Mobile robot obstacle avoidance method and system for risk assessment of narrow channel

Abstract

The invention belongs to the technical field of robot obstacle avoidance, particularly relates to a mobile robot obstacle avoidance method and system for risk assessment of a narrow channel, and solves the technical problem that an existing multi-robot obstacle avoidance method is easy to cause systematic deadlock in the narrow channel due to lack of operation semantic perception. According to the method, path step-by-step screening based on narrow channel constraint, channel capacity modeling based on cleaning state perception and a conflict pre-judgment and coordination mechanism based on a cleaning window are introduced, and operation semantics such as cleaning continuity and task load are fused into a collaborative decision closed loop, so that deadlock risks are actively eliminated before conflicts occur. According to the system, the method is achieved through a cooperative control platform, and the operation continuity and the overall efficiency of the multiple robots in the complex and narrow environment can be effectively guaranteed. The method is suitable for cooperative operation of multiple environmental sanitation cleaning robots in scenes such as parks and roadways.

Description

Technical Field

[0001] This invention relates to the field of robot obstacle avoidance technology, and in particular to a method and system for mobile robot obstacle avoidance for risk assessment in narrow passages. Background Technology

[0002] With the continuous improvement of urban environmental sanitation intelligence, sanitation cleaning robots are gradually being applied to complex environments such as sidewalks, park roads, underpasses, and back streets to replace or assist manual labor in daily cleaning tasks. In practical applications, multiple sanitation cleaning robots are usually required to perform tasks simultaneously in the same area to improve work efficiency. However, the above-mentioned working environments generally have characteristics such as limited passage width, long passage length, dense dynamic obstacles, and frequent intersections of work routes. Especially when multiple robots approach narrow passages at the same time, situations such as mutual waiting, obstruction of passage, or even deadlock can easily occur, seriously affecting the overall cleaning efficiency and system stability.

[0003] In current robotic sanitation operations, control methods often employ preset cleaning paths combined with simple obstacle avoidance strategies. These methods typically rely on static environment assumptions or a single robot's perspective for path planning and obstacle avoidance decisions, primarily using local perception information to handle obstacle detours in real time. However, in real-world applications, sanitation cleaning environments exhibit significant dynamic changes and often involve multiple mobile robots operating simultaneously. Existing methods generally neglect system-level considerations regarding space resource occupancy, passage timing, and mutual influence among multiple robots.

[0004] Specifically, existing technologies often focus only on the immediate adjustment of a single robot's local behavior, lacking a holistic assessment mechanism for the passage capacity of critical areas such as narrow passages, and failing to model and predict the occupancy of passage resources in real time. When multiple robots approach a narrow passage simultaneously, the lack of a clear order determination and coordination strategy can easily lead to situations where robots wait for each other, repeatedly avoid each other, or even block the passage, resulting in interruptions to the cleaning task, reduced efficiency, and in severe cases, systemic deadlock. For example, in scenarios such as park passages, multiple robots attempting to enter a narrow entrance simultaneously are prone to getting stuck due to disordered resource competition, affecting the overall continuity of operations and system reliability.

[0005] Existing methods for obstacle avoidance or path planning of multiple mobile robots are mostly based on path length, travel time or speed information for traffic scheduling. They do not fully consider the actual operating characteristics of sanitation cleaning operations, such as high requirements for cleaning continuity, high costs of cleaning interruption, and significant differences in cleaning operation status. As a result, it is still difficult to predict and effectively avoid the problem of multiple robots deadlock in a timely manner in narrow passage environments. Summary of the Invention

[0006] This invention proposes a mobile robot obstacle avoidance method and system for narrow passage risk assessment, aiming to solve the technical problem that existing multi-robot obstacle avoidance methods are prone to systemic deadlock in narrow passages due to the lack of operational semantic awareness.

[0007] In a first aspect, the present invention provides a method for obstacle avoidance of a mobile robot for risk assessment in narrow passages, comprising: Based on real-time environmental and robot status data, multiple preliminary paths are planned for each robot; Based on the geometric and obstacle constraints of the narrow passage, the preliminary path is filtered step by step to obtain candidate paths; By combining the cleaning operation status and task load, the passage capacity of narrow passages in the candidate path is modeled with state awareness, and the upper limit of passage flow is generated. Based on the upper limit of the passage flow and the differentiated passage characteristics of the robots, the occupancy time sequence of each robot through the narrow passage is predicted; In response to the predicted conflict in the occupancy time of multiple robots in narrow passages, a coordination sequence is generated and the time window is dynamically adjusted based on the constraints of cleaning operation continuity and task urgency. Based on the aforementioned coordination, the path of the conflicting robots is optimized in a spatiotemporal manner through a sequence and dynamic adjustment of the time window; The robot is controlled in a closed loop based on the optimized path, and its state is continuously updated to make decisions for the next cycle.

[0008] The technical effect of the mobile robot obstacle avoidance method for narrow passage risk assessment disclosed in this invention is that by combining passage risk assessment with the continuity of cleaning operations, differences in cleaning status and operation stage control, multiple sanitation cleaning robots can achieve coordinated passage and dynamic control in narrow passage environments, reduce the probability of deadlock, and improve the overall cleaning operation efficiency and system operation reliability.

[0009] Furthermore, the preliminary path is further refined through a step-by-step screening process, including: Calculate the probability of passage for each grid cell in the path. Paths with an excessively high proportion of grid cells whose probability of passage is below a preset threshold are removed. Due to the narrow channel density of this unit Decide: ; in, This represents the density of narrow channels within the local area where the grid cell is located; This corresponds to the narrow channel density threshold; The sensitivity parameter is used to adjust the degree of influence of narrow passage density changes on the passage probability, and its value ranges from [0.5, 5.0].

[0010] Furthermore, the step-by-step screening also includes: Identify continuous narrow passage segments in the path and calculate path-level structural risk assessment values. and remove Paths that exceed a preset risk threshold, including: ; in, The number of narrow passage segments, To characterize the first The weighting coefficient of the impact of narrow passages on overall traffic risk. For the first Segment length, For the maximum allowed length, The indicator function is defined as follows: (x)= .

[0011] Furthermore, the step-by-step screening also includes: Calculate the average obstacle density along the path and to Paths exceeding a preset density threshold trigger replanning, where: ; in, This represents the total number of grid cells traversed by the path; Indicates the first The obstacle occupancy status in a grid cell, when there is an obstacle in that grid cell. Pick Otherwise take ; This represents the total area of ​​the corresponding grid.

[0012] Furthermore, state-aware modeling of traffic capacity includes: Calculate the robot's mobility adaptability in narrow passages, whereby the mobility adaptability is a function of the robot's size matching the passage width, the robot's current task load influence factor, and the cleaning operation continuity constraint factor; Based on the passage adaptability and channel space constraint index, the upper limit of the number of robots allowed to pass through the narrow channel in the same time period is determined, that is, the upper limit of the channel passage flow.

[0013] Furthermore, the predicted occupancy time series includes: Based on the robot's speed distribution, historical priority records, and the density of dynamic obstacles in the passage, the robot is distinguished between cleaning operation and non-cleaning operation states, and the time occupied in the narrow passage is estimated accordingly. A channel occupancy demand timetable is generated based on the expected arrival time of each robot and the duration of occupancy.

[0014] Furthermore, the generation of coordination through sequences and dynamically adjusted time windows includes: When multiple mobile robots are detected to be in conflict in a narrow passage area, the robots are comprehensively evaluated based on the urgency of their tasks, their current motion status, and the difference in distance from the target location, and their initial passage priority is determined. Based on the cleaning operation stage information of the mobile robot, the initial passage priority is dynamically adjusted, and a dynamic adjustment time window with time priority is allocated to the mobile robot in the critical cleaning stage. Based on the adjusted passage priority and the dynamically adjusted time window, a coordinated passage sequence is generated to guide each conflicting robot through the narrow passage in turn, thereby avoiding blockage or deadlock caused by multiple robots entering the narrow passage at the same time.

[0015] Furthermore, spatiotemporal collaborative optimization of the path includes: Based on the coordination sequence, the robot speed is adjusted to match the dynamic adjustment time window; And / or, plan alternative paths for robots that need to detour, and comprehensively evaluate them by combining their task load impact factors and environmental complexity interference values ​​during the planning process.

[0016] Furthermore, the closed-loop control includes: issuing control commands containing optimized path and speed instructions; The robot's execution deviations and environmental changes are monitored in real time. When the deviation or change exceeds the threshold, a new cycle of decision-making process starting from path planning is triggered to achieve continuous obstacle avoidance.

[0017] In a second aspect, the present invention provides a mobile robot obstacle avoidance system for narrow passage risk assessment, used to execute the method, the system comprising: A robot swarm, comprising multiple sanitation cleaning robots equipped with environmental sensors and actuators; The collaborative control platform, which communicates with the robot cluster, is used for: Receive and fuse environmental and status data from robot clusters to build a dynamic environment model; Execute narrow passage risk assessment, deadlock prediction, and obstacle avoidance decision-making processes; The generated decision path and coordination instructions are sent to the robot cluster for execution. Monitor the execution status and implement closed-loop feedback control.

[0018] The technical effect of the system disclosed in this invention is that by integrating the semantics of cleaning operations into the collaborative scheduling of multiple robots through the collaborative control platform, continuous and efficient cleaning operations without deadlock can be achieved in narrow passage environments. Attached Figure Description

[0019] Figure 1 This is a flowchart illustrating a mobile robot obstacle avoidance method for narrow passage risk assessment proposed in an embodiment of the present invention. Figure 2 A schematic diagram of a three-level path filtering mechanism based on narrow channel constraints provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the coordination decision and path optimization process after detecting a traffic conflict, as provided in an embodiment of the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] The purpose of this invention is to overcome the shortcomings of existing technologies and provide a mobile robot obstacle avoidance method for narrow passage risk assessment. By combining passage risk assessment with the continuity of cleaning operations, differences in cleaning status, and operation stage control, this method enables coordinated passage and dynamic control of multiple sanitation cleaning robots in narrow passage environments, reducing the probability of deadlock and improving overall cleaning efficiency and system reliability. (Reference) Figures 1 to 3 As shown, the specific steps include: S1. Using sensor arrays (such as LiDAR, depth cameras, and odometers) deployed on each sanitation cleaning robot, the robot's GPS or SLAM coordinates, orientation angle, instantaneous linear velocity, and angular velocity are collected in real time. Simultaneously, the sensors perceive the surrounding environment, acquiring high-precision point cloud or image data, from which road boundaries, the positions of static obstacles (such as flower beds and railings), predicted trajectories of dynamic obstacles (pedestrians and vehicles), and terrain details such as the width, length, and curvature of passageways are extracted. Specifically, this includes the following two sub-steps: (1) Constructing a dynamic environment layout map: The central control system or robot aggregates all data through vehicle-to-everything (V2X) communication. Set the channel width limit value W. min (e.g., 1.5 times the robot width) and environmental complexity disturbance value C env(Calculated based on the number of dynamic obstacles and movement speed). The raw sensing data is filtered and fused to remove noise, generating a dynamic environment layout map represented in the form of a grid or topology. Each cell or node in this map contains information such as passage cost, obstacle presence probability, and passage width.

[0022] (2) Generate initial motion state description: Extract the velocity distribution (such as average velocity and velocity variance) of each robot from the data, and calculate the robot body width W. robot The width W of the channel currently in or about to be entered channel Match degree M fit =W channel / W robot The initial motion state description of the robot is formed by packaging {position, orientation, velocity distribution, and size matching}.

[0023] This step establishes a globally consistent framework that includes job semantics (via W). min and C env The environmental and robot state model (reflecting the environment) provides accurate and unified input for subsequent collaborative decision-making, overcoming the limitations of a single robot's local perspective.

[0024] S2. Path Planning and Step-by-Step Filtering Based on Narrow Passages. Based on the initial motion state description, a gridded path search technique is used to calculate multiple preliminary routes for each mobile robot from its current position to the target position, forming a preliminary path set. For each route in the preliminary path set, its corresponding grid cell sequence is extracted. Based on grid cell-level narrow passage constraint analysis, routes that do not meet the passage width limitation are filtered out. Furthermore, combining the passage length measurement data of the narrow passage segment in the path and the dynamic obstacle distribution density within the path's area, structural constraint analysis is performed on the remaining routes to obtain candidate paths that meet the narrow passage structural constraints and environmental passage conditions.

[0025] The central controller plans a path for each robot. The dynamic path search algorithm, combined with a real-time updated dynamic environment layout map, plans N (e.g., 3-5) preliminary routes, starting from the robot's current position and ending at the target work area, forming a preliminary path set P. initial .

[0026] Grid cell level filtering: For P initial Each path in i It iterates through the grid cells it passes through. For cell (i,j), based on the narrow channel density D of that cell in the graph... narrow (i,j) (can be derived from width less than W) min(Calculation of unit proportions) to calculate the probability of passage. If the probability of passage in the path Less than threshold P th The proportion of units exceeds the preset value If a path is deemed to have a hard constraint conflict with the passage capacity of a narrow passage, it is eliminated. This step, known as "hard screening," ensures that the remaining paths are physically feasible.

[0027] Among them, the narrow channel density If the passable width within the local neighborhood centered on this grid cell is less than a preset minimum passable width threshold. The proportion of the number of grid cells to the total number of neighboring grid cells. If the probability of passage in the path... Less than the threshold The proportion of grid cells exceeds the preset value. If a path is deemed to have a hard constraint conflict with the passage capacity of a narrow passage, it is eliminated. This step, known as "hard screening," ensures that the remaining paths are basically feasible under physical width constraints. By introducing a grid-level passage probability assessment based on narrow passage density and combining it with path-level proportional constraints for hard screening, paths with passage conflicts under local passage constraints can be eliminated in advance during the path planning stage, thereby reducing path failures caused by insufficient passage width during subsequent dynamic programming.

[0028] in, Due to the narrow channel density of this unit Decide: ; in, This represents the density of narrow channels within the local area where the grid cell is located; This corresponds to the narrow channel density threshold; A sensitivity parameter is used to adjust the degree of influence of narrow passage density changes on the passage probability, with a value range of [0.5, 5.0]. When <0.5, the model is almost insensitive to narrow channels, losing its screening significance; when When the value is greater than 5.0, the model becomes too sensitive and may cause too many paths to be mistakenly removed.

[0029] Narrow channel density Used to characterize grid cells The distribution degree of narrow passages within a local area is obtained by statistically analyzing the number of narrow passage grid cells in the neighborhood of that grid cell and calculating their proportion. The calculation formula is as follows:

[0030] in: : grid cell The density of the narrow channel is dimensionless; : using grid cells The number of grid cells in the local neighborhood centered on the narrow channel; The total number of grid cells within the local neighborhood; Local neighborhood is based on grid cells Centered on, with side length as The neighboring region; When the passable width between obstacles on both sides of a grid cell is less than the minimum passable width threshold When this happens, the grid cell is determined to be a narrow channel grid cell; The result obtained through the above calculation method The range of values ​​is This is used to reflect the density of narrow passages within a local area, and then to calculate the passage probability of the corresponding grid cell. .

[0031] A further approach is path-level screening: for paths that have passed grid-level screening, continuous narrow passage segments are identified. Path-level structural risk is then calculated. For example, a segment of length is =8m narrow passage, if =10m, =1.0, then the contribution of this item is calculated to be 1.0. If the path contains multiple such channels, Accumulate. Set a risk threshold R. th ,like >R th If a path is deemed high-risk, it can be added to a subset of candidates that require careful evaluation, or it can be eliminated directly if there are sufficient paths available. This step assesses the degree of "structural bottleneck" in narrow passages within the path.

[0032] A further step, environment-level screening: This involves considering dynamic environments. The path is calculated. i The average density of obstacles O density If O density Above the threshold O th This indicates that the current path area is too congested and has low reliability. The system will trigger a local replanning for the robot, attempting to find a smoother alternative path. After three levels of screening, a set P of candidate paths with higher safety and feasibility is obtained. candidate .

[0033] This step transforms the risk of narrow passages from an evaluation indicator in path planning into a pre-constraint condition. Through a three-level progressive screening, inferior paths that are physically infeasible, structurally risky, or overly congested in the early stages of planning are eliminated, laying a solid foundation for subsequent refined collaborative scheduling and significantly improving the safety and robustness of the path planning process.

[0034] S3. Channel Capacity Assessment and Traffic Flow Limit Generation. For the candidate paths, extract the path segments that pass through narrow channels. Combine the channel width limit and channel length measurement data to assess the spatial constraint assessment index of the narrow channels. Introduce the cleaning operation continuity constraint factor, comprehensively consider the cleaning interruption cost caused by channel blockage during the cleaning operation of the mobile robot, and refer to the robot size matching degree and task load impact factor to generate a description of the channel traffic flow limit. Specifically, it includes the following seven steps: Step 1: Extract road segment data containing narrow passages from the candidate path set. Using preset filtering rules, classify and process the corresponding passage width limit value and passage length measurement data for each road segment to obtain the preliminary classification results of narrow passages.

[0035] Step 2: Based on the preliminary division results, calculate the corresponding spatial constraint evaluation index for each narrow passage based on the passage width limit and passage length measurement data, combined with the spatial constraint evaluation criteria. Then, classify each narrow passage according to the spatial constraint evaluation index to determine the spatial constraint level of each road segment.

[0036] Step 3: Based on the spatial constraint level, and combined with the robot size matching degree, analyze the mobility adaptability of different mobile robots when performing cleaning operations in corresponding narrow passages, and obtain the mobility adaptability evaluation data of each robot in each narrow passage.

[0037] Step 4: Based on the aforementioned mobility fitness assessment data, a task load impact factor is further introduced to reflect the impact of cleaning task intensity and cleaning coverage requirements on the robot's mobility. A comprehensive analysis is conducted on the degree of restriction on robot traffic flow in each narrow passage to generate a preliminary description of the passage traffic restriction.

[0038] Step 5: When the preliminary description of the channel flow restriction exceeds the preset threshold, the relevant restriction parameters are corrected based on the road segment data of the corresponding narrow channel, combined with the comprehensive influence of robot size matching degree and task load influence factor, to determine the adjusted passage restriction range.

[0039] Step Six: Based on the adjusted traffic restriction range, and combined with the calculation results of the spatial constraint level and spatial constraint evaluation index, generate a description of the upper limit of traffic flow for each narrow channel, which will serve as the basis for subsequent channel resource allocation and traffic scheduling.

[0040] Step 7: Based on the description of the upper limit of traffic flow, for each narrow passage segment in the path set, combined with the traffic adaptability evaluation results of different mobile robots and the task load, generate a passage scheme suitable for different robots and obtain a complete passage flow planning result.

[0041] This step innovatively quantifies job semantics (continuity, load) as a key factor affecting channel capacity, upgrading channel resource assessment from simple geometric measurement to intelligent modeling closely coupled with tasks. This makes subsequent resource allocation more aligned with actual job requirements and improves allocation rationality.

[0042] S4. Channel Occupancy Duration Calculation and Demand Timing: Starting from the description of the channel's maximum traffic flow, this model integrates mobile robot speed distribution information, historical priority data, and dynamic obstacle distribution density to construct a channel occupancy duration calculation model. The dynamic obstacle distribution density characterizes the distribution characteristics of pedestrians, non-motorized vehicles, or other moving objects within the narrow channel when the dynamic obstacle distribution density exceeds a preset threshold. This serves as a time-weighted factor for channel occupancy duration estimation and is distinguished from the average obstacle density used for path feasibility screening in step S2. The model differentiates between when the mobile robot is in a cleaning operation state and when it is not, performing differentiated estimations of the robot's occupancy duration in the narrow channel and generating a channel occupancy demand time series table accordingly. This specifically includes the following seven steps: Step 1: Obtain the flow restriction data corresponding to each narrow channel from the channel flow limit description, and combine it with the speed distribution information of the mobile robot. Classify the speed data of each robot through a preset speed interval division method to obtain the preliminary grouping results of the speed distribution. Among them, for the robot in the cleaning operation state, its speed grouping parameters are corrected according to the cleaning operation characteristics.

[0043] Step 2: Based on the preliminary grouping results of the speed distribution and combined with historical priority record data, the passage order of each robot is initially adjusted using a preset sorting rule to determine the priority sorted robot group list.

[0044] Step 3: For the priority-sorted group list, obtain the distribution density data of dynamic obstacles in the narrow passage; when the distribution density of dynamic obstacles exceeds a preset threshold, perform passage time weighting processing on the robot group in the corresponding passage to reflect the time impact of dynamic obstacles on robot passage efficiency, and obtain the weighted passage time estimate.

[0045] Step 4: Based on the weighted time estimate and combined with the upper limit description data of the passage flow, distinguish whether the mobile robot is in a cleaning operation state or not. Calculate the estimated time occupied by each robot in the narrow passage to obtain the estimated time occupied data of each robot.

[0046] Step 5: Based on the estimated occupancy time data and the priority-sorted robot group list, the passage time of each robot is divided using a time window allocation method to generate a preliminary channel occupancy demand time sequence arrangement.

[0047] Step Six: Based on the preliminary channel occupancy demand time sequence arrangement and combined with the real-time changes in dynamic obstacle distribution density, when the obstacle density within a certain time period causes a conflict in passage time, the robot passage order within that time period is fine-tuned, the corresponding time window allocation result is corrected, and the final channel occupancy demand time sequence table is determined.

[0048] Step 7: Based on the final channel occupancy demand time series table, for the flow restriction data of each narrow channel, the robot passage arrangement is compared and verified with the channel passage flow limit using a time series matching method to generate a complete occupancy demand time series table data that meets the channel capacity constraints.

[0049] S5. Traffic Conflict Prediction and Active Coordination: When the occupancy demand time series shows that multiple mobile robots approach the same narrow passage within the same time period, the urgency of each mobile robot's task, its current movement state, and the distance difference between it and the target location are extracted to comprehensively evaluate the passage demand of the mobile robots and determine a preliminary passage priority order. Based on this, the passage priority order is dynamically adjusted by combining path conflict probability records and historical priority record data. Furthermore, according to the cleaning operation stage of each mobile robot, the cleaning process is divided into critical and non-critical stages. Dynamically adjusted time windows with time priority are allocated to mobile robots in the critical stage, generating a coordinated passage sequence to guide multiple mobile robots to pass through the narrow passage sequentially. The path conflict probability records are used to characterize the frequency of time window overlap or conflict level detected in the historical time series analysis.

[0050] Step 1: Obtain relevant data from the occupancy demand time sequence table. When multiple mobile robots approach the same narrow passage at the same time, extract the urgency of each mobile robot's task, its current motion state, and the distance difference between each robot and the target location. Perform a comprehensive evaluation of the information to determine the priority of each mobile robot's passage demand.

[0051] Step 2: Based on the passage demand assessment results of each mobile robot obtained in Step 1, determine the initial passage order of each mobile robot according to the preset passage priority determination rules; at the same time, record the path conflict probability data corresponding to the current passage scenario as the initial conflict determination basis for subsequent passage coordination.

[0052] Step 3: Combining the preliminary priority passage order and path conflict probability record, when the conflict probability is detected to exceed a preset threshold, the sorting result is locally adjusted to reduce the risk of potential passage conflicts and generate an optimized coordinated passage sequence.

[0053] Step 4: For the optimized coordinated passage sequence, further retrieve historical priority record data for analysis; when the historical record shows that some robots were delayed multiple times in the previous channel coordination process, fine-tune the robot order in the current coordinated passage sequence to avoid long-term imbalance caused by a fixed order, and determine the final passage arrangement.

[0054] Step 5: Based on the final passage arrangement and the cleaning operation stage information of each mobile robot, the cleaning operation process is divided into critical and non-critical stages. For different stages, the passage time of each robot is allocated in detail using a dynamic adjustment window method, generating dynamic adjustment time window data based on the cleaning window.

[0055] Step Six: Based on the dynamically adjusted time window data, update the timing of the occupancy demand in the narrow passage, correct the passage time information of each mobile robot, and finally generate a complete coordinated passage plan as the basis for subsequent path adjustment and control command issuance.

[0056] S6. Path Adjustment and Global Optimization: Based on the coordinated sequence and dynamic adjustment time window, the routes of affected mobile robots in the initial path set are adjusted. During the path adjustment process, considering the task load impact factor, robots with high-load cleaning tasks are prioritized for path optimization. Furthermore, considering the environmental complexity interference value, the potential passage risks of detour paths are analyzed, thereby generating an optimized path distribution with reduced conflict risk. Specifically, this includes the following six steps: Step 1: Obtain relevant data from the coordination sequence, extract the corresponding travel routes from the preliminary path set based on the passage order of the affected mobile robots and the time allocation information of the dynamic adjustment time window, and determine the path segments that need to be corrected.

[0057] Step 2: Based on the path segments that need to be corrected, and in conjunction with the task load impact factor, sort the current task priorities of each affected mobile robot; when the task load impact factor is higher than a preset threshold, prioritize adjusting the travel routes of robots with high-load cleaning tasks to obtain a preliminary adjusted path scheme.

[0058] Step 3: Based on the initially adjusted route plan, analyze the distribution of obstacles that may be encountered during the detour route planning process, taking into account the environmental complexity interference value; when the obstacle distribution causes the complexity of the detour route to exceed the preset range, perform local optimization on the route plan, and screen and determine feasible detour routes.

[0059] Step 4: For the feasible detour routes, obtain the potential intersection data identified during the path conflict analysis; when the intersection data indicates a high risk of traffic conflict, fine-tune the passage order of the mobile robot to generate a path arrangement with reduced conflict risk.

[0060] Step 5: Based on the path arrangement after the reduction of conflict risk, and in conjunction with the time allocation adjustment rules, the time periods in the dynamic adjustment time window are re-divided, the passage time periods of each mobile robot are refined, and the final time allocation scheme is determined.

[0061] Step Six: Based on the final time allocation scheme, the optimized path layout and coordination are matched through a sequence, and the travel routes and travel time periods of each affected mobile robot are updated synchronously to generate a complete path and time coordination plan, which serves as the basis for subsequent control command issuance and execution monitoring.

[0062] S7. Closed-loop control and continuous conflict handling: Based on the optimized path distribution, coordinated control commands are sent to each mobile robot, and the execution process is monitored in real time by integrating the dynamically adjusted time window and the dynamic cleaning window; by continuously collecting and updating the dynamic environment layout map and motion state description, the occupancy status of the narrow passage is cyclically judged, thereby realizing closed-loop control and continuous obstacle avoidance of multiple mobile robots in the narrow passage environment.

[0063] Specifically, using the optimized path distribution data, coordinated control commands are issued to each mobile robot. Combined with dynamically adjusted time window information, real-time execution data for each robot is obtained to determine its current trajectory. Based on the real-time execution data and environmental layout information, the dynamic environmental layout map is updated. Furthermore, considering the changes in the updated map's regions, it is determined whether there are any potential traffic conflicts related to the current motion state.

[0064] When a potential traffic conflict is identified, the occupancy status of the passage within the conflict area is further analyzed using motion state data to obtain the corresponding occupancy time periods and path information, thus determining the distribution of the potential conflict area. Based on the distribution of the conflict area, and in conjunction with the control rules for dynamically adjusting the time window and the dynamic cleaning window, coordinated control commands are reissued to the affected mobile robots to determine the adjusted operating trajectory and traffic time period allocation.

[0065] After completing the above adjustments, the matching relationship between channel occupancy and path distribution is re-analyzed based on the updated running trajectory and passage time allocation results, as well as the latest dynamic environmental layout map data, to generate an optimized channel allocation scheme. For the optimized channel allocation scheme, potential conflict areas are processed cyclically, and the existence of new passage conflicts is continuously determined based on real-time updates of the movement status.

[0066] When a new potential passage conflict is detected, the channel occupancy demand analysis, passage sequence coordination, and path optimization process are re-triggered. The control commands of each mobile robot are continuously updated, thus forming a closed-loop control process that combines channel occupancy regulation, path adjustment, and execution monitoring. Ultimately, this enables multiple mobile robots to cooperate in passage and continuously execute cleaning operations in narrow passage environments.

[0067] Compared with the prior art, the technical solution provided by the present invention has the following advantages: (1) The path screening mechanism based on the feasibility constraint of narrow passage improves the safety and adaptability of path planning: In the preliminary path planning stage, the present invention introduces the feasibility constraint of narrow passage as a hard constraint for step-by-step screening, rather than as a soft indicator for ranking. This can eliminate paths with structural bottlenecks or high risk of blockage in advance, effectively reducing the probability of passage blockage and operation interruption caused by unreasonable path planning, and improving the path safety and environmental adaptability of multiple robots in complex narrow passage environments.

[0068] (2) Introducing a channel capacity modeling mechanism based on cleaning operation status awareness to improve the rationality of resource allocation in narrow channels: This invention, tailored to the characteristics of sanitation cleaning operations, introduces cleaning operation continuity constraints, robot operation status differences, and task load factors into the channel capacity assessment, making the description of the upper limit of channel traffic flow more in line with actual operational needs. This avoids the simplistic treatment of treating narrow channels as fixed capacity, and can dynamically reflect the channel's adaptability to the passage of robots in different states, thereby achieving a rational allocation of channel resources while ensuring cleaning continuity, reducing unnecessary congestion and detours, and improving the overall operating efficiency of the system.

[0069] (3) Enhanced system stability in dynamic environments through a passage coordination and deadlock prediction mechanism based on cleaning window constraints: This invention introduces the concept of a "cleaning window" during the passage occupancy demand analysis phase and actively coordinates and sorts tasks based on task urgency, operation stage, and historical records. This enables early prediction and proactive resolution of potential deadlocks that may occur when multiple robots simultaneously approach a narrow passage. This is fundamentally different from the passive response of existing technologies after a conflict occurs, realizing a shift from "conflict resolution" to "conflict avoidance," thereby significantly improving the operational stability and reliability of multi-robot systems in complex, dynamic, and narrow passage environments.

[0070] Based on the same inventive concept, embodiments of the present invention also provide a mobile robot obstacle avoidance system for narrow passage risk assessment, used to execute the method described above, the system comprising: A robot swarm, comprising multiple sanitation cleaning robots equipped with environmental sensors and actuators; The collaborative control platform, which communicates with the robot cluster, is used for: Receive and fuse environmental and status data from robot clusters to build a dynamic environment model; Execute narrow passage risk assessment, deadlock prediction, and obstacle avoidance decision-making processes; The generated decision path and coordination instructions are sent to the robot cluster for execution. Monitor the execution status and implement closed-loop feedback control.

[0071] Example embodiments have been disclosed herein, and while specific terminology has been used, it is for illustrative purposes only and should be construed as such, and is not intended to be limiting. In some instances, it will be apparent to those skilled in the art that features, characteristics, and / or elements described in conjunction with particular embodiments may be used alone, or in combination with features, characteristics, and / or elements described in conjunction with other embodiments, unless otherwise expressly indicated. Therefore, those skilled in the art will understand that various changes in form and detail may be made without departing from the scope of the invention as set forth in the appended claims.

Claims

1. A method for obstacle avoidance in mobile robots based on risk assessment in narrow passages, characterized in that, include: Based on real-time environmental and robot status data, multiple preliminary paths are planned for each robot; Based on the geometric and obstacle constraints of the narrow passage, the preliminary path is filtered step by step to obtain candidate paths; By combining the cleaning operation status and task load, the passage capacity of narrow passages in the candidate path is modeled with state awareness, and the upper limit of passage flow is generated. Based on the upper limit of the passage flow and the differentiated passage characteristics of the robots, the occupancy time sequence of each robot through the narrow passage is predicted; In response to the predicted conflict in the occupancy time of multiple robots in narrow passages, a coordination sequence is generated and the time window is dynamically adjusted based on the constraints of cleaning operation continuity and task urgency. Based on the aforementioned coordination, the path of the conflicting robots is optimized in a spatiotemporal manner through a sequence and dynamic adjustment of the time window; The robot is controlled in a closed loop based on the optimized path, and its state is continuously updated to make decisions for the next cycle.

2. The method according to claim 1, characterized in that, The preliminary path is filtered step by step, including: Calculate the probability of passage for each grid cell in the path. Paths with an excessively high proportion of grid cells whose probability of passage is below a preset threshold are removed. Due to the narrow channel density of this unit Decide: ； in, This represents the density of narrow channels within the local area where the grid cell is located; This corresponds to the narrow channel density threshold; The sensitivity parameter is used to adjust the degree of influence of narrow passage density changes on the passage probability, and its value ranges from [0.5, 5.0]. Narrow channel density Used to characterize grid cells The distribution degree of narrow passages within a local area is obtained by statistically analyzing the number of narrow passage grid cells in the neighborhood of that grid cell and calculating their proportion. The calculation formula is as follows: ； in: : grid cell The density of the narrow channel is dimensionless; : using grid cells The number of grid cells in the local neighborhood centered on the narrow channel; The total number of grid cells within the local neighborhood; The local neighborhood is defined by grid cells. Centered on, with side length as The neighboring region; When the passable width between obstacles on both sides of a grid cell is less than the minimum passable width threshold If the condition is met, the grid cell is identified as a narrow channel grid cell.

3. The method according to claim 2, characterized in that, The step-by-step screening also includes: Identify continuous narrow passage segments in the path and calculate path-level structural risk assessment values. and remove Paths exceeding a preset risk threshold, including: ； in, The number of narrow passage segments, To characterize the first The weighting coefficient of the impact of narrow passages on overall traffic risk. For the first Segment length, For the maximum allowed length, The indicator function is defined as follows: (x)= 。 4. The method according to claim 3, characterized in that, The step-by-step screening also includes: Calculate the average obstacle density along the path and to Paths exceeding a preset density threshold trigger replanning, where: ； in, This represents the total number of grid cells traversed by the path; Indicates the first The obstacle occupancy status in a grid cell, when there is an obstacle in that grid cell. Pick Otherwise take ; This represents the total area of ​​the corresponding grid.

5. The method according to claim 1, characterized in that, State-aware modeling of traffic capacity includes: Calculate the robot's mobility adaptability in narrow passages, whereby the mobility adaptability is a function of the robot's size matching the passage width, the robot's current task load influence factor, and the cleaning operation continuity constraint factor; Based on the passage adaptability and channel space constraint index, the upper limit of the number of robots allowed to pass through the narrow channel in the same time period is determined, that is, the upper limit of the channel passage flow.

6. The method according to claim 1, characterized in that, The occupancy timing includes: Based on the robot's speed distribution, historical priority records, and the density of dynamic obstacles in the passage, the robot is distinguished between cleaning operation and non-cleaning operation states, and the time occupied in the narrow passage is estimated accordingly. A channel occupancy demand timetable is generated based on the expected arrival time of each robot and the duration of occupancy.

7. The method according to claim 1, characterized in that, The generation coordination includes the following through a sequence and dynamic adjustment of the time window: When multiple mobile robots conflict in a narrow passage area, the conflicting robots are comprehensively evaluated based on the urgency of their tasks, their current motion status, and the difference in distance from the target location to determine the initial passage priority. Based on the cleaning operation stage information of the mobile robot, the initial passage priority is dynamically adjusted, and a dynamic adjustment time window with time priority is allocated to the mobile robot in the critical cleaning stage. Based on the adjusted passage priority and the dynamically adjusted time window, a coordinated passage sequence is generated to guide each conflicting robot through the narrow passage in turn, so as to avoid the narrow passage from being blocked or deadlocked.

8. The method according to claim 1, characterized in that, Spatiotemporal co-optimization of paths includes: Based on the coordination sequence, the robot speed is adjusted to match the dynamic adjustment time window; And / or, plan alternative paths for robots that need to detour, and comprehensively evaluate them by combining their task load impact factors and environmental complexity interference values ​​during the planning process.

9. The method according to claim 1, characterized in that, The closed-loop control includes: issuing control commands containing optimized path and speed instructions; The robot's execution deviations and environmental changes are monitored in real time. When the deviation or change exceeds the threshold, a new cycle of decision-making process starting from path planning is triggered to achieve continuous obstacle avoidance.

10. A mobile robot obstacle avoidance system for narrow passage risk assessment, used to perform the method as described in any one of claims 1-9, characterized in that, The system includes: A robot swarm, comprising multiple sanitation cleaning robots equipped with environmental sensors and actuators; The collaborative control platform, which communicates with the robot cluster, is used for: Receive and fuse environmental and status data from robot clusters to build a dynamic environment model; Execute narrow passage risk assessment, deadlock prediction, and obstacle avoidance decision-making processes; The generated decision path and coordination instructions are sent to the robot cluster for execution. Monitor the execution status and implement closed-loop feedback control.

Images