaptive weight values, For the first Search step size in the next iteration For the first The iteration coefficient at the next iteration;

[0014] Adaptive weight values The calculation formula is:

[0015]

[0016] in, Based on the weighting coefficient, The maximum number of iterations is preset for the algorithm, where t is the number of searches;

[0017] Search Step Size The calculation formula is:

[0018]

[0019] in, The step size scaling constant. For the first During the nth iteration The fitness value of each search individual. For the first The optimal fitness value of the population at the next iteration. It is a nonlinear function;

[0020] Iteration coefficients The calculation formula is:

[0021]

[0022] in, This is the iterative scaling constant. For the first The location of the global optimal solution for the population in the next iteration. The variable adjustment coefficient, For the first The random perturbation value during the next iteration, wherein the perturbation value takes a value between 0 and 1.

[0023] In some implementations, the improved large cane rat algorithm uses the following formulas to adjust the population position based on leader guidance and to adjust the search direction based on sine and cosine functions:

[0024]

[0025] or

[0026]

[0027] In the formula, the sine or cosine function is randomly selected before each calculation. Let be the position of the i-th individual in the (t+1)-th search. Let be the position of the i-th individual in the t-th search. For adaptive weight values, For the first Search step size in the next iteration For the first The iteration coefficients for the next iteration; r1 is the decreasing adjustment coefficient; r2 and r3 are two random angle parameters;

[0028] The formula for calculating the decreasing adjustment coefficient r1 is:

[0029]

[0030] The two random angle parameters are as follows:

[0031]

[0032]

[0033] They are respectively Different random numbers uniformly distributed within an interval.

[0034] In some implementations, the formula for updating the population position based on the new optimal position using peer guidance in the improved cane rat algorithm is as follows:

[0035]

[0036] In the formula, Let be the position of the i-th individual in the (t+1)-th search. Let be the position of the i-th individual in the t-th search. For the search scaling constant, The variable adjustment coefficient, For adaptive weight values, For the first Search step size in the next iteration For the first The average position of all searched individuals at the next iteration.

[0037] In some implementations, the formula for generating the candidate next generation population in the improved cane rat algorithm is:

[0038]

[0039] in, This represents the position of the i-th individual in the next generation population. This represents the position of the i-th individual in the previous generation population. Cross factor For the first The iteration coefficient at the next iteration;

[0040] The formula for selecting the next generation of the population is:

[0041]

[0042] in, for fitness coefficient, for The fitness coefficient.

[0043] In some implementations, scoring each of the predicted paths based on proximity, safety, and speed smoothness includes:

[0044] The score for each predicted path is calculated using a path evaluation formula, which is:

[0045]

[0046] in, A score for the predicted path. , , Here, G represents the target approach rate, O represents the safety factor, and V represents the velocity smoothness factor.

[0047] The formula for calculating the goal-approachability G is:

[0048]

[0049] in, The angle between the robot's forward direction and the "vector from the current position to the local target point";

[0050] The formula for calculating safety O is:

[0051]

[0052] in, To establish a safe distance, This represents the maximum possible distance within the window.

[0053] The formula for calculating velocity smoothness V is:

[0054]

[0055] in, The desired linear velocity guided by the global path. Let be the actual speed of the i-th predicted path.

[0056] A second aspect of this invention provides a path planning system for an inspection robot based on an improved Big Cane Rat algorithm. The system includes: a path planning module for acquiring data within a workshop and determining the optimal inspection path based on the improved Big Cane Rat algorithm; a window selection module for determining a velocity window and an angular velocity window based on robot motion constraints and safety braking constraints, and sampling multiple candidate velocities and multiple candidate angular velocities within the velocity and angular velocity windows; a path generation module for generating multiple predicted paths based on the candidate velocities and candidate angular velocities, scoring each predicted path based on proximity, safety, and velocity smoothness, and determining the predicted path with the highest score as the motion path; and further for enabling the robot to move along the motion path to a local target point, recalculating the velocity and angular velocity windows, and re-determining the motion path until the robot moves to a global target point.

[0057] In some implementations, the path generation module is further configured to, if the distance between the local target point and the optimal inspection path exceeds a distance threshold, use the local target point as the starting point and the global target point as the ending point to call the improved Big Cane Mouse algorithm to recalculate the optimal inspection path.

[0058] This invention discloses a path planning method for factory inspection robots based on the fusion of an improved Big Cane Mouse algorithm and a dynamic window method, aiming to solve the technical problems of low global path planning accuracy, slow local dynamic obstacle avoidance response, and difficulty in adapting to the complex dynamic environment of the workshop in traditional factory inspection robots. The method first constructs a three-dimensional environment model of the factory workshop, clarifying the spatial distribution and motion characteristics of inspection target points, static obstacles, and dynamic obstacles. Second, it makes three improvements to the basic Big Cane Mouse algorithm: designing adaptive weights that decrease linearly with the number of iterations to dynamically balance global exploration and local development capabilities; introducing a sine and cosine strategy, enriching the search dimensions through specific parameter adjustment and sine and cosine direction optimization to avoid local optima; and embedding adaptive factors and introducing nonlinear functions into the core calculation formula to improve the accuracy and convergence speed of global path planning. Based on this, a "global-local" two-layer path planning architecture is constructed by fusing the dynamic window method: the global layer generates the optimal collision-free path from the starting point to the inspection point using the improved Big Cane Mouse algorithm, while the local layer uses the dynamic window method to obtain the motion state of dynamic obstacles in the workshop in real time, dynamically adjusting the robot's motion parameters based on velocity space sampling and safety, smoothness, and target approach evaluation functions to achieve dynamic obstacle avoidance. Experiments show that this method reduces global path planning error by 15%-20% and shortens dynamic obstacle avoidance response time to less than 0.3s in complex dynamic factory environments, meeting the robot's all-weather inspection needs and ensuring inspection efficiency and operational safety. Attached Figure Description

[0059] Figure 1 A flowchart illustrating a path planning method for an inspection robot based on an improved large cane rat algorithm, provided in an embodiment of the present invention;

[0060] Figure 2 This is a schematic diagram of the structure of an inspection robot path planning system based on an improved sugarcane rat algorithm, provided in an embodiment of the present invention. Detailed Implementation

[0061] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0062] The specific technical features described in the various embodiments in the detailed implementation can be combined in various ways without contradiction. For example, different implementation methods can be formed by combining different specific technical features. In order to avoid unnecessary repetition, the various possible combinations of the specific technical features in this invention will not be described separately.

[0063] It should also be noted that, in order to avoid obscuring the present invention with unnecessary details, only the structures and / or processing steps closely related to the present invention are shown in the accompanying drawings, while other details that are not closely related to the present invention are omitted.

[0064] Additionally, it should be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. In the following description, the terms "first," "second," etc., are used merely to distinguish different objects and do not indicate any similarity or connection between them. It should be understood that the directional descriptions such as "above," "below," "inside," and "outside" refer to the orientation under normal use conditions.

[0065] In some embodiments, such as Figure 1 As shown, the steps of the inspection robot path planning method based on the improved Big Cane Mouse algorithm include:

[0066] Step S101: Obtain data from within the workshop and determine the optimal inspection path based on the improved sugarcane rat algorithm.

[0067] The process of acquiring data within the workshop is used to identify static and dynamic obstacles and the locations of inspection tasks, thereby constructing a grid environment model of the factory workshop and clarifying the spatial distribution of core elements in the inspection scenario. For example, the steps for constructing this grid environment model of the factory workshop include:

[0068] Static obstacle labeling: Static obstacles such as fixed equipment, shelves, and walls in the workshop are mapped to obstacle areas in a grid map. Grid indexes are used to distinguish between passable areas and obstacle areas, ensuring that the model accurately reflects the actual layout of the workshop and providing clear obstacle location information for the algorithm. Dynamic obstacle characteristic definition: For dynamic obstacles such as pedestrians and moving equipment in the workshop, their initial positions, moving speeds, and movement direction ranges are preset, clarifying the boundaries of the dynamic obstacle's movement patterns and avoiding decision-making biases caused by ambiguity in subsequent obstacle avoidance calculations. Inspection task parameter determination: The robot inspection start point (e.g., charging pile location), end point (e.g., workshop end inspection point), and key inspection target points (e.g., equipment detection points) are set, clarifying the spatial coordinates of each point to ensure that path planning can cover the complete inspection task requirements.

[0069] It should be noted that the improved Big Cane Mouse algorithm is an improvement based on the Big Cane Mouse algorithm. The search range of the improved Big Cane Mouse algorithm shrinks with the number of iterations. That is, a more aggressive strategy is used in the early stage of the search to speed up the search, and a more cautious strategy is used in the later stage of the search to improve the search accuracy. In addition, nonlinear and random factors are added in the search process to prevent the search from getting stuck in local optima. The improvements to the Big Cane Mouse algorithm are given below.

[0070] In the exploratory phase of the Big Cane Rat algorithm, the formula for adjusting population position based on leader guidance is:

[0071]

[0072] in, Let be the position of the i-th individual in the (t+1)-th search. Let be the position of the i-th individual in the t-th search. Influence coefficient of optimal solution; For adaptive weight values, For the first Search step size in the next iteration For the first The iteration coefficient at the next iteration;

[0073] Adaptive weight values The calculation formula is:

[0074]

[0075] in, Based on the weighting coefficient, The maximum number of iterations is preset for the algorithm, and t is the number of searches; it should be noted that adaptive weights are designed that decrease linearly with the number of iterations. By dynamically changing the weights, the algorithm balances its global exploration capability in the early stages of iteration with its local development capability in the later stages. , For the first The adaptive weight value in the next iteration ranges from close to the value of the next iteration as the iteration progresses. Gradually decrease to 0;

[0076] Search Step Size The calculation formula is:

[0077]

[0078] in, The step size scaling constant. For the first During the nth iteration The fitness value of each search individual. For the first The optimal fitness value of the population at the next iteration. It is a nonlinear function; nonlinear function such as the exponential function power functions ,in These are nonlinear coefficients; through Dynamically adapting nonlinear characteristics to step size requirements—taking the exponential function as an example, in the early stage of iteration ( When it is smaller), Approaching 1, step size Smaller step size avoids missing local optima due to excessively large step sizes; later stages of iteration ( near hour), near Step length Appropriately increasing the step size accelerates the approach to the global optimum, addressing the weakness of linear step size convergence in the later stages. This can be understood as an adaptive step size. Used to control the magnitude of an individual's displacement "towards the leader / companion," it doesn't change "where to go" (determined by the direction vector), but rather "how far to go." Worse-performing individuals... Larger steps will be assigned to larger steps, thus more aggressively approaching a better solution; the later the steps (…). (The step size decreases or increases to a certain threshold, and then shrinks, which is beneficial for convergence.)

[0079] Iteration coefficients The calculation formula is:

[0080]

[0081] in, This is the iterative scaling constant. For the first The location of the global optimal solution for the population in the next iteration. The variable adjustment coefficient, For the first The random perturbation value at each iteration, taking values ​​between 0 and 1. Iteration coefficients. Used to adjust the bias in individual position updates, its most direct effect is to shift the focus away from the "global optimum" (i.e., the bias in individual position updates). The expansion of the "" into an attraction center with controllable random offset both preserves exploitation (contraction around the optimum) and... It provides exploration.

[0082] Furthermore, during the search process, the search direction of the population is randomly adjusted using sine or cosine functions, with the specific formula as follows:

[0083]

[0084] or

[0085]

[0086] In the formula, the sine or cosine function is randomly selected before each calculation. Let be the position of the i-th individual in the (t+1)-th search. Let be the position of the i-th individual in the t-th search. For adaptive weight values, For the first Search step size in the next iteration For the first The iteration coefficients for the next iteration; r1 is the decreasing adjustment coefficient; r2 and r3 are two random angle parameters. The influence coefficient of the optimal solution;

[0087] The formula for calculating the decreasing adjustment coefficient r1 is:

[0088]

[0089] The two random angle parameters are as follows:

[0090]

[0091]

[0092] They are respectively Different random numbers uniformly distributed within an interval.

[0093] It should be noted that, With the number of iterations Linearly decreasing, with the value range decreasing from 1 to 0, it is used to dynamically control the adjustment intensity of the sine and cosine strategies, realizing the transition of "strong exploration in the early stage and strong development in the later stage"; , The range of values ​​is It is used to generate diverse trigonometric function phases, ensuring the randomness and diversity of the search direction.

[0094] If a better population location than the existing optimal population is found, all population members need to migrate to this new optimal location. The formula for updating the population location based on the new optimal location and peer guidance is as follows:

[0095]

[0096] In the formula, Let be the position of the i-th individual in the (t+1)-th search. Let be the position of the i-th individual in the t-th search. For the search scaling constant, The variable adjustment coefficient, For adaptive weight values, For the first Search step size in the next iteration For the first The average position of all searched individuals at the next iteration.

[0097] The above methods are for searching populations during the non-rainy season. The following describes a method for generating new populations during the rainy season. First, candidate next-generation populations are generated using the following formula:

[0098]

[0099] in, This represents the position of the i-th individual in the next generation population. This represents the position of the i-th individual in the previous generation population. Cross factor For the first The iteration coefficient at the next iteration;

[0100] Then, based on the greedy pruning principle, the next generation population is generated by selecting the best candidate from the previous generation population and the candidate next generation population, using the following formula:

[0101]

[0102] in, for fitness coefficient, for The fitness coefficient.

[0103] It should also be noted that the fitness coefficient is related to the path length and the distance to obstacles. The shorter the path length, the lower the fitness coefficient. The farther the distance to obstacles, the lower the fitness coefficient. A lower fitness coefficient indicates that the population is in a better position.

[0104] Step S102: Determine the velocity window and angular velocity window based on robot motion constraints and safety braking constraints, and sample multiple candidate velocities and multiple candidate angular velocities within the velocity window and angular velocity window.

[0105] Specifically, the dynamic window is the range of speeds the robot can currently execute, determined by both dynamic constraints and safety braking constraints, ensuring that the speed conforms to physical capabilities and obstacle avoidance safety.

[0106] Dynamic constraint formula: Considering the robot's current speed and maximum acceleration / deceleration limits, calculate the feasible speed boundary:

[0107]

[0108] in, , Given the robot's current linear velocity and angular velocity, For time step, , These are the maximum linear acceleration and angular acceleration.

[0109] Safety braking constraint formula: To ensure the robot can safely stop with maximum deceleration before an obstacle, calculate the upper limit of safe linear velocity:

[0110]

[0111] in, To predict the minimum distance from the trajectory to the nearest obstacle, This is the maximum deceleration (take the absolute value).

[0112] Final dynamic window: Combining the two types of constraints, the effective window for linear velocity and angular velocity is obtained:

[0113]

[0114] In dynamic window Inside, according to the set step size (Linear velocity sampling resolution) and Discrete sampling is performed at (angular velocity sampling resolution) to generate several candidate velocity pairs. .

[0115] Step S103: Generate multiple predicted paths based on candidate velocities and candidate angular velocities, score each predicted path by proximity, safety, and velocity smoothness, and determine the predicted path with the highest score as the motion path.

[0116] Predicted paths are used to guide the robot to local target positions. This can be understood as discretizing the overall path into multiple continuous paths by having the robot gradually approach each local target position and ultimately reach the overall target position. The formula for generating multiple predicted paths is given below:

[0117] For each candidate velocity pair Using a differential drive robot motion model to predict the future period The trajectory within 1-3 seconds consists of a series of discrete points. The composition and position update formula are as follows:

[0118]

[0119] in, To predict the first The position and orientation of the step, For predicting time intervals, , This represents the current candidate velocity pair.

[0120] After generating multiple predicted paths, each predicted path needs to be scored based on its approachability, safety, and velocity smoothness. The predicted path with the highest score is then determined as the motion path. The specific scoring formula is as follows:

[0121]

[0122] in, , , Weighting coefficient (adjustable according to the inspection scenario requirements, such as increasing it in a narrow workshop). ), , , The scores are for target approach, safety, and speed smoothness, respectively.

[0123] Target proximity (G): Measures the directional consistency between the trajectory endpoint and the local target point, calculated using the cosine of the direction angle; the closer the value is to 1, the better.

[0124]

[0125] in, This is the angle between the robot's forward direction and the "vector from the current position to the local target point".

[0126] Safety (O): Ensures the trajectory maintains a safe distance from obstacles, calculated using the normalized value of the closest distance; a higher value indicates greater safety.

[0127]

[0128] in, To establish a safe distance, This represents the maximum possible distance within the window.

[0129] Speed ​​smoothness (V): Maintains motion continuity and avoids abrupt acceleration and deceleration; calculated using the deviation between the current speed and the desired speed.

[0130]

[0131] in, The desired linear velocity guided by the global path. Let be the actual speed of the i-th predicted path.

[0132] Step S104: Move the robot along the motion path to the local target point, recalculate the velocity window and angular velocity window, and redetermine the motion path until the robot moves to the global target point.

[0133] This can be understood as follows: after the robot moves along the motion path to the local target point, steps S102 and S103 are re-executed based on the current position and the position of the current obstacle, thereby recalculating the velocity window and angular velocity window, resampling and calculating multiple predicted paths, scoring them, and obtaining a new motion path so that the robot can continue to move along the motion path to the next local target point.

[0134] In some other embodiments, step S104 includes: if the distance between the local target point and the optimal inspection path exceeds a distance threshold, then the local target point is used as the starting point and the global target point is used as the ending point to recalculate the optimal inspection path using the improved Big Cane Mouse algorithm.

[0135] This can be understood as performing global path feedback correction during robot inspection. If local obstacle avoidance causes the robot to deviate from the global path by more than a threshold (such as 2 grid units in a grid map), the current position and obstacle status are fed back to the improved Big Cane Mouse algorithm, triggering a local replanning of the global path. Starting from the current position and ending at the global target point, a corrected path adapted to the dynamic obstacle is generated to ensure that "local obstacle avoidance does not deviate from the global target".

[0136] Optionally, during robot movement, the current position, speed, and direction of movement of dynamic obstacles in the workshop are acquired in real time using LiDAR and vision sensors. The minimum distance between the predicted trajectory and the obstacle is recalculated, and the movement path is also recalculated. This can be understood as dynamic obstacle perception and updating during the inspection process: real-time acquisition of workshop environmental data using LiDAR and vision sensors to update the current position of dynamic obstacles. Speed ​​of movement and direction And recalculate the minimum distance between the predicted trajectory and the obstacle. ,like This immediately triggers a new round of speed sampling and evaluation.

[0137] In other words, during the inspection process, the DWA (Local Path Planning) and SCGCRA (Improved Large Cane Mouse Algorithm) global path planning are run in tandem. The SCGCRA global path provides long-term directional guidance for the DWA, avoiding path detours caused by the DWA's "short-sightedness." The DWA local obstacle avoidance provides dynamic obstacle feedback for the SCGCRA global path, enabling the global path to adapt to changes in the dynamic workshop environment. Through this mechanism, the robot can achieve a dynamic obstacle response time of ≤0.2 seconds and a global path deviation rate of ≤5%, meeting the requirements for efficient and safe inspection in complex workshop environments.

[0138] Optionally, during the inspection process, the motion window is calculated by scrolling, and local target points can be determined based on the motion window. The scrolling window is set as follows: with the robot's current position as the center, the radius of the scrolling window is set according to the dynamic obstacle density in the workshop and the robot's movement speed (to ensure that the window can cover the robot's short-term movement range and surrounding dynamic obstacles). The window range is updated in real time as the robot moves.

[0139] Local target point determination: Calculate the intersection of the scrolling window and the SCGCRA global path. If there are multiple intersections, select the intersection closest to the inspection endpoint and without obstacle occlusion as the local target point of DWA. If there is no intersection of the global path within the window or the robot is close to the endpoint, the global endpoint is directly used as the local target point to ensure that the local obstacle avoidance of DWA always revolves around the direction of the global path and avoids the problem of "local optimum, global deviation".

[0140] This invention also provides a path planning system for an inspection robot based on an improved large cane rat algorithm, such as... Figure 2 As shown, the inspection robot path planning system based on the improved Big Cane Rat algorithm includes: a path planning module 100, used to acquire data in the workshop and determine the optimal inspection path based on the improved Big Cane Rat algorithm; a window selection module 200, used to determine the velocity window and angular velocity window based on robot motion constraints and safety braking constraints, and sample multiple candidate velocities and multiple candidate angular velocities within the velocity window and angular velocity window; a path generation module 300, used to generate multiple predicted paths based on candidate velocities and candidate angular velocities, score each predicted path by proximity, safety, and velocity smoothness, and determine the predicted path with the highest score as the motion path; it is also used to make the robot move along the motion path to a local target point, recalculate the velocity window and angular velocity window and redetermine the motion path until the robot moves to the global target point.

[0141] In some embodiments, such as Figure 2 As shown, the path generation module 300 is also used to recalculate the optimal inspection path by taking the local target point as the starting point and the global target point as the ending point when the distance between the local target point and the optimal inspection path exceeds the distance threshold.

[0142] The above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A path planning method for an inspection robot based on an improved large cane rat algorithm, characterized in that, The path planning method for the inspection robot includes: Data from within the workshop is acquired, and the optimal inspection path is determined based on the improved large cane rat algorithm, wherein the search range of the improved large cane rat algorithm decreases with the number of iterations; Based on robot motion constraints and safety braking constraints, velocity windows and angular velocity windows are determined, and multiple candidate velocities and multiple candidate angular velocities are sampled in the velocity windows and the angular velocity windows. Multiple predicted paths are generated based on the candidate velocities and candidate angular velocities. These predicted paths are used to move the robot to a local target point. Each predicted path is scored based on approachability, safety, and speed smoothness. The predicted path with the highest score is determined as the motion path. Approachability indicates the consistency of the direction between the target point of the predicted path and the local target point of the optimal inspection path. Safety indicates the ability to ensure that the trajectory maintains a safe distance from obstacles. Speed ​​smoothness indicates the ability to maintain motion continuity. The robot is made to move along the motion path to a local target point, the velocity window and angular velocity window are recalculated and the motion path is redefined until the robot moves to the global target point. In the improved large cane rat algorithm, the formula for updating the population position based on the new optimal position through peer guidance is as follows: In the formula, Let be the position of the i-th individual in the (t+1)-th search. Let be the position of the i-th individual in the t-th search. For the search scaling constant, The variable adjustment coefficient, For adaptive weight values, For the first Search step size in the next iteration For the first The average position of all search individuals at the time of the next iteration.

2. The inspection robot path planning method according to claim 1, characterized in that, Moving the robot along the motion path to a local target point, recalculating the velocity window and angular velocity window, and redetermining the motion path until the robot reaches the global target point includes: If the distance between the local target point and the optimal inspection path exceeds a distance threshold, the improved Big Cane Mouse algorithm is invoked to recalculate the optimal inspection path, with the local target point as the starting point and the global target point as the ending point.

3. The inspection robot path planning method according to claim 1, characterized in that, The step of moving the robot along the motion path to a local target point, recalculating the velocity window and angular velocity window, and redetermining the motion path until the robot moves to the global target point includes: The system uses lidar and vision sensors to obtain the current position, speed, and direction of movement of dynamic obstacles in the workshop in real time, recalculates the minimum distance between the predicted trajectory and the obstacle, and recalculates the movement path.

4. The inspection robot path planning method according to claim 1, characterized in that, In the improved large cane rat algorithm, the formula for adjusting population position based on leader guidance is: in, Let be the position of the i-th individual in the (t+1)-th search. Let be the position of the i-th individual in the t-th search. The influence coefficient of the optimal solution; For adaptive weight values, For the first Search step size in the next iteration For the first The iteration coefficient at the next iteration; Adaptive weight values The calculation formula is: in, Based on the weighting coefficient, The maximum number of iterations is preset for the algorithm, where t is the number of searches; Search Step Size The calculation formula is: in, The step size scaling constant. For the first During the nth iteration The fitness coefficient of each search individual. For the first The optimal fitness coefficient of the population at the next iteration. It is a nonlinear function; the fitness coefficient is related to the path length and the distance to obstacles. The shorter the path length, the lower the fitness coefficient. The farther the distance to obstacles, the lower the fitness coefficient. The lower the fitness coefficient, the better the position of the population. Iteration coefficients The calculation formula is: in, This is the iterative scaling constant. For the first The location of the global optimal solution for the population in the next iteration. The variable adjustment coefficient, For the first The random perturbation value during the next iteration, wherein the perturbation value takes a value between 0 and 1.

5. The inspection robot path planning method according to claim 1, characterized in that, In the improved large cane rat algorithm, the formulas for adjusting the population position based on leader guidance and adjusting the search direction based on sine and cosine functions are as follows: or In the formula, the sine or cosine function is randomly selected before each calculation. Let be the position of the i-th individual in the (t+1)-th search. Let i be the position of the i-th individual in the t-th search. For adaptive weight values, For the first Search step size in the next iteration For the first The iteration coefficients for the next iteration; r1 is the decreasing adjustment coefficient; r2 and r3 are two random angle parameters. The influence coefficient of the optimal solution; The formula for calculating the decreasing adjustment coefficient r1 is: The two random angle parameters are as follows: They are respectively Different random numbers uniformly distributed within an interval.

6. The path planning method for inspection robots according to claim 1, characterized in that, In the improved large cane rat algorithm, the formula for generating the candidate next generation population is: in, This represents the position of the i-th individual in the next generation population. This represents the position of the i-th individual in the previous generation population. Cross factor For the first The iteration coefficient at the next iteration; The formula for selecting the next generation of the population is: in, for fitness coefficient, for The fitness coefficient is related to the path length and the distance to obstacles. The shorter the path length, the lower the fitness coefficient. The farther the distance to obstacles, the lower the fitness coefficient. A lower fitness coefficient indicates that the population is in a better position.

7. The inspection robot path planning method according to claim 1, characterized in that, The scoring of each predicted path based on proximity, safety, and speed smoothness includes: The score for each predicted path is calculated using a path evaluation formula, which is: in, A score for the predicted path. , , Here, G represents the target approach rate, O represents the safety factor, and V represents the velocity smoothness factor. The formula for calculating the goal-approachability G is: in, The angle between the robot's forward direction and the "vector from the current position to the local target point"; The formula for calculating safety O is: in, To establish a safe distance, This represents the maximum possible distance within the window. To predict the minimum distance from the trajectory to the nearest obstacle; The formula for calculating velocity smoothness V is: in, The desired linear velocity guided by the global path. Let be the actual speed of the i-th predicted path.

8. A path planning system for an inspection robot based on an improved cane rat algorithm, characterized in that, The inspection robot path planning system includes: The path planning module is used to acquire data within the workshop and determine the optimal inspection path based on the improved sugarcane rat algorithm. The window selection module is used to determine a velocity window and an angular velocity window based on robot motion constraints and safety braking constraints, and to sample multiple candidate velocities and multiple candidate angular velocities in the velocity window and the angular velocity window. The path generation module is used to generate multiple predicted paths based on the candidate velocities and candidate angular velocities, score each predicted path by proximity, safety, and velocity smoothness, and determine the predicted path with the highest score as the motion path; it is also used to make the robot move along the motion path to a local target point, recalculate the velocity window and angular velocity window and redetermine the motion path until the robot moves to the global target point. In the improved large cane rat algorithm, the formula for updating the population position based on the new optimal position through peer guidance is as follows: In the formula, Let be the position of the i-th individual in the (t+1)-th search. Let be the position of the i-th individual in the t-th search. For the search scaling constant, The variable adjustment coefficient, For adaptive weight values, For the first Search step size in the next iteration For the first The average position of all searched individuals at the next iteration.

9. The inspection robot path planning system according to claim 8, characterized in that, The path generation module is further configured to, when the distance between the local target point and the optimal inspection path exceeds a distance threshold, use the local target point as the starting point and the global target point as the ending point to call the improved Big Cane Mouse algorithm to recalculate the optimal inspection path.

Images