A multi-target task-oriented inspection robot path planning method and system

Abstract

The application discloses a kind of multi-target task-oriented inspection robot path planning method and system, it is related to mobile robot autonomous navigation technical field, including for ACO algorithm, introduce obstacle dynamic penalty and pheromone self-adapting control mechanism, real-time feedback environmental information, optimization pheromone update and evaporation factor, optimal access sequence is efficiently solved, avoid path detour.For Q-RRT algorithm, introduce Gaussian target-oriented sampling, establish probability gradient field to guide node directional extension;Adopt environment perception and target distance bimodal dynamic step, adaptive regulation exploration step;Application path fold point dynamic radius tangent circle smoothing strategy, extract angle feature realizes curvature continuous transition, improve path smoothness and motion stability.The application uses above-mentioned method and system, improves the optimality of inspection path, obstacle avoidance reliability and the continuity of motion trajectory.

Description

Technical Field

[0001] This invention relates to the field of autonomous navigation technology for mobile robots, and in particular to a path planning method and system for inspection robots oriented towards multi-objective tasks. Background Technology

[0002] With the continuous improvement of industrial automation, mobile robots are increasingly widely used in facility inspection, intelligent manufacturing, warehousing and logistics, and other fields. In these application scenarios, robots typically need to visit multiple target points sequentially to complete inspection or operation tasks, which places higher demands on path planning technology.

[0003] Traditional path planning methods are primarily designed for single-objective point navigation. When applied to multi-objective inspection tasks, they face numerous challenges: Firstly, directly extending single-objective planning algorithms to multi-objective planning often results in circuitous paths, low efficiency, and difficulty in obtaining optimal solutions in complex obstacle environments. Secondly, existing multi-objective planning methods often employ a simple "sequence first, path later" hierarchical strategy. In the sequence planning stage, the environment is typically treated as free space, with only Euclidean distance as the optimization objective, ignoring the impact of actual obstacle distribution on the feasibility of the access sequence. This leads to theoretically optimal sequences potentially being unexecutable or requiring significant detours in real-world environments due to obstacles. At the path generation level, random sampling-based planning algorithms such as RRT and RRT2... While possessing high computational speed, it suffers from problems such as strong randomness, slow convergence speed, and tortuous generation paths. Although subsequent Q-RRT proposals... While the algorithm has improved convergence efficiency, it still suffers from shortcomings in complex obstacle environments, such as unstable path quality and a lack of consideration for motion smoothness, making it difficult to meet the requirements of inspection robots for path execution accuracy and motion stability. Therefore, there is an urgent need for a multi-objective path planning method that can balance global sequence optimization and local path quality to achieve efficient, safe, and smooth inspection path planning in complex industrial environments. Summary of the Invention

[0004] The purpose of this invention is to provide a path planning method and system for inspection robots oriented towards multi-objective tasks, so as to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, this invention provides a path planning method for inspection robots oriented towards multi-objective tasks, comprising: S1. Based on the improved ACO algorithm, while considering the geometric distance between target points, the influence of environmental obstacle distribution on path feasibility is introduced. By establishing a dynamic obstacle penalty mechanism and a dynamic adjustment mechanism for pheromone and volatile factors, environmental geometric constraints are quantified into evaluation parameters for sequence optimization, thus obtaining the optimal target point access sequence that balances efficiency and feasibility. S2. Based on the optimal target point access sequence obtained in step S1, path planning is performed on adjacent target point pairs in the sequence in turn. For each adjacent target point pair, a multi-strategy fusion path search method is adopted. Oriented sampling guidance is achieved by establishing a probability gradient field centered on the target point. Adaptive step size adjustment is achieved based on the density of environmental obstacles and the robot's motion state. The path smoothing is achieved by adopting the dynamic radius tangent circle method, and a high-quality collision-free path is generated. S3. Connect the high-quality, collision-free paths planned for each pair of adjacent target points in the optimal access sequence in step S2 in sequence to form a complete inspection path that conforms to the robot's kinematic constraints.

[0006] Preferably, step S1 specifically includes: S11. Model the multi-target point inspection task as a traveling salesman problem with obstacle constraints, where nodes represent inspection target points and edge weights represent the estimated cost of point-to-point paths. S12. Improve ACO algorithm initialization, set ant colony algorithm parameters, including the number of ants. Pheromones as heuristic factors Expected heuristic factors pheromone basic volatile factors pheromone maximum volatile factor pheromone release amount Maximum number of iterations, etc.; S13. By establishing a dynamic penalty mechanism for obstacles and a dynamic adjustment mechanism for pheromones and volatile factors, environmental geometric constraints are quantified into evaluation parameters for sequence optimization. S14. Output the optimal (i.e., the one with the lowest total path cost and the least penalty from obstacles) target point visit sequence.

[0007] Preferably, the specific content of step S13 is as follows: After each ant constructs its path, each segment of that path connecting adjacent target points is uniformly sampled to generate... One sampling point; based on the preset error tolerance ε, obstacle overlap detection is performed; Define a path penalty factor ω(m) to quantify the degree of conflict between the path and obstacles: ; in, For the first Total overlap length of ant paths This is the total path length. () represents the number of times the path overlaps with obstacles. This represents the theoretical maximum number of overlaps (equal to the total number of obstacles). Weighting coefficients ( This is used to balance the effects of length overlap and number overlap. The closer the value is to 1, the more severe the path conflict. Dynamic adjustment mechanism of pheromones and volatile factors: Dynamic volatility factor: based on path penalty factor Dynamic adjustment of volatile factors : ; When there are no path conflicts ( )hour, Maintain high-quality path pheromones; when path conflicts are severe ( When it approaches 1), asymptotic This accelerates the evaporation of pheromones from poor-quality paths and enhances the algorithm's ability to escape local optima.

[0008] Global pheromone update: Chapter After iteration, path segment The pheromone concentration on the screen has been updated to: ; in, For all processes in this iteration The sum of pheromones released by ants.

[0009] Adaptive pheromone increment: the first Only ants on the path segment The incremental amount of pheromones released is: ; in, As a pathway quality modulator, it achieves negative correlation regulation between pheromone enhancement intensity and pathway conflict degree. A lower limit for pheromone concentration is set to ensure exploration capability. .

[0010] Preferably, step S2 contains the following: S21. Construct a path decision model based on obstacle detection; S22. Detect whether the straight path between the current adjacent target points collides with the expanded obstacle generated in step S21. If there is no collision, directly use the straight line segment as a feasible path and discretize it according to the preset step size. If there is a collision, enable the probability sampling method based on Gaussian distribution, establish a probability gradient field centered on the target point, and guide the sampling nodes to be concentrated in the target area. S23. Dynamically adjust the expansion step size of the random tree based on obstacle density and target distance; S24, Improvement of Q-RRT Each inflection point in the polyline path generated by the algorithm is smoothed using a tangent circle with a dynamic radius, and combined with step S23 to expand and optimize the random tree, finally generating an optimized smooth path from the starting point to the target point.

[0011] Preferably, step S21 contains the following: An obstacle expansion strategy is adopted, with a preset safety threshold. The original obstacle boundary is expanded to generate an expanded region as a new restricted area.

[0012] Direct Path Strategy: During the path planning initialization phase, first check whether the straight path between the starting point and the target point collides with the expanded obstacles. If there is no collision, the straight line segment is directly used as a feasible path and discretized according to a preset step size; if there is a collision, proceed to step S22.

[0013] Preferably, the probability density function of the Gaussian distribution in step S22 is: ; in, The coordinates of the target point; The standard deviation is used to control the degree of clustering of sampling points.

[0014] Preferably, step S23 specifically includes: The baseline step size is initially corrected by calculating the obstacle density. The calculation formula is as follows: ; in, This represents the ratio of the area of ​​the obstacles after expansion to the total area of ​​the map. As the reference step size, This is the density weighting coefficient.

[0015] The goal-oriented step size correction step involves a secondary correction based on the distance between the current node and the target point. The calculation formula is as follows: ; in, ; This is the lower limit threshold of the target distance factor.

[0016] The step size boundary constraint step applies upper and lower limit constraints to the corrected step size to ensure that the step size is within a reasonable range: .

[0017] Preferably, step S24 specifically includes: The path bend angle is calculated by extracting adjacent vectors at path nodes and normalizing them. ; in , These are the forward and backward unit vectors, respectively; The dynamic radius determination step is based on the path bend angle. Calculate the diameter of the tangent circle: ; in Use the reference diameter and apply boundary constraints to the calculation results: ; The steps for solving geometric parameters are as follows: calculate the direction vector of the angle bisector, and determine the coordinates of the center of the circle based on geometric relationships. ,in , And solve for the tangent point. , coordinate; The smooth path reconstruction process uses a three-segment splicing method to preserve the original path segments. With arc segment Replace the original inflection point Connecting path segments This forms a smooth path composed of straight line segments and tangent circular arcs.

[0018] A path planning system for an inspection robot oriented towards multi-objective tasks, comprising: The global sequence planning module is configured to be based on the improved ACO algorithm. While considering the geometric distance between target points, it introduces the impact of environmental obstacle distribution on path feasibility. By establishing a dynamic obstacle penalty mechanism and a dynamic adjustment mechanism for pheromones and volatile factors, the environmental geometric constraints are quantified into evaluation parameters for sequence optimization, resulting in the optimal target point access sequence that balances efficiency and feasibility. The local path generation module is configured to perform path planning on adjacent target point pairs in the sequence based on the optimal target point access sequence output by the global sequence planning module. For each adjacent target point pair, a multi-strategy fusion path search method is adopted. Oriented sampling guidance is achieved by establishing a probability gradient field centered on the target point. Adaptive step size adjustment is achieved based on the density of environmental obstacles and the robot's motion state. The path smoothing is achieved by using the dynamic radius tangent circle method to generate a high-quality collision-free path. The global path integration module is configured to sequentially connect the high-quality, collision-free paths planned by the local path generation module for each pair of adjacent target points in the optimal access sequence, forming a complete inspection path that conforms to the robot's kinematic constraints.

[0019] Therefore, the present invention employs the above-mentioned path planning method and system for inspection robots oriented towards multi-objective tasks, which has the following beneficial effects: (1) By introducing a dynamic obstacle penalty mechanism, the overlap length ratio and overlap frequency between the path and the obstacle are quantified into a path penalty factor. This allows the ACO algorithm to not only consider Euclidean distance when planning the order of visiting target points, but also to predict the collision risk of the path.

[0020] (2) By using a goal-oriented Gaussian probability sampling strategy, the random tree is guided to expand rapidly toward the target direction. Combined with the dual-modal dynamic step size adjustment of obstacle density and target distance, the algorithm can use small step size for fine search in dense obstacle areas and large step size for rapid exploration in open areas.

[0021] (3) The dynamic radius tangent circle method is used to smooth the turning point. The diameter of the tangent circle is dynamically adjusted based on the size of the turning angle (the larger the turning angle, the smaller the radius, and the smaller the turning angle, the larger the radius). This achieves a smooth transition of curvature continuity, so that the generated path can meet the actual motion control requirements of the inspection robot and avoids the problem of needing to stop suddenly at the turning point in the traditional broken line path.

[0022] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0023] Figure 1 This is a flowchart illustrating an embodiment of a path planning method and system for inspection robots oriented towards multi-objective tasks according to the present invention. Figure 2 This is a schematic diagram illustrating the principle of the RRT algorithm in an embodiment of the present invention; Figure 3 RRT is an embodiment of the present invention. Algorithm principle diagram; Figure 4 This is an embodiment of the Q-RRT invention. Algorithm principle diagram; where (a) is a diagram of reselecting the parent node; (b) is a diagram of pruning; Figure 5 The flowchart of the improved ACO algorithm of this invention is shown in the embodiment of the invention. Figure 6 This is a schematic diagram of the path decision model based on obstacle detection in an embodiment of the present invention (including obstacle expansion, direct connection strategy, and Gaussian sampling). Figure 7 This is a schematic diagram of obstacle boundary expansion according to an embodiment of the present invention; Figure 8 This is a flowchart illustrating the calculation process of the dynamic radius tangent circle smoothing strategy according to an embodiment of the present invention. Figure 9 This is a schematic diagram of path bend angle calculation according to an embodiment of the present invention; Figure 10 This is a schematic diagram of smooth path reconstruction according to an embodiment of the present invention; Figure 11 This is a simulation environment map for Example 1 of the present invention, wherein (a) is a digital twin platform; and (b) is a test scenario. Figure 12 The improved Q-RRT in Example 1 of this invention A schematic diagram of the path corresponding to the algorithm; Figure 13 This is a schematic diagram of the paths corresponding to different algorithms in Example 1 of the present invention; wherein, (a) is the improved Q-RRT. The corresponding path diagrams for the algorithms are shown in (b) and (c). The corresponding path diagram of the algorithm, (d) is Q-RRT The corresponding path diagrams for the algorithms are shown in (e) and (f). The path diagram corresponding to the _connect algorithm. Detailed Implementation

[0024] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0025] Unless otherwise defined, the technical or scientific terms used in this invention shall have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0026] Example Please see Figures 1-13 This invention provides a path planning method for inspection robots oriented towards multi-objective tasks, including: S1. Based on the improved ACO algorithm, while considering the geometric distance between target points, the influence of environmental obstacle distribution on path feasibility is introduced. By establishing a dynamic obstacle penalty mechanism and a dynamic adjustment mechanism for pheromones and volatile factors, environmental geometric constraints are quantified into evaluation parameters for sequence optimization, thus obtaining the optimal target point access sequence that balances efficiency and feasibility.

[0027] Problem Modeling: The multi-target point inspection task is modeled as a traveling salesman problem with obstacle constraints, where nodes represent inspection target points and edge weights represent the estimated cost of point-to-point paths.

[0028] Improved ACO algorithm initialization: Set ant colony algorithm parameters, including the number of ants. Pheromones as heuristic factors Expected heuristic factors pheromone basic volatile factors pheromone maximum volatile factor pheromone release amount Maximum number of iterations, etc.

[0029] Obstacle dynamic penalty mechanism: After each ant constructs its path, each segment of that path connecting adjacent target points is uniformly sampled to generate... One sampling point.

[0030] Obstacle overlap detection is performed based on a preset error tolerance ε.

[0031] Define a path penalty factor ω(m) to quantify the degree of conflict between the path and obstacles: ; in, For the first Total overlap length of ant paths This is the total path length. () represents the number of times the path overlaps with obstacles. This represents the theoretical maximum number of overlaps (equal to the total number of obstacles). Weighting coefficients ( This is used to balance the effects of length overlap and number overlap. The closer the value is to 1, the more severe the path conflict.

[0032] Dynamic adjustment mechanism of pheromones and volatile factors: Dynamic volatility factor: based on path penalty factor Dynamic adjustment of volatile factors : ; When there are no path conflicts ( )hour, Maintain high-quality path pheromones; when path conflicts are severe ( When it approaches 1), asymptotic This accelerates the evaporation of pheromones from poor-quality paths and enhances the algorithm's ability to escape local optima.

[0033] Global pheromone update: Chapter After iteration, path segment The pheromone concentration on the screen has been updated to: ; in, For all processes in this iteration The sum of pheromones released by ants.

[0034] Adaptive pheromone increment: the first Only ants on the path segment The incremental amount of pheromones released is: ; in, As a pathway quality modulator, it achieves negative correlation regulation between pheromone enhancement intensity and pathway conflict degree. A lower limit for pheromone concentration is set to ensure exploration capability. .

[0035] Sequence output: After iteration, output the optimal (i.e., the one with the lowest total path cost and the least penalty from obstacles) target point visit sequence.

[0036] S2. Based on the optimal target point access sequence obtained in step S1, path planning is performed on adjacent target point pairs in the sequence in turn. For each adjacent target point pair, a multi-strategy fusion path search method is adopted. Oriented sampling guidance is achieved by establishing a probability gradient field centered on the target point. Adaptive step size adjustment is achieved based on the density of environmental obstacles and the robot's motion state. The path smoothing is achieved by adopting the dynamic radius tangent circle method, generating a high-quality collision-free path.

[0037] Based on the optimal target point access sequence obtained in step S1, path planning is performed sequentially for adjacent target point pairs in the sequence. For each pair of starting and target points, the improved Q-RRT is executed. algorithm: Path decision model based on obstacle detection: Obstacle expansion and safe zone construction: To ensure the safe operation of the robot, an obstacle expansion strategy is adopted, based on a preset safety threshold. The original obstacle boundary is expanded to generate an expanded region as a new restricted area.

[0038] Direct Path Strategy: During the path planning initialization phase, it first checks whether the straight path between the starting point and the target point collides with the expanded obstacles. If there is no collision, the straight line segment is directly adopted as a feasible path and discretized according to a preset step size; if there is a collision, subsequent target-oriented sampling is performed.

[0039] Goal-oriented Gaussian probability sampling strategy: To improve search efficiency, a sampling method based on Gaussian distribution is adopted to establish a probability gradient field centered on the target point.

[0040] The probability density function is: ; in, The coordinates of the target point are used as the center of the mean of the Gaussian distribution; The standard deviation controls the clustering of sampling points. This distribution makes sampling points appear more densely near the target point, achieving a gradient distribution of "dense near and sparse far," guiding the random tree to expand rapidly in the target direction.

[0041] Dual-modal dynamic step size adjustment strategy based on environmental perception and target distance: Environmental perception layer: First, assess the complexity of the environment.

[0042] Obstacle density: ; in, The total area of ​​all expanded obstacles. This represents the total area of ​​the map.

[0043] Preliminary dynamic step size: ; As the reference step size, This is the density weighting coefficient. This formula establishes a negative correlation between step size and obstacle density.

[0044] Target-oriented layer: performs secondary correction based on the distance between the current extended node and the target point.

[0045] Target distance factor: ; in, The lower threshold is... This represents the Euclidean distance from the current node to the target point. This represents the maximum possible distance, such as the length of the map's diagonal.

[0046] Secondary correction step size: This step achieves a positive correlation between step size and target distance: increasing step size accelerates exploration when moving away from the target, and decreasing step size improves accuracy when moving closer to the target.

[0047] Boundary constraints: Imposing constraints on the final step size to ensure reasonableness. Avoid step size that is too small, resulting in insufficient exploration ability, or step size that is too large, resulting in path dispersion.

[0048] Path inflection point dynamic radius tangent circle smoothing strategy: Angle calculation: For inflection points in the path Obtain its forward vector and backward vector After unitization, we get and Calculate the angle: , .

[0049] Dynamic diameter calculation: Establishing the diameter of the tangent circle and angle Inverse proportional relationship model: radius The anchoring coefficient ( (Base diameter).

[0050] Therefore, diameter .

[0051] Apply boundary constraints to the diameter: .

[0052] Solving for the parameters of the tangent circle: Calculate the direction of the angle bisector: .

[0053] like (Small threshold), then normalized to get ;like ,illustrate Skip smoothing.

[0054] Calculate the coordinates of the center of the circle: radius Distance between centers center .

[0055] Calculate the coordinates of the tangent point: tangent point It is the center of the circle To path segment The dangling foot, It is the center of the circle To path segment The foot of the perpendicular is determined by vector projection.

[0056] Smooth path reconstruction: using straight line segments Circular arc segment straight line segment Replace the original inflection point And its adjacent path segments, to achieve a smooth transition of curvature continuity.

[0057] Q-RRT Core process enhancement: In standard Q-RRT In the parent node selection and rewiring stage of the algorithm, the algorithm expands the candidate parent node set (including neighboring nodes and their multi-level ancestors) by taking advantage of its feature. Combined with the dynamic step size and smoothing strategy proposed in this invention, the random tree is expanded and optimized, and finally an optimized smooth path from the starting point to the target point is generated.

[0058] S3. Connect the high-quality, collision-free paths planned for each pair of adjacent target points in the optimal access sequence in step S2 in sequence to form a complete inspection path that conforms to the robot's kinematic constraints.

[0059] The local fine-grained paths planned for each pair of adjacent target points in the optimal access sequence in step S2 are connected sequentially to form a complete, smooth, and safe global inspection path that starts from the starting point, traverses all target points, and finally returns to the starting point (or reaches the end point).

[0060] A path planning system for an inspection robot oriented towards multi-objective tasks, comprising: The global sequence planning module is configured to be based on the improved ACO algorithm. While considering the geometric distance between target points, it introduces the impact of environmental obstacle distribution on path feasibility. By establishing a dynamic obstacle penalty mechanism and a dynamic adjustment mechanism for pheromones and volatile factors, the environmental geometric constraints are quantified into evaluation parameters for sequence optimization, resulting in the optimal target point access sequence that balances efficiency and feasibility. The local path generation module is configured to perform path planning on adjacent target point pairs in the sequence based on the optimal target point access sequence output by the global sequence planning module. For each adjacent target point pair, a multi-strategy fusion path search method is adopted. Oriented sampling guidance is achieved by establishing a probability gradient field centered on the target point. Adaptive step size adjustment is achieved based on the density of environmental obstacles and the robot's motion state. The path smoothing is achieved by using the dynamic radius tangent circle method to generate a high-quality collision-free path. The global path integration module is configured to sequentially connect the high-quality, collision-free paths planned by the local path generation module for each pair of adjacent target points in the optimal access sequence, forming a complete inspection path that conforms to the robot's kinematic constraints.

[0061] Example 1 This embodiment uses the actual layout of a PC component intelligent production workshop as a background, and is conducted in a high-precision digital twin simulation environment of 200m × 100m. Figure 11 (a) verifies the effectiveness of the method described in this invention in complex multi-target point inspection tasks.

[0062] S1: Simulation environment and problem initialization.

[0063] 1. Environmental Modeling: Based on actual factory drawings, a 1:100 accurate geometric model is constructed in the simulation platform, including static obstacles such as production line equipment and material storage areas, forming a model like... Figure 11 The standardized test scenario is shown in (b) of the diagram.

[0064] 2. Target point setting: In this embodiment, 12 target points to be inspected are set, and their coordinate set is: [(80,140),(235,185),(350,140),(280,40),(210,90),(190,100),(120,50),(20,20),(45,100),(350,100),(250,130),(175,120)].

[0065] 3. Task Definition: The inspection robot’s task is to start from a preset starting point, visit the remaining 11 target points in sequence (the visiting order is determined by global sequence planning), and finally return to the starting point or end at the specified endpoint. The planned global path must meet the requirements of no collision, short length, and high smoothness.

[0066] S2: Global target point access sequence planning based on improved ACO algorithm.

[0067] This step aims to find a target point visit sequence that minimizes the total travel cost while considering obstacle constraints.

[0068] 1. Algorithm parameter initialization: The running parameters of the improved ACO algorithm are set as follows: Ant count pheromone concentration =1; Information weighting factor =5; Basic volatile factor of pheromones ; Maximum volatile factor of pheromones pheromone release amount Maximum number of iterations = 60.

[0069] 2. Iterative optimization process: Path Construction and Evaluation: In each iteration, each ant constructs a path traversing all 12 target points based on the state transition probabilities. For each straight line connection in the path, uniform sampling is performed, and its overlap with inflated obstacles is detected. The path penalty factor for this path is then calculated. This factor comprehensively reflects the proportion of overlap length and overlap frequency between path segments and obstacles.

[0070] Pheromones are dynamically updated: Dynamic evaporation: Based on the penalty factor for each ant's path. Dynamically calculate its volatility factor: .

[0071] This mechanism causes pheromones on poor paths through obstacles to evaporate rapidly.

[0072] Adaptive enhancement: The pheromone update formula is: ; Among them, pheromone increment Linked to path quality, This means that safe, short-path, high-quality solutions will receive more pheromone enhancements, thus guiding subsequent ants to gather in a better search space.

[0073] 3. Sequence Output: After 60 iterations, the average path length steadily decreases with increasing iteration count, and the volatile factor dynamically adjusts according to the degree of path conflict. The algorithm outputs the optimal target point access sequence (i.e., the sequence with the minimum combined cost of actual path length and obstacle penalty) across all iterations. Assume the optimal sequence obtained in this embodiment is: S→P8→P7→P5→P6→P12→P4→P11→P2→P1→P10→P3→P9→S (where P1-P12 correspond to the coordinate points mentioned above).

[0074] S3: Based on improved Q-RRT Local fine-grained path planning in the algorithm.

[0075] Based on the optimal access sequence obtained in S2, the improved Q-RRT is used. The algorithm plans a safe and smooth local path for each pair of adjacent target points in the sequence (e.g., S→P8, P8→P7, ..., P9→S).

[0076] 1. Algorithm parameter initialization: Improved Q-RRT The key parameters of the algorithm are set as follows: security threshold Standard deviation of Gaussian probability sampling Reference step size Density weighting coefficient Lower limit threshold of target distance factor Anchoring coefficient Preset reference diameter ; ; ; The target is close to the threshold ; Backtracking depth .

[0077] 2. Single-segment path planning process (taking P1→P10 as an example): Step 1: Direct Connection Detection. First, check whether the straight connection between P1 and P10 collides with an expanding obstacle. If not, directly use the straight path and jump to the path smoothing step; if so, execute the following search steps.

[0078] Step 2: Target-oriented sampling. Using the target point P10 as the center, perform probability sampling using a Gaussian distribution to generate random points. This makes it more likely that sampling points will appear near P10, accelerating the tree's growth towards the target direction.

[0079] Step 3: Dynamic step size expansion.

[0080] Environmental perception: Calculating global obstacle density Initial adjustment of step size: .

[0081] Goal-oriented: Calculate the distance factor from the current expanded node to P10. A second revision was made: .

[0082] Boundary constraints: final step size Constrained Within a certain range, to balance exploration efficiency and safety.

[0083] Step 4: Q-RRT Core operation. After a new node is generated, the optimal parent node is selected within its extended neighborhood (including multi-level ancestor nodes), and rewiring optimization is performed to ensure the asymptotic optimality of the path.

[0084] Step 5: Path Smoothing. Apply a dynamic radius tangent circle smoothing strategy to the initial path obtained from the search. For each inflection point in the path... : Calculate its steering angle .

[0085] According to the formula Dynamically calculate the diameter of the tangent circle and constrain it to... Within the range. The larger the angle, the smaller the diameter of the smooth circle used, achieving a more refined transition.

[0086] Calculate the center and tangent point of the circle, and replace the original inflection points with a combination of straight line-circular arc-straight line to generate a smooth trajectory with continuous curvature, such as... Figure 10 As shown by the black path in the middle.

[0087] S4: Global path integration.

[0088] The smooth local paths planned for each pair of adjacent target points in the optimal sequence in S3 are seamlessly stitched together in sequence order to form a complete global inspection path that starts from the starting point S, efficiently traverses all 12 target points, and satisfies the robot's motion constraints. A schematic diagram of the path planning results in the simulation environment can be found here. Figure 12 "Improved Q-RRT" The corresponding path.

[0089] S5: Performance verification and results analysis.

[0090] To objectively evaluate the performance of the method of this invention, a comparative experiment was conducted with five mainstream path planning algorithms under the same 12 target point sequence and simulation environment. All algorithms were run 20 times and the average value was taken.

[0091] 1. Evaluation indicators: Path planning quality: Measured by average path length (in pixels or analog units).

[0092] Algorithm running efficiency: Measured by average running time (in seconds).

[0093] Search performance: Measured by the average number of effective nodes. Fewer nodes usually indicate higher search efficiency.

[0094] 2. Experimental Results: The quantitative results are recorded in Table 1, and their visualizations are as follows: Figure 12 As shown.

[0095] Table 1 Comparative experimental data of different algorithms

[0096] 3. Results Analysis: Path Quality: This invention improves Q-RRT The algorithm plans the shortest path (1375.6), compared to traditional RRT and RRT. And the original Q-RRT The algorithm reduced path length by 16.3%, 9.2%, and 6.8%, respectively. This demonstrates the significant effect of the goal-oriented sampling and path smoothing strategies in this invention on improving path optimality.

[0097] Planning efficiency: The average running time of this invention is only 3.56117 seconds, far lower than other algorithms. Especially compared to RRT... and Q-RRT The algorithm improved efficiency by 75.9% and 77.1%. This is mainly attributed to the Gaussian goal-oriented sampling mechanism and the dual-modal dynamic step size adjustment strategy, which greatly reduced unnecessary random exploration and computational redundancy.

[0098] Search performance: The average number of effective nodes in this invention is only 63.85, the lowest among all algorithms, and 61.7% less than the RRT_connect algorithm with the highest number of nodes. This indicates that the search directionality of this invention is extremely strong, and it can quickly find feasible paths with fewer expansion nodes, resulting in excellent search performance.

[0099] This specific embodiment fully demonstrates that the fusion of improved ACO and improved Q-RRT provided by the present invention The collaborative path planning method can significantly improve path quality, planning efficiency and search performance simultaneously when solving multi-target inspection tasks in complex industrial environments, thus fully achieving the intended purpose of this invention.

[0100] Therefore, the present invention adopts the above-mentioned path planning method and system for inspection robots oriented to multi-target tasks. Through the above-mentioned multi-dimensional mechanism improvement and algorithm synergy, it effectively solves key problems such as path optimization, dynamic obstacle avoidance and trajectory smoothing in multi-target inspection, ensuring that the robot can efficiently, safely and stably complete multi-target point inspection tasks in complex inspection environments, and significantly improves the optimality of inspection path, obstacle avoidance reliability and motion trajectory continuity.

[0101] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. 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 still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A path planning method for inspection robots oriented towards multi-objective tasks, characterized in that, include: S1. Based on the improved ACO algorithm, while considering the geometric distance between target points, the influence of environmental obstacle distribution on path feasibility is introduced. By establishing a dynamic obstacle penalty mechanism and a dynamic adjustment mechanism for pheromones and volatile factors, environmental geometric constraints are quantified into evaluation parameters for sequence optimization, resulting in an optimal target point access sequence that balances efficiency and feasibility; specifically including: S11. Model the multi-target point inspection task as a traveling salesman problem with obstacle constraints, where nodes represent inspection target points and edge weights represent the estimated cost of point-to-point paths. S12. Improve ACO algorithm initialization, set ant colony algorithm parameters, including the number of ants. Pheromones as heuristic factors Expected heuristic factors pheromone basic volatile factors pheromone maximum volatile factor pheromone release amount Maximum number of iterations; S13. By establishing a dynamic penalty mechanism for obstacles and a dynamic adjustment mechanism for pheromones and volatile factors, environmental geometric constraints are quantified into evaluation parameters for sequence optimization; the specific content is as follows: After each ant constructs its path, each segment of that path connecting adjacent target points is uniformly sampled to generate... One sampling point; based on the preset error tolerance ε, obstacle overlap detection is performed; Define a path penalty factor ω(m) to quantify the degree of conflict between the path and obstacles: ； in, For the first Total overlap length of ant paths This is the total path length. This represents the number of times the path overlaps with obstacles. This represents the theoretical maximum number of overlaps. These are the weighting coefficients. This is used to balance the effects of length overlap and number overlap. The closer the value is to 1, the more severe the path conflict. Dynamic adjustment mechanism of pheromones and volatile factors: Dynamic volatility factor: based on path penalty factor Dynamic adjustment of volatile factors : ； When there are no path conflicts hour, Maintain high-quality path pheromones; when path conflicts are severe, When it approaches 1, asymptotic This accelerates the evaporation of pheromones from poor-quality paths and enhances the algorithm's ability to escape local optima. No. After iteration, path segment The pheromone concentration on the screen has been updated to: ； in, For all processes in this iteration The sum of pheromones released by ants; No. Only ants on the path segment The incremental amount of pheromones released is: ； in, As a pathway quality adjustment factor, a lower limit for pheromone concentration is set. ; S14. Output the optimal target point access sequence; S2. Based on the optimal target point access sequence obtained in step S1, perform path planning on adjacent target point pairs in the sequence in turn to generate a high-quality collision-free path. S3. Connect the high-quality, collision-free paths planned for each pair of adjacent target points in the optimal access sequence in step S2 in sequence to form a complete inspection path that conforms to the robot's kinematic constraints.

2. The path planning method for an inspection robot oriented towards multi-objective tasks according to claim 1, characterized in that, Step S2 is as follows: S21. Construct a path decision model based on obstacle detection; S22. Detect whether the straight path between the current adjacent target points collides with the expanded obstacle generated in step S21. If there is no collision, directly use the straight line segment as a feasible path and discretize it according to the preset step size. If there is a collision, enable the probability sampling method based on Gaussian distribution, establish a probability gradient field centered on the target point, and guide the sampling nodes to be concentrated in the target area. S23. Dynamically adjust the expansion step size of the random tree based on obstacle density and target distance; S24, Improvement of Q-RRT Each inflection point in the polyline path generated by the algorithm is smoothed using a tangent circle with a dynamic radius, and combined with step S23 to expand and optimize the random tree, finally generating an optimized smooth path from the starting point to the target point.

3. The path planning method for an inspection robot oriented towards multi-objective tasks according to claim 2, characterized in that, Step S21 is as follows: An obstacle expansion strategy is adopted, with a preset safety threshold. The original obstacle boundary is expanded to generate an expanded region as a new restricted area; In the path planning initialization phase, first check whether the straight path between the starting point and the target point collides with the expanded obstacle; if there is no collision, the straight line segment is directly used as a feasible path and discretized according to the preset step size; if there is a collision, then execute step S22.

4. The path planning method for an inspection robot oriented towards multi-objective tasks according to claim 3, characterized in that, The probability density function of the Gaussian distribution in step S22 is: ； in, The coordinates of the target point; The standard deviation is used to control the degree of clustering of sampling points.

5. The path planning method for an inspection robot oriented towards multi-objective tasks according to claim 4, characterized in that, The specific content of step S23 is as follows: The baseline step size is initially corrected by calculating the obstacle density. The calculation formula is as follows: ； in, This represents the ratio of the area of ​​the obstacles after expansion to the total area of ​​the map. As the reference step size, This refers to the density weighting coefficient; The goal-oriented step size correction step involves a secondary correction based on the distance between the current node and the target point. The calculation formula is as follows: ； in, ; This is the lower limit threshold of the target distance factor; Apply upper and lower limits to the corrected step size to ensure it remains within a reasonable range: 。 6. The path planning method for an inspection robot oriented towards multi-objective tasks according to claim 5, characterized in that, Step S24 specifically includes: The path bend angle is calculated by extracting adjacent vectors at path nodes and normalizing them. ； in , These are the forward and backward unit vectors, respectively; The dynamic radius determination step is based on the path bend angle. Calculate the diameter of the tangent circle: ； in Use the reference diameter and apply boundary constraints to the calculation results: ； The steps for solving geometric parameters are as follows: calculate the direction vector of the angle bisector, and determine the coordinates of the center of the circle based on geometric relationships. ,in , And solve for the tangent point. , coordinate; The smooth path reconstruction process uses a three-segment splicing method to preserve the original path segments. With arc segment Replace the original inflection point Connecting path segments This forms a smooth path composed of straight line segments and tangent circular arcs.

7. A path planning system for an inspection robot oriented towards multi-objective tasks, comprising a path planning method for an inspection robot oriented towards multi-objective tasks according to any one of claims 1-6, characterized in that, include: The global sequence planning module is configured to be based on the improved ACO algorithm. While considering the geometric distance between target points, it introduces the impact of environmental obstacle distribution on path feasibility. By establishing a dynamic obstacle penalty mechanism and a dynamic adjustment mechanism for pheromones and volatile factors, the environmental geometric constraints are quantified into evaluation parameters for sequence optimization, resulting in the optimal target point access sequence that balances efficiency and feasibility. The local path generation module is configured to perform path planning on adjacent target point pairs in the sequence based on the optimal target point access sequence output by the global sequence planning module. For each adjacent target point pair, a multi-strategy fusion path search method is adopted. Oriented sampling guidance is achieved by establishing a probability gradient field centered on the target point. Adaptive step size adjustment is achieved based on the density of environmental obstacles and the robot's motion state. The path smoothing is achieved by using the dynamic radius tangent circle method to generate a high-quality collision-free path. The global path integration module is configured to sequentially connect the high-quality, collision-free paths planned by the local path generation module for each pair of adjacent target points in the optimal access sequence, forming a complete inspection path that conforms to the robot's kinematic constraints.

Images