Coverage maintenance type hierarchical dynamic obstacle avoidance method and system for unmanned cleaning vehicle

Abstract

The application discloses a kind of covering keeping type hierarchical dynamic obstacle avoidance method and system for unmanned cleaning vehicle, it is related to unmanned cleaning vehicle technical field, the method includes the following steps: obtaining full coverage path and obstacle information;Calculate TTC, and according to this, obstacle is divided into five risk levels of no risk, low risk, medium risk, high risk and blockage;Differentiated obstacle avoidance strategy is generated according to risk level, including keeping path, deceleration, parking or DWA local obstacle avoidance and A* bypass;TTC is used as driving parameter Dynamic adjustment improves the weight of coverage path keeping term and TTC risk penalty term in DWA evaluation function;Judge whether deviate from full coverage path, deviate when record unswept area and guide regression or supplementary cleaning;Blockage level uses A* bypass and records the missing scanning track point to missing scanning queue;Until full coverage path and all track points in missing scanning queue are cleaned up.The application maintains the continuity of cleaning path while ensuring obstacle avoidance safety, effectively reduces the missing scanning rate.

Description

Technical Field

[0001] This invention relates to the field of unmanned sweeping vehicle technology, and in particular to a coverage-maintaining hierarchical dynamic obstacle avoidance method and system for unmanned sweeping vehicles. Background Technology

[0002] Unmanned cleaning vehicles, as a typical type of service mobile robot, are widely used in open or semi-open environments such as campuses, parks, squares, and roads. When performing cleaning tasks, unmanned cleaning vehicles typically need to travel back and forth along a pre-planned full-coverage path to ensure that the target area is cleaned as thoroughly as possible.

[0003] In the autonomous navigation process of unmanned cleaning vehicles, path planning and dynamic obstacle avoidance are crucial factors affecting their operational efficiency, safety, and coverage integrity. Existing unmanned cleaning vehicles typically employ full-coverage path planning methods to generate cleaning routes, combined with local obstacle avoidance algorithms to handle dynamic obstacles such as pedestrians, bicycles, and electric vehicles. Among these, the Dynamic Window (DWA) method is a commonly used approach for local obstacle avoidance in mobile robots. This algorithm samples linear and angular velocities in velocity space and selects the optimal trajectory based on an evaluation function, offering advantages such as strong real-time performance, low computational complexity, and ease of engineering implementation.

[0004] However, traditional DWA still has the following shortcomings in scenarios where unmanned sweepers cover the entire area: First, traditional DWA relies solely on obstacle distance to assess collision risk, neglecting obstacle speed and movement trends. This leads to sluggish response to rapidly approaching distant obstacles and excessive avoidance of stationary near obstacles, resulting in unnecessary detours. Second, traditional DWA uses only the target point direction as guidance, causing significant deviations from the preset coverage path during obstacle avoidance, leading to the accumulation of missed areas. While existing sweeper obstacle avoidance solutions incorporate missed area compensation, they employ post-event area calculations, making the problem of missed area accumulation still significant in scenarios with frequent obstacle appearances. Third, traditional DWA uses a uniform detour strategy for all obstacles. Detours for momentarily crossing obstacles disrupt the continuity of the coverage path; for persistent, obstructive obstacles, waiting strategies cause task interruptions, lacking a tiered and differentiated response mechanism.

[0005] While existing technologies have explored applications of predictive collision time (TTC) and risk classification, most methods do not model TTC risk classification in a unified manner with full-coverage path-keeping mechanisms, making it difficult to simultaneously meet the requirements of dynamic obstacle avoidance safety and full-coverage sweep integrity. Summary of the Invention

[0006] The purpose of this invention is to provide a coverage-maintaining hierarchical dynamic obstacle avoidance method and system for unmanned sweepers, in order to solve the technical problems of insufficient risk assessment, poor coverage continuity and single strategy response in the prior art, and to achieve the organic unity of safe obstacle avoidance and high coverage operation of unmanned sweepers in dynamic environments.

[0007] The technical solution adopted by this invention to solve the above-mentioned technical problems is: a coverage-maintaining hierarchical dynamic obstacle avoidance method for unmanned sweeping vehicles, comprising the following steps: S1. Obtain the full coverage path, current pose and motion status of the unmanned sweeping vehicle; S2. Obtain information on the position, speed, direction of motion, and category of surrounding obstacles through sensors, and construct an obstacle state set; S3. Based on the obstacle information obtained in S2, calculate the predicted collision time between the unmanned cleaning vehicle and each obstacle, and classify the obstacles into five risk levels: no risk, low risk, medium risk, high risk, and blockage, according to the relationship between the predicted collision time and a preset threshold. The risk level is triggered when the duration of the obstacle stay exceeds the occupation time threshold. S4. Based on the risk level determined in S3, generate the corresponding obstacle avoidance strategy: No-risk level: Maintain coverage path; Low-risk level: Reduce speed; Medium-risk level: Continuously reduce speed; High-risk level: Stop or use dynamic window method for local obstacle avoidance; Blockage level: Switch to A* detour mode and simultaneously implement a missed scan recording mechanism. S5. For low-risk, medium-risk, high-risk, and blocking risk levels, the weights of each item in the dynamic window method evaluation function are dynamically adjusted and improved, using the current predicted collision time as the driving parameter. S6. Determine whether the current trajectory will cause the unmanned cleaning vehicle to deviate from the full coverage path. If it does not deviate from the path, continue to run. If it deviates from the full coverage path, record the uncleaned area and guide the unmanned cleaning vehicle back to the full coverage path or perform supplementary cleaning on the uncleaned area after the obstacle risk is reduced or disappears.

[0008] S7. For risk levels, the A* algorithm is used to plan detour routes, and during the detour, all unswept track points that deviate from the full coverage route are recorded in real time and added to the missed sweep area and merged into the missed sweep queue. S8. Determine whether all track points in the full coverage path and the missed scan queue have been cleaned. If all are cleaned, the task ends; otherwise, return to S2 to continue execution.

[0009] Further, in step S5, the improved dynamic window method evaluation function is:

[0010] in, This indicates the directional consistency evaluation item. This indicates the obstacle distance evaluation item. Indicates the speed evaluation item. This represents the trajectory smoothness evaluation item; Indicates a path-preserving item with full coverage; The term "time of collision" indicates the risk penalty for predicted collisions; TTC indicates the predicted collision time. This represents the weight of each item, where Both represent dynamic weighting functions with the predicted collision time as the independent variable.

[0011] Furthermore, the expression for the full-coverage path-preserving term is:

[0012] Where, d path This represents the minimum distance from the endpoint of the candidate trajectory to the original fully covered path; The expression for predicting the collision time risk penalty is:

[0013] in, This indicates that the value is divided by a small amount to prevent division by zero.

[0014] Further, in step S1, the full coverage path is represented as:

[0015] in, Indicates the first One coverage point, Indicates the total number of covered waypoints; The current pose state and motion state are represented as follows:

[0016] Where, x r and y r θ represents the current location of the unmanned cleaning vehicle. r Let v be the heading angle. r Let ω be the current linear velocity. r The value represents the current angular velocity. The superscript T indicates the transpose of a matrix or vector.

[0017] Further, in step S2, the obstacle state set is represented as:

[0018] Where k represents the k-th obstacle, and This indicates the current coordinates of the obstacle. This indicates the speed of the obstacle. Indicates the direction of the obstacle's movement. This indicates the category of obstacle; the obstacle category includes at least one of pedestrians, motor vehicles, non-motor vehicles, and temporary static obstacles.

[0019] Further, in step S3, the calculation function for the predicted collision time is:

[0020] in, Let be the position vector of the k-th obstacle relative to the unmanned sweeper. Let be the velocity vector of the k-th obstacle relative to the unmanned sweeper.

[0021] Furthermore, when This indicates that the distance between the two is decreasing, posing a potential risk of collision; when This indicates that the distance between the two obstacles is increasing or they are no longer getting closer, and there is no immediate risk of collision. When there are multiple dynamic obstacles, the minimum predicted collision time is taken as the risk assessment criterion for the current moment, and its expression is:

[0022] in, This indicates the number of dynamic obstacles currently detected.

[0023] Furthermore, in step S6, when the unmanned sweeping vehicle performs a detour and deviates from the original full-coverage path, the skipped coverage path segment is recorded in the missed sweeping queue:

[0024] in, This represents the set of coverage track points corresponding to the missed scan area; After recording the missed areas, the nearest unswept coverage track point to the current location of the unmanned sweeper is searched within the original full-coverage path as the regression point. The expression for this regression point is:

[0025] in, p represents the set of cover track points that have not yet been cleared. i Indicates the first in the full coverage path One coverage point, Indicates the current location of the unmanned sweeper; After determining the return point, a return path is generated based on the passable area between the current location of the unmanned sweeper and the return point:

[0026] in, Indicates the first step in the regression path At each path point, the unmanned sweeper returns to the original full-coverage path along the regression path; When the missed scan queue When the location is not empty, the unmanned sweeper generates a re-sweeping path based on the missed sweeping points:

[0027] in, Indicates the scan path, Indicates the first in the scan path One scan path point.

[0028] This invention also proposes a coverage-maintaining hierarchical dynamic obstacle avoidance system for unmanned sweeping vehicles, comprising: The information acquisition module is used to acquire the full coverage path, current pose and motion status of the unmanned sweeper, as well as the position, speed, direction of motion and category information of surrounding obstacles through sensors; The risk classification module is used to calculate the predicted collision time between the unmanned sweeper and each obstacle based on obstacle information, and classify obstacles into five risk levels: no risk, low risk, medium risk, high risk, and blockage based on the relationship between the predicted collision time and a preset threshold. The strategy generation module is used to generate corresponding obstacle avoidance strategies based on the risk level. The strategies include maintaining the covered path, decelerating, continuously reducing speed, stopping or using the dynamic window method for local obstacle avoidance, and A* detour. The trajectory planning module is used to dynamically adjust the weights of various items in the improved dynamic window method evaluation function with the predicted collision time as the driving parameter. The improved dynamic window method evaluation function includes a coverage path preservation term and a predicted collision time risk penalty term. It generates candidate trajectories and outputs local obstacle avoidance trajectories through dynamic window sampling. The missed sweeping compensation module is used to determine whether the current trajectory will cause the unmanned sweeping vehicle to deviate from the full coverage path. When it deviates, it records the uncleaned area and guides the unmanned sweeping vehicle back to the full coverage path or performs supplementary sweeping on the uncleaned area after the obstacle risk is reduced or disappears. The task judgment module is used to determine whether all waypoints in the full coverage path and the missed sweep queue have been cleaned. If so, the task ends; otherwise, the obstacle avoidance and cleaning process continues.

[0029] According to the above technical solution, the beneficial effects of the present invention are: 1. This invention introduces TTC to classify the risks of dynamic obstacles, comprehensively considering the relative position, relative speed and movement trend between the obstacle and the unmanned sweeper. Compared with the traditional DWA that only relies on the distance of the obstacle, the risk assessment is more accurate, which can avoid unnecessary detours and improve the operating efficiency of the unmanned sweeper.

[0030] 2. This invention adopts differentiated obstacle avoidance strategies based on different risk levels, including maintaining the coverage path, slowing down, stopping and waiting, partial detour, and A* detour, so that the unmanned sweeper can make full use of the active obstacle avoidance behavior of dynamic obstacles while ensuring safety, thereby reducing path deviation and the number of stops.

[0031] 3. This invention introduces a coverage path preservation term into the DWA evaluation function, which enables the unmanned sweeper to prioritize maintaining the original full-coverage sweeping path under low-risk or medium-risk conditions, reducing trajectory deviations caused by local obstacle avoidance, thereby improving the continuity of the full-coverage path and the integrity of sweeping.

[0032] 4. This invention introduces a TTC risk penalty term into the DWA evaluation function and combines it with dynamic weight adjustment, enabling the unmanned sweeper to adaptively adjust its movement speed and obstacle avoidance trajectory according to the degree of obstacle risk, thereby improving safety and robustness in dynamic scenarios.

[0033] 5. This invention designs a mechanism for recording missed areas, generating return points, and performing supplementary cleaning. When the unmanned cleaning vehicle deviates from its original coverage path due to detours, it can record the uncleaned areas and automatically return to the coverage path or perform supplementary cleaning tasks after the obstacle risk is reduced or disappears, thereby reducing the missed cleaning rate and improving the coverage rate.

[0034] 6. This invention is applicable to unmanned cleaning scenarios such as campus roads, park roads, squares, and pedestrian walkways where there are dynamic obstacles such as pedestrians, bicycles, and electric vehicles, and has strong practicality and promotional value. Attached Figure Description

[0035] Figure 1 This is a schematic diagram of the method flow of the present invention; Figure 2 This is a schematic diagram of the functional module relationship of the coverage-preserving hierarchical dynamic obstacle avoidance method in an embodiment of the present invention; Figure 3 This is a schematic diagram of the mapping between TTC risk classification and obstacle avoidance strategy in an embodiment of the present invention; Figure 4 This is a diagram showing the change in the safe distance between the unmanned cleaning vehicle and the nearest obstacle in an embodiment of the present invention; Figure 5 This is a graph showing the changes in the linear velocity and angular velocity of the unmanned sweeper in an embodiment of the present invention. Figure 6 This is a diagram showing the full coverage and dynamic obstacle avoidance trajectory of the unmanned sweeper in an embodiment of the present invention. Detailed Implementation

[0036] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0037] like Figure 1-2As shown, the method described in this embodiment includes functional steps such as full-coverage path and vehicle status acquisition, obstacle status perception, TTC calculation and risk level determination, generation of graded obstacle avoidance strategies, improved DWA local trajectory planning, A* detour path planning, missed sweep recording and regression and re-sweeping, and completion judgment of sweeping tasks. The results of full-coverage path and vehicle status acquisition, as well as obstacle status perception, are input into the TTC calculation and risk level determination step. Based on the determined risk level, a corresponding graded obstacle avoidance strategy is generated, and either improved DWA local trajectory planning or A* detour path planning is selected. Missed sweep track points generated during local obstacle avoidance or detour are processed by the missed sweep recording, regression, and re-sweeping steps. After completing the regression or re-sweeping, it is determined whether all sweeping tasks in the full-coverage path and the missed sweep queue have been completed.

[0038] This invention provides a coverage-maintaining hierarchical dynamic obstacle avoidance method for unmanned sweeping vehicles, comprising the following steps: S1. Obtain the map of the unmanned sweeper's operating area, the full coverage path, and its current pose and motion status.

[0039] Specifically, the work area is represented as a two-dimensional raster map:

[0040] in, Indicates the first line, number Column grid status, and These represent the number of grid cells in the horizontal and vertical directions of the raster map, respectively. When the grid is a free area When the grid is an obstacle area, .

[0041] In this embodiment, the unmanned sweeper first generates a full-coverage path based on the work area map. :

[0042] in, Indicates the first One coverage point, This indicates the total number of covered waypoints.

[0043] Simultaneously obtain the current pose status of the unmanned sweeper:

[0044] Where, x r and y r θ represents the current location of the unmanned cleaning vehicle. r Let v be the heading angle. r Let ω be the current linear velocity.r The value represents the current angular velocity. The superscript T indicates the transpose of a matrix or vector.

[0045] S2: Obtain the status information of surrounding obstacles through sensors, including position, speed, direction of movement and category, and construct an obstacle status set.

[0046] Dynamic obstacle status information is obtained through LiDAR, cameras, millimeter-wave radar, or multi-sensor fusion.

[0047] No. The state of a dynamic obstacle can be represented as:

[0048] in, and Indicates the current position coordinates of the obstacle. Indicates the speed of the obstacle. Indicates the direction of the obstacle's movement. This indicates the category of obstacle. The obstacle category includes, but is not limited to, pedestrians, motor vehicles, non-motor vehicles, or temporary static obstacles.

[0049] Based on the obstacle's velocity and direction of motion, the obstacle's velocity vector is obtained:

[0050] The velocity vector of the unmanned sweeper is:

[0051] S3: Calculate TTC and classify the risk.

[0052] For each obstacle, its relative position and relative velocity with the unmanned cleaning vehicle are calculated to obtain the predicted collision time (TTC). Based on the TTC value and the obstacle's occupancy time on the full coverage path, the risk level is divided into five levels: no risk, low risk, medium risk, high risk, and obstruction risk. A risk level is triggered when the obstacle's dwell time exceeds the occupancy time threshold.

[0053] Let the current position of the sweeper be: No. The current positions of the obstacles are: The relative position vector between the sweeper and the obstacle is defined as: The relative velocity vector is defined as: TTC can then be represented as:

[0054] if This indicates that the distance between the two is decreasing, posing a potential risk of collision.

[0055] if This indicates that the distance between the two is increasing or they are no longer getting closer, and it can generally be considered that there is no immediate risk of collision.

[0056] When multiple dynamic obstacles exist, the minimum predicted collision time (TTC) is taken. min As the basis for risk assessment at the current moment, its expression is:

[0057] in, This indicates the number of dynamic obstacles currently detected.

[0058] S4: Generate obstacle avoidance strategies based on risk levels.

[0059] Set TTC threshold , and ,satisfy:

[0060] The risk threshold is determined comprehensively based on the unmanned sweeper's common driving speed, braking response time, maximum braking capacity, sensor detection cycle, controller calculation and execution delay, and reserved safety margin in this embodiment. The threshold selection is an exemplary value in this embodiment. In practical applications, it can be calibrated and adjusted according to the unmanned sweeper's operating speed, braking performance, vehicle size, sensor refresh frequency, road conditions, and obstacle types.

[0061] In this embodiment, the risk threshold can be set as follows: , , When satisfied At that time, it was determined to be risk-free; When satisfied At that time, it was determined to be low risk; When satisfied At that time, it was determined to be a medium-risk area; When satisfied At that time, it was determined to be high risk; When an obstacle continuously occupies the current coverage channel or the area ahead to be cleared, and the occupancy time meets the following conditions... At that time, it was determined to be a risk of congestion. Among them, This indicates the duration for which an obstacle occupies the coverage path. This indicates the congestion risk assessment threshold. The congestion risk assessment threshold is set to 5 to 6 seconds, primarily to distinguish between short-term obstacle crossings by pedestrians and non-motorized vehicles and congestion that continuously occupies the coverage path, preventing the unmanned sweeping vehicle from prematurely initiating A* detours due to short-term obstacle passage. The threshold is determined based on the actual obstacle dwell time in the scenario.

[0062] Based on the risk levels identified above, corresponding obstacle avoidance strategies are generated: When the risk level is no risk, the unmanned sweeper continues to travel along the fully covered path. When the risk level is low, the unmanned sweeper should reduce its speed appropriately while maintaining its original coverage path. When the risk level is medium risk, the unmanned cleaning vehicle reduces its speed and increases its distance constraints on obstacles; When the risk level is high, the unmanned sweeper will stop and wait or perform short-distance DWA (Depth-Way Alternative) local avoidance. When the risk level is congestion risk, the unmanned sweeper initiates A* detour and records the skipped coverage path segments.

[0063] When a blockage risk is identified, it means that the obstacle continues to occupy the original full coverage path. Relying solely on DWA for partial avoidance may result in the unmanned sweeper stopping and waiting for a long time or experiencing local vibrations.

[0064] At this point, the A* bypass strategy is activated, and the missed scan areas caused by the bypass are recorded. For example... Figure 3 As shown, this embodiment is based on the minimum predicted collision time (TTC). min The obstacle risk is categorized into no risk, low risk, medium risk, high risk, and obstruction risk based on the duration of obstacle occupancy on the full coverage path. Corresponding obstacle avoidance strategies include maintaining full coverage path travel, decelerating, continuously reducing speed and increasing obstacle distance constraints, stopping and waiting or partial obstacle avoidance using DWA (Distance-Wait-Avoidance), and obstacle avoidance using A* (Aspect-Driven Detour). Through this mapping of risk levels to obstacle avoidance strategies, the unmanned sweeper achieves a graded response to obstacles with different motion states and varying degrees of obstruction.

[0065] Specifically, the area where the obstacle is located is first expanded:

[0066] in, Indicates the area occupied by the obstacle. Indicates the safe expansion radius. This indicates the expansion operation.

[0067] Then, using the current position of the unmanned sweeper as the starting point of the A* search and the passable waypoint behind the blocked area as the search endpoint, a local A* detour path is generated in the local grid map:

[0068] in, Indicates the first detour path in A*. 1 path point This indicates the number of detour path points.

[0069] The evaluation function of the A* algorithm is:

[0070] in, Represents a node The total cost, Indicates the distance from the starting point to the node. The actual cost, Represents a node Heuristic cost estimation to the target node.

[0071] Preferably, the heuristic function uses Euclidean distance:

[0072] in, Indicates the coordinates of the current node. Indicates the coordinates of the target node.

[0073] During the A* detour of the unmanned sweeping vehicle, the covered track points that are skipped and not yet swept are recorded as missed areas and added to the missed sweep queue.

[0074] S5: Construct improved DWA evaluation function weights based on risk level.

[0075] An improved DWA evaluation function is constructed by introducing a coverage preservation term and a TTC risk penalty term into the traditional DWA evaluation function.

[0076] in, For the directional consistency evaluation item, For obstacle distance evaluation items, For speed evaluation items, For trajectory smoothness evaluation; To cover path preservation items, This is a TTC risk penalty item. Here are the corresponding weights for each item, where All are based on For the dynamic weight function of the independent variable; The weights are dynamically adjusted based on the risk level. When the risk is low, the coverage path maintenance weight is increased; when the risk is high, the obstacle distance weight and TTC risk penalty weight are increased, while the speed weight and coverage path maintenance weight are decreased.

[0077] The coverage path preservation term measures the degree of deviation between the candidate trajectory and the original full coverage path, while the TTC risk penalty term measures the collision risk corresponding to the candidate trajectory.

[0078] The coverage preservation term can be represented as:

[0079] Where, d path Indicates the path from the endpoint of the candidate trajectory to the original full coverage path. The minimum distance. d path The smaller the value, the closer the candidate trajectory is to the original full coverage path, and the higher its coverage retention evaluation.

[0080] TTC risk penalties can be represented as follows:

[0081] in, A very small positive number set to prevent the denominator from being zero.

[0082] S6: Determine if the path deviates from the full coverage path, record the missed areas, and perform regression or rescanning.

[0083] After the unmanned cleaning vehicle executes a partial obstacle avoidance trajectory, it is determined whether it has deviated from the original full-coverage path. Specifically, the minimum distance from the current position of the unmanned cleaning vehicle to the original full-coverage path is calculated:

[0084] in, Indicates the path deviation distance. p indicates the current location of the unmanned sweeper. i Indicates the full coverage path P cov The Middle One coverage point.

[0085] Path deviation distance Deviation from Path Judgment Value Comparison: When satisfied If the unmanned sweeper is determined to have not deviated from the full coverage path or the degree of deviation is acceptable, the judgment result is "no"; the unmanned sweeper continues to perform the sweeping task along the original full coverage path.

[0086] When satisfied If the unmanned cleaning vehicle is found to have deviated from the full coverage path, the judgment result is "yes"; it is necessary to record the uncleaned area and perform regression control.

[0087] The missed scan areas are recorded as the coverage track points that were not covered during the deviation period. The set of missed scan track points is represented as follows:

[0088] Add the missed scan track points to the missed scan queue: in, This represents the set of coverage track points corresponding to the missed scan area. This indicates that the cleaning is complete. This represents the missed scan queue. This indicates the number of areas missed during scanning.

[0089] After recording the missed areas, the nearest unswept coverage track point to the current location of the unmanned sweeper is searched within the original full-coverage path as the regression point. The regression point is defined as:

[0090] in, This represents the set of covered track points that have not yet been cleared.

[0091] After determining the return point, a return path is generated based on the passable area between the current location of the unmanned sweeper and the return point:

[0092] in, Indicates the first step in the regression path At each path point, the unmanned sweeper returns to the original full-coverage path along the regression path.

[0093] When the missed scan queue When the location is not empty, the unmanned sweeper generates a re-sweeping path based on the missed sweeping points:

[0094] in, Indicates the first in the scan path One scan path point.

[0095] After completing the supplementary cleaning, the unmanned cleaning vehicle returns to its original full-coverage path to perform subsequent cleaning tasks.

[0096] S7: When A* detours, it will deviate from the original coverage path, and therefore the skipped path segments will also be added to the missed scan queue. And mark the corresponding areas as to be scanned in the raster map.

[0097] S8: After the unmanned sweeper continues to perform coverage sweeping or completes return sweeping, determine whether the sweeping task of the current work area has been completed.

[0098] If the following conditions are met: Alternatively, if all necessary trackpoints along the full coverage path have been cleaned, the cleaning task is considered complete and the process ends.

[0099] If there are still un-cleaned and covered track points, the cleaning task is determined to be incomplete. The unmanned cleaning vehicle continues to perform the coverage cleaning and returns to step S2 to reacquire obstacle information and continue to perform TTC risk assessment and dynamic obstacle avoidance planning.

[0100] To verify the obstacle avoidance safety, motion stability, and coverage path maintenance capability of the obstacle avoidance method in this embodiment under dynamic obstacle scenarios, dynamic obstacles were set in the test scenario, and the safe distance between the unmanned sweeper and the nearest obstacle, the linear velocity and angular velocity of the unmanned sweeper, and the actual running trajectory of the unmanned sweeper were recorded.

[0101] like Figure 4 The diagram illustrates the changes in the safe distance between the unmanned cleaning vehicle and the nearest obstacle during operation. As the obstacle approaches and the risk level increases, the unmanned cleaning vehicle decelerates, stops, or partially avoids the obstacle based on the corresponding risk level, maintaining a necessary safe distance between the vehicle and the obstacle. Once the obstacle's risk decreases or disappears, the unmanned cleaning vehicle resumes its journey along the full-coverage path.

[0102] like Figure 5 The diagram illustrates the changes in linear velocity and angular velocity of the unmanned sweeper during full-coverage sweeping and dynamic obstacle avoidance. As the risk level increases from low to high, the linear velocity of the unmanned sweeper decreases accordingly. During partial obstacle avoidance, detour, or return control, the angular velocity of the unmanned sweeper is adjusted according to the planned trajectory. This demonstrates that this embodiment can dynamically adjust the motion state of the unmanned sweeper based on the obstacle risk level.

[0103] like Figure 6 As shown, the diagram illustrates the full-coverage path and dynamic obstacle avoidance trajectory of the unmanned cleaning vehicle. When a dynamic obstacle affects the passage of the original full-coverage path, the unmanned cleaning vehicle performs partial avoidance or A* detour, and records the unswept coverage track points skipped due to the detour; after the obstacle risk is reduced or disappears, the unmanned cleaning vehicle returns to the original full-coverage path, or performs supplementary cleaning on the areas corresponding to the missed track points, thereby ensuring the integrity of the full-coverage cleaning task while ensuring obstacle avoidance safety.

[0104] This invention also provides a coverage-maintaining hierarchical dynamic obstacle avoidance system for unmanned sweeping vehicles, comprising: The information acquisition module is used to acquire the full coverage path, current pose and motion status of the unmanned sweeper, as well as the position, speed, direction of motion and category information of surrounding obstacles through sensors; The risk classification module is used to calculate the predicted collision time between the unmanned sweeper and each obstacle based on obstacle information, and classify obstacles into five risk levels: no risk, low risk, medium risk, high risk, and blockage based on the relationship between the predicted collision time and a preset threshold. The strategy generation module is used to generate corresponding obstacle avoidance strategies based on the risk level. The strategies include maintaining the covered path, decelerating, continuously reducing speed, stopping or using the dynamic window method for local obstacle avoidance, and A* detour. The trajectory planning module is used to dynamically adjust the weights of various items in the improved dynamic window method evaluation function with the predicted collision time as the driving parameter. The improved dynamic window method evaluation function includes a coverage path preservation term and a predicted collision time (TTC) risk penalty term. It generates candidate trajectories and outputs local obstacle avoidance trajectories through dynamic window sampling. The missed sweeping compensation module is used to determine whether the current trajectory will cause the unmanned sweeping vehicle to deviate from the full coverage path. When it deviates, it records the uncleaned area and guides the unmanned sweeping vehicle back to the full coverage path or performs supplementary sweeping on the uncleaned area after the obstacle risk is reduced or disappears. The task judgment module is used to determine whether all waypoints in the full coverage path and the missed sweep queue have been cleaned. If so, the task ends; otherwise, the obstacle avoidance and cleaning process continues.

[0105] It should be noted that the above embodiments are only used to illustrate the present invention, but the present invention is not limited to the above embodiments. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention shall fall within the protection scope of the present invention.

Claims

1. A hierarchical dynamic obstacle avoidance method for unmanned sweeping vehicles with coverage maintenance, characterized in that, Includes the following steps: S1. Obtain the full coverage path, current pose and motion status of the unmanned sweeping vehicle; S2. Obtain information on the position, speed, direction of motion, and category of surrounding obstacles through sensors, and construct an obstacle state set; S3. Based on the obstacle information obtained in S2, calculate the predicted collision time between the unmanned cleaning vehicle and each obstacle, and classify the obstacles into five risk levels: no risk, low risk, medium risk, high risk, and blockage, according to the relationship between the predicted collision time and a preset threshold. The risk level is triggered when the duration of the obstacle stay exceeds the occupation time threshold. S4. Based on the risk level determined in S3, generate the corresponding obstacle avoidance strategy: No-risk level: Maintain coverage path; Low-risk level: Reduce speed; Medium-risk level: Continuously reduce speed; High-risk level: Stop or use dynamic window method for local obstacle avoidance; Blockage level: Switch to A* detour mode and simultaneously implement a missed scan recording mechanism. S5. For low-risk, medium-risk, high-risk, and blocking risk levels, the weights of each item in the dynamic window method evaluation function are dynamically adjusted and improved, using the current predicted collision time as the driving parameter. S6. Determine whether the current trajectory will cause the unmanned cleaning vehicle to deviate from the full coverage path. If it does not deviate from the path, continue to run. If it deviates from the full coverage path, record the uncleaned area and guide the unmanned cleaning vehicle back to the full coverage path or perform supplementary cleaning on the uncleaned area after the obstacle risk is reduced or disappears. S7. For risk levels, the A* algorithm is used to plan detour routes, and during the detour, all unswept track points that deviate from the full coverage route are recorded in real time and added to the missed sweep area and merged into the missed sweep queue. S8. Determine whether all track points in the full coverage path and the missed scan queue have been cleaned. If all are cleaned, the task ends; otherwise, return to S2 to continue execution.

2. The coverage-maintaining hierarchical dynamic obstacle avoidance method for unmanned sweeping vehicles according to claim 1, characterized in that: In step S5, the improved dynamic window method evaluation function is: in, This indicates the directional consistency evaluation item. This indicates the obstacle distance evaluation item. Indicates the speed evaluation item. This represents the trajectory smoothness evaluation item; Indicates a path-preserving item with full coverage; The term "time of collision" indicates the risk penalty for predicted collisions; TTC indicates the predicted collision time. This represents the weight of each item, where Both represent dynamic weighting functions with the predicted collision time as the independent variable.

3. The coverage-maintaining hierarchical dynamic obstacle avoidance method for unmanned sweeping vehicles according to claim 2, characterized in that: The expression for the full-coverage path-preserving term is: Where, d path This represents the minimum distance from the endpoint of the candidate trajectory to the original fully covered path; The expression for predicting the collision time risk penalty is: in, This indicates that the value is divided by a small amount to prevent division by zero.

4. The coverage-maintaining hierarchical dynamic obstacle avoidance method for unmanned sweeping vehicles according to claim 1, characterized in that: In step S1, the full coverage path is represented as: in, Indicates the first One coverage point, Indicates the total number of covered waypoints; The current pose state and motion state are represented as follows: Where, x r and y r θ represents the current location of the unmanned cleaning vehicle. r Let v be the heading angle. r Let ω be the current linear velocity. r The value represents the current angular velocity. The superscript T indicates the transpose of a matrix or vector.

5. The coverage-maintaining hierarchical dynamic obstacle avoidance method for unmanned sweeping vehicles according to claim 1, characterized in that: In step S2, the obstacle state set is represented as follows: Where k represents the k-th obstacle, and This indicates the current coordinates of the obstacle. This indicates the speed of the obstacle. Indicates the direction of the obstacle's movement. This indicates the category of obstacle; the obstacle category includes at least one of pedestrians, motor vehicles, non-motor vehicles, and temporary static obstacles.

6. The coverage-maintaining hierarchical dynamic obstacle avoidance method for unmanned sweeping vehicles according to claim 1, characterized in that: In step S3, the function for calculating the predicted collision time is: in, Let be the position vector of the k-th obstacle relative to the unmanned sweeper. Let be the velocity vector of the k-th obstacle relative to the unmanned sweeper.

7. The coverage-maintaining hierarchical dynamic obstacle avoidance method for unmanned sweeping vehicles according to claim 6, characterized in that: when This indicates that the distance between the two is decreasing, posing a potential risk of collision; when This indicates that the distance between the two obstacles is increasing or they are no longer getting closer, and there is no immediate risk of collision. When there are multiple dynamic obstacles, the minimum predicted collision time is taken as the risk assessment criterion for the current moment, and its expression is: in, This indicates the number of dynamic obstacles currently detected.

8. The coverage-maintaining hierarchical dynamic obstacle avoidance method for unmanned sweeping vehicles according to claim 1, characterized in that: In step S6, when the unmanned sweeping vehicle performs a detour and deviates from the original full-coverage path, the skipped coverage path segment is recorded in the missed sweeping queue: in, This represents the set of coverage track points corresponding to the missed scan area; After recording the missed areas, the nearest unswept coverage track point to the current location of the unmanned sweeper is searched within the original full-coverage path as the regression point. The expression for this regression point is: in, p represents the set of cover track points that have not yet been cleared. i Indicates the first in the full coverage path One coverage point, Indicates the current location of the unmanned sweeper; After determining the return point, a return path is generated based on the passable area between the current location of the unmanned sweeper and the return point: in, Indicates the first step in the regression path At each path point, the unmanned sweeper returns to the original full-coverage path along the regression path; When the missed scan queue When the location is not empty, the unmanned sweeper generates a re-sweeping path based on the missed sweeping points: in, Indicates the scan path, Indicates the first in the scan path One scan path point.

9. A coverage-maintaining hierarchical dynamic obstacle avoidance system for unmanned sweeping vehicles, characterized in that, include: The information acquisition module is used to acquire the full coverage path, current pose and motion status of the unmanned sweeper, as well as the position, speed, direction of motion and category information of surrounding obstacles through sensors; The risk classification module is used to calculate the predicted collision time between the unmanned sweeper and each obstacle based on obstacle information, and classify obstacles into five risk levels: no risk, low risk, medium risk, high risk, and blockage based on the relationship between the predicted collision time and a preset threshold. The strategy generation module is used to generate corresponding obstacle avoidance strategies based on the risk level. The strategies include maintaining the covered path, decelerating, continuously reducing speed, stopping or using the dynamic window method for local obstacle avoidance, and A* detour. The trajectory planning module is used to dynamically adjust the weights of various items in the improved dynamic window method evaluation function with the predicted collision time as the driving parameter. The improved dynamic window method evaluation function includes a coverage path preservation term and a predicted collision time risk penalty term. It generates candidate trajectories and outputs local obstacle avoidance trajectories through dynamic window sampling. The missed sweeping compensation module is used to determine whether the current trajectory will cause the unmanned sweeping vehicle to deviate from the full coverage path. When it deviates, it records the uncleaned area and guides the unmanned sweeping vehicle back to the full coverage path or performs supplementary sweeping on the uncleaned area after the obstacle risk is reduced or disappears. The task judgment module is used to determine whether all waypoints in the full coverage path and the missed sweep queue have been cleaned. If so, the task ends; otherwise, the obstacle avoidance and cleaning process continues.

Images