System and method for monitoring working data of AGV robot

Abstract

The invention discloses a monitoring system and method for working data of an AGV robot, and belongs to the technical field of artificial intelligence. The method comprises the following steps: preprocessing and integrating the driving track, freight weight, running state and task execution data of the AGV robot to form a robot working data set; clustering analysis is carried out based on driving tracks of all robots, and similar tracks are clustered into one class; constructing a dynamic driving track planning model by using an A * algorithm, and generating a plurality of candidate optimal driving tracks for each task combination; constructing a task demand list, and carding a task sequence logic relationship; performing task allocation on the AGV robot by using a greedy algorithm; generating a task allocation list by using a simulated annealing mechanism; a task instruction is issued to a corresponding AGV robot control system, a task execution process is started, and whole-course tracking is started; and after the AGV robot completes a single task and a round of task sequence, a model self-learning updating mechanism is triggered.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a monitoring system and method for working data of an AGV robot. Background Art

[0002] In the process of the vigorous development of modern logistics distribution centers, automated guided vehicles (AGVs) have become a key automated transportation force, aiming to improve cargo transportation efficiency, reduce labor costs and enhance operational accuracy. However, traditional technical solutions around AGVs have many difficult shortcomings, which restrict the deep release of logistics efficiency.

[0003] In terms of positioning and track recording, some early AGVs rely on magnetic stripe navigation or QR code navigation. Magnetic stripe navigation requires the pre-laying of magnetic strips on site. Not only is the initial installation cumbersome and the subsequent layout change costly, but the magnetic strips are easily disturbed by iron filings, dust, etc. in the environment, resulting in impaired positioning accuracy. In the existing technology, task priority is often simply judged by manual experience. Faced with massive, complex orders with varying degrees of urgency, it is very easy to misjudge or miss, resulting in a backlog of urgent goods and delays; the allocation process does not comprehensively consider the real-time status and environmental factors of the AGV, but only mechanically allocates them in a fixed order or on the principle of proximity, resulting in the coexistence of idle resources and task congestion, uneven loads on different AGVs, and time-consuming and inefficient overall task execution, and collision accidents occur from time to time. Summary of the invention

[0004] The object of the present invention is to provide a monitoring system and method for AGV robot working data to solve the problems raised in the prior art.

[0005] To achieve the above object, the present invention provides the following technical solutions: A method for monitoring working data of an AGV robot, the method comprising the following steps: S100, using laser navigation to obtain the real-time position of the AGV robot in the logistics distribution center, record its driving trajectory, monitor the weight of each shipment through the pressure sensor installed on the AGV cargo platform, monitor the operating status of the AGV robot, and record the task execution data of the AGV robot; pre-process and integrate the driving trajectory, cargo weight, operating status and task execution data of the AGV robot to form a robot work data set; S200, based on the robot work data set, cluster analysis is used based on the driving trajectories of all robots to cluster similar trajectories into one category, and a cargo handling trajectory requirement set including a number of start-end points is obtained; based on the cargo handling trajectory requirement set, a dynamic driving trajectory planning model is constructed using the A* algorithm, and a number of candidate optimal driving trajectories are generated for each task combination by integrating the driving trajectory distance, real-time road conditions and avoidance rules; S300. Classify the newly added cargo handling tasks of the logistics distribution center in the current period, set the task priority according to the urgency of cargo delivery and inventory storage requirements, build a task requirement list, and sort out the logical relationship of the tasks; use a greedy algorithm to preferentially assign high-priority tasks to AGV robots that are close, in normal operation, with a suitable rated cargo weight, and without congestion in their driving trajectories; use a simulated annealing mechanism to evaluate the comprehensive task execution efficiency, robot load balancing, and collision risks of different candidate optimal driving trajectories, and generate a task assignment list; S400. Based on the task allocation list, the task instructions are issued to the corresponding AGV robot control system to start the task execution process. At the same time, the task start status is marked on the monitoring end to start full-process tracking. When the AGV robot completes a single task and a round of task sequence, its actual task execution data is collected and compared with the predicted data in the task allocation stage to trigger the model self-learning update mechanism.

[0006] According to step S100, the AGV robot is equipped with a laser scanner on the top. The laser scanner emits a laser beam. Using the triangulation method, the robot can calculate its own coordinates in the two-dimensional plane coordinate system of the logistics distribution center in real time and send them to the monitoring server. Whenever the monitoring server receives new coordinates, it stores them in the local cache in the order of timestamps. The cache is rolled and stored according to the first-in-first-out principle to obtain the driving trajectory of the AGV robot. The pressure sensor installed on the AGV cargo platform monitors the weight of each shipment. Monitor the running status of the AGV robot, monitor the voltage, charge and discharge current and battery temperature of each battery in real time, and transmit the data to the AGV main controller through the CAN bus; install Hall sensors and temperature sensors on the motor part, the Hall sensors monitor the motor speed, and the temperature sensors monitor the motor winding temperature; use the vibration acceleration sensor on the moving part of the AGV robot to capture transmission vibration anomalies based on strain, vibration frequency and amplitude characteristics; obtain the task execution data of the AGV robot from the built-in task management module of the AGV robot; The driving trajectory, cargo weight, operating status and task execution data of the AGV robot are preprocessed and integrated to form a robot work data set.

[0007] According to step S200, for each driving trajectory, a set of cargo handling features is extracted, and the cargo handling features include geometric features, statistical features, and spatiotemporal features; wherein, the coordinates of the starting point and the end point of the driving trajectory are calculated, and the difference between the coordinates of the starting point and the end point is used as a vector component as a geometric feature; the total length, the number of stop points, and the average speed of the driving trajectory are counted as statistical features; the time period of the start time and the end time of the trajectory is obtained as a spatiotemporal feature; the cargo handling features are integrated to convert each driving trajectory into a multidimensional vector in the form of: [starting horizontal coordinate difference, starting vertical coordinate difference, trajectory length, number of stop points, average speed, starting time period code, ending time period code], and all multidimensional vectors constitute a cluster analysis input matrix; A density-based spatial clustering algorithm combined with a hierarchical clustering algorithm is used to cluster similar trajectories in the cluster analysis input matrix into one category; the spatial clustering algorithm traverses the trajectory vector set according to the set neighborhood radius and the minimum number of points, identifies the core points and boundary points, clusters the connected points into initial clusters, and divides the initial trajectory categories; hierarchical clustering is introduced to merge and split the preliminary clusters, and a clustering dendrogram is constructed from the bottom up using the Euclidean distance as an indicator. The clustering quality is evaluated based on the silhouette coefficient, the optimal number of clusters is determined, and similar trajectories are clustered into one category; similar trajectories are traversed, the common starting point and end point coordinate ranges are extracted, and summarized into several start-end points pairs to form a cargo handling trajectory demand set.

[0008] According to step S200, the environmental map model of the logistics distribution center is obtained and gridded, and each grid unit is used as the basic node for the A* algorithm search; based on the driving trajectory of the AGV robot and the layout information of the logistics distribution center, the attributes of each grid are annotated; based on the cargo handling trajectory requirement set, the coordinate positions of each start-end point pair are clearly corresponding to the map grid nodes, and the start and end anchor points are anchored for the driving trajectory; for real-time road condition information, according to the frequency and time distribution analysis of historical driving trajectory data, the congestion probability and average travel time of each section at different time periods are statistically analyzed to simulate the actual traffic congestion level; based on the multi-machine collaborative operation records in the robot's task execution data, the robot's avoidance rules at intersections and parallel sections are sorted out, quantified as waiting time or driving trajectory adjustment costs under different scenarios, and integrated into the model rule set; The A* algorithm is based on the heuristic function f(n) = g(n) + h(n), where g(n) represents the actual cost from the starting node to the current node n, which is quantified by the combined driving trajectory distance and the real-time road condition additional cost; h(n) is used as a heuristic function, which uses Manhattan distance to estimate the expected cost from the current node to the end point, guiding the search direction toward the end point; During the search process, each time a node is expanded, the model rule set is verified. When there are running trajectories of other robots around the expanded node, the current driving trajectory expansion branch is paused and the alternative branch driving trajectory is explored first. The A* algorithm is used to perform several search iterations to generate several candidate optimal driving trajectories for each task combination.

[0009] According to step S300, from the perspective of cargo type, cargo handling tasks are divided into fresh and perishable, high-value precision electronics, ordinary daily necessities, and large industrial parts; from the perspective of cargo source, cargo handling tasks are divided into supplier replenishment tasks, online e-commerce order delivery tasks, and offline store transfer tasks; the urgency of cargo delivery is calculated based on the remaining time from the promised delivery time, and the shorter the remaining time, the higher the priority; When constructing a task requirement list, it is presented in a structured table format, with each row recording the information of a single handling task; task-related constraints are reflected in the cargo matching and process sequence, sorting out the logical relationship between tasks.

[0010] According to step S300, task priorities are set according to the urgency of goods delivery and the quantified inventory storage requirements. , the value range is set to [1, N], where N is the number of priority levels divided according to actual operations, and is dynamically adjusted through the real-time inventory system and order delivery time monitoring data; based on the coordinate information obtained by the laser navigation system, the Euclidean distance formula is used to calculate that in the two-dimensional plane of the distribution center, if the current coordinate of the AGV robot j is , the starting coordinates of task i are , then the distance between the AGV robot and the starting point of the task is , in meters, distance data is updated in real time; comprehensive multi-dimensional operation status monitoring data is quantified to set the AGV operation status score ,in For electrical system health, Is the mechanical structure integrity, Indicates the reliability of the braking and steering systems. is the weight of the electrical system health, is the weight of the mechanical structure integrity, is the weight of the braking and steering system reliability. When the AGV is in full operation state ; Estimated cargo weight for task i AGV robot rated load weight , then the rated freight weight adaptability , with a value between [0,1]; based on the historical driving trajectory big data analysis combined with real-time road condition monitoring, the congestion probability and duration of each section on the driving trajectory of AGV robot j to the starting point of task i are counted, and the congestion index is set , where m is the number of driving track sections, is the congestion probability of road section k, Estimate additional time for congestion; Initialize Task Set , and AGV robot , sort the tasks in descending order by priority, where n is the total number of new tasks and m is the number of robots that can be deployed; For each task in the task set , traverse in priority order and calculate each AGV robot For tasks Comprehensive adaptation index , the formula is: ; in is the weight coefficient of task priority, is the weight coefficient of the distance between the AGV robot and the task starting point, is the weight coefficient of AGV operation status score, is the weight coefficient of the rated freight weight adaptability, is the weight coefficient of the congestion index; Selection The most valuable AGV robot Assigning tasks , update the robot's task execution data and task set.

[0011] According to step S300, a set of initial task allocation plans are randomly generated, each plan covers the allocation of various types of cargo handling tasks to different AGV robots, and specifies the driving trajectory; based on the task execution data of the AGV robots, the average time spent on different types of tasks by each robot in the past is counted, and the execution time of a single task is estimated in combination with the characteristics of the current task; when evaluating the allocation plan, the estimated time of all tasks is accumulated, and the time spent on the connection between tasks is considered to obtain the total task execution time of the overall plan; the shorter the plan time, the higher the score in the execution efficiency dimension; In the task allocation scheme evaluation, the cumulative load weight of the tasks assigned to each robot is analyzed to measure whether the load distribution is balanced; the smaller the standard deviation, the more balanced the load of each robot; for the driving trajectory selected by each robot in each allocation scheme, the probability of encounter within the same road section and time window is analyzed, and historical collision events and congestion point data are reviewed. Combined with real-time location tracking, points are deducted for high-risk driving trajectory combinations; The entire operating space of the logistics distribution center is divided into fine two-dimensional grid units based on the actual site layout, channel direction, shelf distribution and other factors. Each unit represents a certain actual area. At the same time, it is divided into discrete time segments along the time axis at a certain time interval to construct a space-time grid system. In this way, the driving trajectory of the AGV robot can be accurately located and described in this space-time grid framework, and each trajectory point corresponds to a specific space grid and time segment. For each driving trajectory selected by each AGV robot under each task allocation scheme, it is mapped point by point to the space-time grid in chronological order. The algorithm scans the frequency of occurrence of trajectory points in the same space grid unit in the same time segment. Traverse all space-time grid slices corresponding to the entire task execution time period, count the total number of encounters of different robot combinations in each section and each time window, and divide it by the theoretically possible maximum number of encounters to obtain the encounter probability value. The higher the probability, the greater the potential risk of collision.

[0012] Construct a comprehensive deduction index system, including encounter probability deduction, historical high-risk area deduction, and real-time warning deduction, and assign weights to each index based on actual operational focus. For each robot driving trajectory combination in each task allocation plan, the deductions are calculated one by one according to the set indicators and weights, and the total deductions for the combination are accumulated. The higher the total deduction, the greater the risk and lower the quality of the driving trajectory combination. When evaluating the entire task allocation plan, trajectory combinations with high deductions will lower the overall score of the plan, prompting the system to eliminate or adjust such plans during optimization processes such as simulated annealing, screen out safer and more efficient task allocation strategies, and ensure smooth and low-risk operation of logistics distribution.

[0013] After several rounds of simulated annealing, until the number of iterations reaches the target, the optimal solution is sorted and converted into a task assignment list.

[0014] According to step S400, task instructions are customized for each AGV robot involved based on the task allocation list. The task instructions are presented in the form of digital data packets, clearly specifying the starting coordinates and the end coordinates of the task; the weight threshold that the robot needs to carry is set in the instruction in combination with the average weight of similar goods in the past counted by the pressure sensor of the cargo platform and the estimated weight of the current order; at the monitoring end, when the task instruction is successfully issued to the AGV robot, the supporting monitoring software system synchronously triggers the task start state marking action, and on the distribution center map interface presented in two-dimensional form, the icon representing the corresponding AGV robot will change, and a pop-up window will be displayed next to the icon to display the task information.

[0015] According to step S400, in the time dimension, the start time and end time of the task are recorded to obtain the actual time taken by the task; the actual driving trajectory of the robot is obtained, and the operating status data is synchronously summarized; in the time dimension comparison, the estimated task execution time is compared with the actual time taken, and the path data is compared to verify the predicted optimal driving trajectory and the actual driving trajectory, and analyze the reasons for the deviation; in terms of operating status comparison, the estimated battery power consumption and motor temperature range are obtained. Differences with actual data; re-execute steps S100-S300 to trigger the model self-learning update mechanism.

[0016] A monitoring system for AGV robot working data, comprising: Data collection and integration module: including: task execution data collection unit and robot work data set integration unit; the task execution data collection unit uses laser navigation to obtain the real-time position of the AGV robot in the logistics distribution center, records its driving trajectory, monitors the weight of each shipment through the pressure sensor installed on the AGV cargo platform, monitors the operating status of the AGV robot, and records the task execution data of the AGV robot; the robot work data set integration unit pre-processes and integrates the driving trajectory, cargo weight, operating status and task execution data of the AGV robot to form a robot work data set; Trajectory analysis and path planning module: including: trajectory clustering analysis unit and dynamic path planning unit; among them, the trajectory clustering analysis unit is based on the robot work data set, and uses clustering analysis based on the driving trajectories of all robots to cluster similar trajectories into one category, and obtains a cargo handling trajectory requirement set containing several start-end pairs; the dynamic path planning unit is based on the cargo handling trajectory requirement set, and uses the A* algorithm to build a dynamic driving trajectory planning model, which integrates the driving trajectory distance, real-time road conditions and avoidance rules to generate several candidate optimal driving trajectories for each task combination; Task scheduling and allocation module: including: task classification and priority setting unit, greedy algorithm allocation unit and simulated annealing optimization unit; among them, the task classification and priority setting unit classifies the newly added cargo handling tasks of the logistics distribution center in the current period, sets the task priority according to the urgency of cargo delivery and inventory storage requirements, builds a task requirement list, and sorts out the logical relationship of task sequence; the greedy algorithm allocation unit uses a greedy algorithm to preferentially allocate high-priority tasks to AGV robots that are close, in normal operation, with suitable rated cargo weight and no congestion in the driving trajectory; the simulated annealing optimization unit uses a simulated annealing mechanism to evaluate the comprehensive task execution efficiency, robot load balance and collision risk of different candidate optimal driving trajectories, and generate a task allocation list; Instruction issuing and monitoring feedback module: including: instruction issuing unit, monitoring and tracking unit and data comparison and model updating unit; among them, the instruction issuing unit issues task instructions to the corresponding AGV robot control system based on the task allocation list, starts the task execution process, and the monitoring and tracking unit marks the task start status at the monitoring end and starts full-process tracking; the data comparison and model updating unit collects the actual task execution data after the AGV robot completes a single task and a round of task sequence, compares and analyzes it with the predicted data in the task allocation stage, and triggers the model self-learning update mechanism.

[0017] Compared with the prior art, the present invention has the following beneficial effects: 1. The present invention uses laser navigation, with the help of precise interaction between laser beam and reflective plate and advanced algorithms, to achieve millimeter-level high-precision positioning of AGV in the distribution center, and record the driving trajectory in real time and stably without fear of interference from environmental dust and tiny obstacles. The precise trajectory data lays a solid foundation for subsequent path optimization and operation tracing, and the entire transportation process can be reviewed at any time, and the route can be dynamically adjusted to adapt to complex and changeable working conditions.

[0018] 2. In the dimension of operating status monitoring, the present invention integrates electrical and mechanical multi-level data, relies on big data analysis and intelligent algorithms to predict failure risks in advance, and provides early warning intervention when the battery power drops slightly or mechanical parts begin to wear out, thus changing post-maintenance to pre-prevention.

[0019] 3. In the closed loop of task allocation and execution, the present invention scientifically divides task priorities based on rigorous data analysis and integrates inventory and order timeliness; uses clustering and A* algorithms to explore transportation rules and plan high-quality paths, and then combines greedy simulated annealing algorithms to fully weigh AGV real-time performance, location, and path conditions to achieve intelligent task allocation. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 It is a schematic diagram of the steps of a method for monitoring working data of an AGV robot according to the present invention; Figure 2 It is a system structure diagram of a monitoring system for working data of an AGV robot according to the present invention. DETAILED DESCRIPTION

[0021] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0022] Example: Figure 1-Figure 2As shown, the present invention provides a technical solution. According to one embodiment of the present invention, Figure 1 A method for monitoring working data of an AGV robot is shown in a schematic diagram of steps. A method for monitoring working data of an AGV robot comprises the following steps: S100, using laser navigation to obtain the real-time position of the AGV robot in the logistics distribution center, record its driving trajectory, monitor the weight of each shipment through the pressure sensor installed on the AGV cargo platform, monitor the operating status of the AGV robot, and record the task execution data of the AGV robot; pre-process and integrate the driving trajectory, cargo weight, operating status and task execution data of the AGV robot to form a robot work data set; S200, based on the robot work data set, cluster analysis is used based on the driving trajectories of all robots to cluster similar trajectories into one category, and a cargo handling trajectory requirement set including a number of start-end points is obtained; based on the cargo handling trajectory requirement set, a dynamic driving trajectory planning model is constructed using the A* algorithm, and a number of candidate optimal driving trajectories are generated for each task combination by integrating the driving trajectory distance, real-time road conditions and avoidance rules; S300. Classify the newly added cargo handling tasks of the logistics distribution center in the current period, set the task priority according to the urgency of cargo delivery and inventory storage requirements, build a task requirement list, and sort out the logical relationship of the tasks; use a greedy algorithm to preferentially assign high-priority tasks to AGV robots that are close, in normal operation, with a suitable rated cargo weight, and without congestion in their driving trajectories; use a simulated annealing mechanism to evaluate the comprehensive task execution efficiency, robot load balancing, and collision risks of different candidate optimal driving trajectories, and generate a task assignment list; S400. Based on the task allocation list, the task instructions are issued to the corresponding AGV robot control system to start the task execution process. At the same time, the task start status is marked on the monitoring end to start full-process tracking. When the AGV robot completes a single task and a round of task sequence, its actual task execution data is collected and compared with the predicted data in the task allocation stage to trigger the model self-learning update mechanism.

[0023] According to step S100, the AGV robot is equipped with a laser scanner on the top. The laser scanner emits a laser beam. Using the triangulation method, the robot can calculate its own coordinates in the two-dimensional plane coordinate system of the logistics distribution center in real time and send them to the monitoring server. Whenever the monitoring server receives new coordinates, it stores them in the local cache in the order of timestamps. The cache is rolled and stored according to the first-in-first-out principle to obtain the driving trajectory of the AGV robot. The pressure sensor installed on the AGV cargo platform monitors the weight of each shipment. Monitor the running status of the AGV robot, monitor the voltage, charge and discharge current and battery temperature of each battery in real time, and transmit the data to the AGV main controller through the CAN bus; install Hall sensors and temperature sensors on the motor part, the Hall sensors monitor the motor speed, and the temperature sensors monitor the motor winding temperature; use the vibration acceleration sensor on the moving part of the AGV robot to capture transmission vibration anomalies based on strain, vibration frequency and amplitude characteristics; obtain the task execution data of the AGV robot from the built-in task management module of the AGV robot; The driving trajectory, cargo weight, operating status and task execution data of the AGV robot are preprocessed and integrated to form a robot work data set.

[0024] According to step S200, for each driving trajectory, a set of cargo handling features is extracted, and the cargo handling features include geometric features, statistical features, and spatiotemporal features; wherein, the coordinates of the starting point and the end point of the driving trajectory are calculated, and the difference between the coordinates of the starting point and the end point is used as a vector component as a geometric feature; the total length, the number of stop points, and the average speed of the driving trajectory are counted as statistical features; the time period of the start time and the end time of the trajectory is obtained as a spatiotemporal feature; the cargo handling features are integrated to convert each driving trajectory into a multidimensional vector in the form of: [starting horizontal coordinate difference, starting vertical coordinate difference, trajectory length, number of stop points, average speed, starting time period code, ending time period code], and all multidimensional vectors constitute a cluster analysis input matrix; A density-based spatial clustering algorithm combined with a hierarchical clustering algorithm is used to cluster similar trajectories in the cluster analysis input matrix into one category; the spatial clustering algorithm traverses the trajectory vector set according to the set neighborhood radius and the minimum number of points, identifies the core points and boundary points, clusters the connected points into initial clusters, and divides the initial trajectory categories; hierarchical clustering is introduced to merge and split the preliminary clusters, and a clustering dendrogram is constructed from the bottom up using the Euclidean distance as an indicator. The clustering quality is evaluated based on the silhouette coefficient, the optimal number of clusters is determined, and similar trajectories are clustered into one category; similar trajectories are traversed, the common starting point and end point coordinate ranges are extracted, and summarized into several start-end points pairs to form a cargo handling trajectory demand set.

[0025] In this embodiment, the logistics distribution center is a rectangular area, 100 meters long and 80 meters wide, divided into 20×20 grid units, each grid side length is 5 meters, the simulation covers functional areas such as cargo storage area, sorting area, and delivery platform, and obstacles such as shelves and pillars are scattered in the area, which are reflected by grid attribute annotation. The layout information shows that the cargo storage area is concentrated on the left side of the site (horizontally 0-30 meters range), the delivery platform is on the right side (horizontally 70-100 meters range), the middle area is the main channel and sorting operation space, and some shelves are distributed in a determinant, affecting the path passage. Five AGV robots are put into operation, numbered from R1 to R5, and their rated loads are [200kg, 250kg, 300kg, 200kg, 250kg] respectively. The initial positions are randomly distributed at different grid nodes in the site. A certain amount of driving trajectory data has been accumulated through the early laser navigation for analyzing road conditions and annotating grid attributes. Set 10 cargo handling task combinations (T1 - T10). The cargo weight is estimated to fluctuate in the range of [50kg - 280kg] based on historical order data. The start-end point pairs of each task are distributed from different shelves in the storage area to different warehouses at the corresponding delivery platform, so as to construct a cargo handling trajectory requirement set.

[0026] According to step S200, the environmental map model of the logistics distribution center is obtained and gridded, and each grid unit is used as the basic node for the A* algorithm search; based on the driving trajectory of the AGV robot and the layout information of the logistics distribution center, the attributes of each grid are annotated; based on the cargo handling trajectory requirement set, the coordinate positions of each start-end point pair are clearly corresponding to the map grid nodes, and the start and end anchor points are anchored for the driving trajectory; for real-time road condition information, according to the frequency and time distribution analysis of historical driving trajectory data, the congestion probability and average travel time of each section at different time periods are statistically analyzed to simulate the actual traffic congestion level; based on the multi-machine collaborative operation records in the robot's task execution data, the robot's avoidance rules at intersections and parallel sections are sorted out, quantified as waiting time or driving trajectory adjustment costs under different scenarios, and integrated into the model rule set; Based on the past AGV driving trajectory data, it is found that the frequent stop (loading and unloading) areas are concentrated in the grids around the shelves in the storage area, which are marked as "loading and unloading point" attributes; the channel area is divided into "main channel" (width 5 meters, low probability of congestion in the past) and "secondary channel" (width 3 meters, occasional congestion) according to the width and historical traffic smoothness, corresponding to different traffic cost coefficients; the grids occupied by pillars or shelves are marked as "obstacle areas" and are not accessible. The historical driving trajectory data of the past month (divided into 10 hours of operation per day and 6 10-minute periods per hour) are analyzed to count the congestion of each section (connection between adjacent grids). It is found that the main channel (30-40 meters on the horizontal axis and the entire vertical axis) from the storage area to the sorting area has a congestion probability of 30% during the two periods of 9-10 am and 3-4 pm every day, and the average travel time is 2-3 minutes longer than the smooth period. This is used to quantify the real-time road condition additional cost of each section and store it in the road condition database for A* algorithm path cost accounting. Reviewing the multi-machine collaborative operation records in the task execution data, it was found that at intersection sections (such as the intersection of two main channels), the "first come, first served, low-speed avoidance" rule was followed, and the waiting time was 30 seconds on average; if the distance between parallel sections was less than 2 meters, the driving trajectory needed to be adjusted to a lateral offset of 1 meter, and the additional path length cost was estimated to be 5 meters. This type of waiting time and trajectory adjustment cost was converted into the corresponding node expansion penalty value and integrated into the rule set.

[0027] The A* algorithm is based on the heuristic function f(n) = g(n) + h(n), where g(n) is the actual distance of the comprehensive driving trajectory (accumulated based on the basic unit of grid length of 5 meters, with an additional cost of 1 meter for each "sub-channel" grid, a detour cost of 2 meters for grids around "obstacle areas", and an average delay length of the corresponding period for sections in congested periods) and the additional cost of real-time traffic conditions (calculated based on statistical congestion probability and time consumption); h(n) uses Manhattan distance to estimate the expected cost to the destination (ignoring the impact of intermediate obstacles, simply estimating the sum of the absolute values ​​of the difference between the horizontal and vertical coordinates multiplied by 5 meters / grid). For each task combination, 5 search iterations are performed, and each iteration tries different initial expansion directions and exploration strategies to explore diverse candidate paths.

[0028] During the search process, each time a node is expanded, the model rule set is verified. When there are running trajectories of other robots around the expanded node, the current driving trajectory expansion branch is paused and the alternative branch driving trajectory is explored first. The A* algorithm is used to perform several search iterations to generate several candidate optimal driving trajectories for each task combination.

[0029] After 5 iterations, task T1 generates 3 candidate optimal driving trajectories: Trajectory 1: Total length 95 meters (including avoidance cost conversion), passing through 19 grids, estimated travel time 8 minutes (including waiting and congestion delays), crossing 2 "sub-channels" and 1 intersection (performing avoidance operations).

[0030] Trajectory 2: Length 105 meters, passing through 21 grids, estimated time 9 minutes, avoiding the most congested sections, but detouring a longer distance, with 3 parallel section adjustment operations.

[0031] Track 3: 88 meters long, passing 18 grids, and taking 7.5 minutes. This path is relatively direct, but it encounters slight congestion in a "secondary channel". It cleverly uses the time difference to move forward.

[0032] Similarly, a corresponding number of candidate optimal trajectories are generated for the T2-T10 task combinations to construct a task-trajectory candidate set, which provides a rich selection space for the subsequent task allocation link to select the best one based on factors such as robot status and load, thereby improving the overall operation efficiency and collaborative safety of AGVs in the logistics distribution center.

[0033] According to step S300, from the perspective of cargo type, cargo handling tasks are divided into fresh and perishable, high-value precision electronics, ordinary daily necessities, and large industrial parts; from the perspective of cargo source, cargo handling tasks are divided into supplier replenishment tasks, online e-commerce order delivery tasks, and offline store transfer tasks; the urgency of cargo delivery is calculated based on the remaining time from the promised delivery time, and the shorter the remaining time, the higher the priority; When constructing a task requirement list, it is presented in a structured table format, with each row recording the information of a single handling task; task-related constraints are reflected in the cargo matching and process sequence, sorting out the logical relationship between tasks.

[0034] In this embodiment, 10 new cargo handling tasks (i.e., n=10) are generated, which are marked as The task types include emergency replenishment of fresh products (need to be quickly transported to the refrigerated storage area to replenish inventory), bulk shipment of popular e-commerce products (need to be sent to the shipping platform in time), and replenishment of regular daily necessities. The estimated weight of the goods is between 30 kg and 250 kg based on historical order data and product types.

[0035] The logistics distribution center is equipped with 5 AGV robots (i.e. m=5) to participate in this task allocation simulation, numbered , with rated load capacities of 100 kg, 150 kg, 200 kg, 120 kg and 180 kg respectively. Their initial positions are randomly distributed at different grid nodes in the distribution center. Each robot is equipped with a complete operation status monitoring system and wireless communication module, and exchanges data with the central dispatching system in real time.

[0036] According to the actual operation precision, the priority level is divided into N=5, the smaller the number, the higher the priority. Combined with the real-time inventory system feedback, fresh food replenishment tasks Due to the shortage of inventory (only enough to sustain sales for 2 hours), set ; E-commerce product shipment tasks At the peak of daily shipments and with only 20% of inventory left, set ; The other routine daily necessities replenishment tasks are prioritized according to the relative sufficiency of inventory and daily replenishment plan. , Different levels are formed to form a task priority list, and then arranged in descending order to prepare for subsequent task assignment.

[0037] by For example, the starting coordinates are (10,10) (the grid coordinates correspond to the actual location of the distribution center). The current coordinates are (30,20), using the Euclidean distance formula rice.

[0038] For the running status score, the weight is , , . After monitoring, The electrical system health is 0.8 (battery charge 70%, motor temperature normal), the mechanical structure integrity is 0.9 (slight vibration, no obvious wear), and the braking and steering system reliability is 0.85. . Similarly, the running status scores of other robots are calculated.

[0039] For the task (estimated cargo weight is 50 kg) and (rated cargo weight is 100 kg), then .

[0040] Based on the big data analysis of historical driving trajectories, we know Current location Go to The starting point passes through 3 sections, and the congestion probability and estimated additional time of each section are: Section 1 (congestion probability is 0.2, additional time is 10 seconds), Section 2 (congestion probability is 0.1, additional time is 5 seconds), Section 3 (congestion probability is 0.3, additional time is 15 seconds). After conversion to a unified unit (meters / second, the average speed of the robot is 1 meter / second), meters, in order to quantify the congestion of each robot going to the starting point of each task.

[0041] Set the weight coefficients: task priority weight coefficient is 0.3, distance weight coefficient is 0.2, operation status score weight coefficient is 0.2, rated cargo weight adaptability weight coefficient is 0.2, and congestion index weight coefficient is 0.1. Calculate the weight coefficient of each robot for the task. Comprehensive adaptation indicators.

[0042] According to the above process, all tasks are traversed in sequence, and the final task allocation result is as follows: Assigned to , Assigned to , Assigned to , Assigned to , Assigned to , Assigned to (at this time Finish You can take on new tasks later). Assigned to , Assigned to , Assigned to , Assigned to .

[0043] According to step S300, a set of initial task allocation plans are randomly generated, each plan covers the allocation of various types of cargo handling tasks to different AGV robots, and specifies the driving trajectory; based on the task execution data of the AGV robots, the average time spent on different types of tasks by each robot in the past is counted, and the execution time of a single task is estimated in combination with the characteristics of the current task; when evaluating the allocation plan, the estimated time of all tasks is accumulated, and the time spent on the connection between tasks is considered to obtain the total task execution time of the overall plan; the shorter the plan time, the higher the score in the execution efficiency dimension; In the task allocation scheme evaluation, the cumulative load weight of the tasks assigned to each robot is analyzed to measure whether the load distribution is balanced; the smaller the standard deviation, the more balanced the load of each robot; for the driving trajectory selected by each robot in each allocation scheme, the probability of encounter within the same road section and time window is analyzed, and historical collision events and congestion point data are reviewed. Combined with real-time location tracking, points are deducted for high-risk driving trajectory combinations; After several rounds of simulated annealing, until the number of iterations reaches the target, the optimal solution is sorted and converted into a task assignment list.

[0044] According to step S400, task instructions are customized for each AGV robot involved based on the task allocation list. The task instructions are presented in the form of digital data packets, clearly specifying the starting coordinates and the end coordinates of the task; the weight threshold that the robot needs to carry is set in the instruction in combination with the average weight of similar goods in the past counted by the pressure sensor of the cargo platform and the estimated weight of the current order; at the monitoring end, when the task instruction is successfully issued to the AGV robot, the supporting monitoring software system synchronously triggers the task start state marking action, and on the distribution center map interface presented in two-dimensional form, the icon representing the corresponding AGV robot will change, and a pop-up window will be displayed next to the icon to display the task information.

[0045] According to step S400, in the time dimension, the start time and end time of the task are recorded to obtain the actual time taken by the task; the actual driving trajectory of the robot is obtained, and the operating status data is synchronously summarized; in the time dimension comparison, the estimated task execution time is compared with the actual time taken, and the path data is compared to verify the predicted optimal driving trajectory and the actual driving trajectory, and analyze the reasons for the deviation; in terms of operating status comparison, the estimated battery power consumption and motor temperature range are obtained. Differences with actual data; re-execute steps S100-S300 to trigger the model self-learning update mechanism.

[0046] A monitoring system for AGV robot working data, comprising: Data collection and integration module: including: task execution data collection unit and robot work data set integration unit; the task execution data collection unit uses laser navigation to obtain the real-time position of the AGV robot in the logistics distribution center, records its driving trajectory, monitors the weight of each shipment through the pressure sensor installed on the AGV cargo platform, monitors the operating status of the AGV robot, and records the task execution data of the AGV robot; the robot work data set integration unit pre-processes and integrates the driving trajectory, cargo weight, operating status and task execution data of the AGV robot to form a robot work data set; Trajectory analysis and path planning module: including: trajectory clustering analysis unit and dynamic path planning unit; among them, the trajectory clustering analysis unit is based on the robot work data set, and uses clustering analysis based on the driving trajectories of all robots to cluster similar trajectories into one category, and obtains a cargo handling trajectory requirement set containing several start-end pairs; the dynamic path planning unit is based on the cargo handling trajectory requirement set, and uses the A* algorithm to build a dynamic driving trajectory planning model, which integrates the driving trajectory distance, real-time road conditions and avoidance rules to generate several candidate optimal driving trajectories for each task combination; Task scheduling and allocation module: including: task classification and priority setting unit, greedy algorithm allocation unit and simulated annealing optimization unit; among them, the task classification and priority setting unit classifies the newly added cargo handling tasks of the logistics distribution center in the current period, sets the task priority according to the urgency of cargo delivery and inventory storage requirements, builds a task requirement list, and sorts out the logical relationship of task sequence; the greedy algorithm allocation unit uses a greedy algorithm to preferentially allocate high-priority tasks to AGV robots that are close, in normal operation, with suitable rated cargo weight and no congestion in the driving trajectory; the simulated annealing optimization unit uses a simulated annealing mechanism to evaluate the comprehensive task execution efficiency, robot load balance and collision risk of different candidate optimal driving trajectories, and generate a task allocation list; Instruction issuing and monitoring feedback module: including: instruction issuing unit, monitoring and tracking unit and data comparison and model updating unit; among them, the instruction issuing unit issues task instructions to the corresponding AGV robot control system based on the task allocation list, starts the task execution process, and the monitoring and tracking unit marks the task start status at the monitoring end and starts full-process tracking; the data comparison and model updating unit collects the actual task execution data after the AGV robot completes a single task and a round of task sequence, compares and analyzes it with the predicted data in the task allocation stage, and triggers the model self-learning update mechanism.

[0047] In this embodiment, the instruction issuing unit customizes a detailed instruction data packet for each task for the corresponding AGV robot based on the task assignment list. Taking T1 assigned to R2 as an example, the instruction data packet contains precise starting coordinates (5, 8) and end coordinates (18, 15) (corresponding to the actual location of the distribution center, based on the grid positioning system), and clearly stipulates that the cargo weight threshold is set to 80kg (considering a certain margin to ensure safe handling), and also attaches operating specifications, such as the need to maintain a stable low-speed driving (speed limited to 0.5m / s) for handling fresh products, and avoid sudden braking and violent shaking.

[0048] On the monitoring side (based on a powerful logistics monitoring software platform with 2D / 3D visualization map display, real-time data panel and other functions), when the task instruction is successfully issued to each AGV robot, the system automatically triggers the task start state marking action. On the visual map interface, the icon representing R2 executing T1 task (identified by a unique color and number, such as blue R2 - T1) changes from gray in standby state to orange in execution, and a pop-up window displays detailed task information next to the icon, including task number T1, start time (accurately recorded as 2024 - 08 - 15 10:05:00, accurate to seconds, based on the monitoring system clock synchronization), cargo type "fresh products", and estimated completion time (based on the previous path planning model combined with task priority, robot performance, etc., a comprehensive estimate of 15 minutes).

[0049] When R2 successfully completed the T1 task, the system automatically collected its actual task execution data. The actual time consumption was recorded as 17 minutes (exceeding the estimated 15 minutes), mainly because when passing through a narrow passage, it encountered other robots and temporarily crossed to avoid (additional consumption of about 2 minutes), the actual driving track length was 98 meters (detoured a distance longer than the planned path, due to the avoidance operation to change the path), and the final cargo weight was stabilized at 60kg (consistent with the estimate), and the battery power consumption was 70% (originally planned to consume 75%, due to additional avoidance and path extension increased energy consumption).

[0050] Based on the above comparative analysis results, the time deviation, path deviation, power deviation and corresponding task scenarios (such as narrow channel intersection scenarios), robot status (speed, load, etc. at the time) data are classified and organized, and injected into the training set of the corresponding models (path planning model, task estimation time model, battery energy consumption model).

[0051] The precise command push of the command issuing unit and the real-time control of the monitoring and tracking unit ensured the orderly start and execution of the tasks. Although there were certain time and path deviations in T1, the overall completion rate of the 10 tasks reached 100%, and the average task execution time was shortened by about 15% compared with the previous time when the monitoring feedback mechanism was not optimized (the average time in the past was about 20 minutes, and the average time in this simulation was about 17 minutes). This was mainly due to the precise task allocation and real-time monitoring and adjustment, which effectively reduced the time waste caused by incorrect operations, getting lost and other problems.

[0052] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

Claims

1. A method for monitoring working data of an AGV robot, characterized in that: The method comprises the following steps: S100, obtaining the real-time position of the AGV robot in the logistics distribution center, recording its driving trajectory, monitoring the weight of each shipment, monitoring the operating status of the AGV robot, and recording the task execution data of the AGV robot; pre-processing and integrating the driving trajectory, cargo weight, operating status, and task execution data of the AGV robot to form a robot work data set; S200, based on the robot work data set, cluster analysis is used to cluster similar trajectories into one category based on the driving trajectories of all robots, and a cargo handling trajectory requirement set is obtained; based on the cargo handling trajectory requirement set, a dynamic driving trajectory planning model is constructed using the A* algorithm, and a number of candidate optimal driving trajectories are generated for each task combination by integrating the driving trajectory distance, real-time road conditions and avoidance rules; S300. Classify the newly added cargo handling tasks of the logistics distribution center in the current period, set the task priority according to the urgency of cargo delivery and inventory storage requirements, and build a task requirement list; use a greedy algorithm to preferentially assign high-priority tasks to AGV robots that are close, in normal operation, with a suitable rated cargo weight, and without congestion in their driving trajectories; use a simulated annealing mechanism to evaluate the comprehensive task execution efficiency, robot load balancing, and collision risks of different candidate optimal driving trajectories, and generate a task assignment list; S400. Based on the task allocation list, the task instructions are issued to the corresponding AGV robot control system, the task execution process is started, the task start status is marked on the monitoring end, and the whole process tracking is started; when the AGV robot completes a single task and a round of task sequence, its actual task execution data is collected, compared and analyzed with the predicted data in the task allocation stage, and the model self-learning update mechanism is triggered.

2. A method for monitoring working data of an AGV robot according to claim 1, characterized in that: According to step S100, the AGV robot is equipped with a laser scanner on the top. The laser scanner emits a laser beam. Using the triangulation method, the robot can calculate its own coordinates in the two-dimensional plane coordinate system of the logistics distribution center in real time and send them to the monitoring server. Whenever the monitoring server receives new coordinates, it stores them in the local cache in the order of timestamps. The cache is rolled and stored according to the first-in-first-out principle to obtain the driving trajectory of the AGV robot. The pressure sensor installed on the AGV cargo platform monitors the weight of each shipment. Monitor the running status of the AGV robot, monitor the voltage, charge and discharge current and battery temperature of each battery in real time, and transmit the data to the AGV main controller through the CAN bus; install Hall sensors and temperature sensors on the motor part, the Hall sensors monitor the motor speed, and the temperature sensors monitor the motor winding temperature; use the vibration acceleration sensor on the moving part of the AGV robot to capture transmission vibration anomalies based on strain, vibration frequency and amplitude characteristics; obtain the task execution data of the AGV robot from the built-in task management module of the AGV robot; The driving trajectory, cargo weight, operating status and task execution data of the AGV robot are preprocessed and integrated to form a robot work data set.

3. A method for monitoring working data of an AGV robot according to claim 1, characterized in that: According to step S200, for each driving trajectory, a set of cargo handling features is extracted, and the cargo handling features include geometric features, statistical features, and spatiotemporal features; wherein, the coordinates of the starting point and the end point of the driving trajectory are calculated, and the difference between the coordinates of the starting point and the end point is used as a vector component as a geometric feature; the total length, the number of stop points, and the average speed of the driving trajectory are counted as statistical features; the time period of the start time and the end time of the trajectory is obtained as a spatiotemporal feature; the cargo handling features are integrated to convert each driving trajectory into a multidimensional vector in the form of: [starting horizontal coordinate difference, starting vertical coordinate difference, trajectory length, number of stop points, average speed, starting time period code, ending time period code], and all multidimensional vectors constitute a cluster analysis input matrix; A density-based spatial clustering algorithm combined with a hierarchical clustering algorithm is used to cluster similar trajectories in the cluster analysis input matrix into one category; the spatial clustering algorithm traverses the trajectory vector set according to the set neighborhood radius and the minimum number of points, identifies the core points and boundary points, clusters the connected points into initial clusters, and divides the initial trajectory categories; hierarchical clustering is introduced to merge and split the preliminary clusters, and a clustering dendrogram is constructed from the bottom up using the Euclidean distance as an indicator. The clustering quality is evaluated based on the silhouette coefficient, the optimal number of clusters is determined, and similar trajectories are clustered into one category; similar trajectories are traversed, the common starting point and end point coordinate ranges are extracted, and summarized into several start-end points pairs to form a cargo handling trajectory demand set.

4. A method for monitoring working data of an AGV robot according to claim 3, characterized in that: According to step S200, the environmental map model of the logistics distribution center is obtained and gridded, and each grid unit is used as the basic node for A* algorithm search; based on the driving trajectory of the AGV robot and the layout information of the logistics distribution center, the attributes of each grid are annotated; based on the cargo handling trajectory requirement set, the coordinate positions of each start-end point pair are clearly corresponding to the map grid nodes, and the start and end anchor points are anchored for the driving trajectory; for real-time road condition information, according to the frequency and time distribution analysis of historical driving trajectory data, the congestion probability and average travel time of each section at different time periods are statistically analyzed to simulate the actual traffic congestion level; based on the multi-machine collaborative operation records in the robot's task execution data, the robot's avoidance rules at intersections and parallel sections are sorted out, quantified as waiting time or driving trajectory adjustment costs under different scenarios, and integrated into the model rule set; During the search process, each time a node is expanded, the model rule set is verified. When there are running trajectories of other robots around the expanded node, the current driving trajectory expansion branch is paused and the alternative branch driving trajectory is explored first. The A* algorithm is used to perform several search iterations to generate several candidate optimal driving trajectories for each task combination.

5. A method for monitoring working data of an AGV robot according to claim 1, characterized in that: According to step S300, from the perspective of cargo type, cargo handling tasks are divided into fresh and perishable, high-value precision electronics, ordinary daily necessities, and large industrial parts; from the perspective of cargo source, cargo handling tasks are divided into supplier replenishment tasks, online e-commerce order delivery tasks, and offline store transfer tasks; the urgency of cargo delivery is calculated based on the remaining time from the promised delivery time, and the shorter the remaining time, the higher the priority; When constructing a task requirement list, it is presented in a structured table format, with each row recording the information of a single handling task; task-related constraints are reflected in the cargo matching and process sequence, sorting out the logical relationship between tasks.

6. A method for monitoring working data of an AGV robot according to claim 5, characterized in that: According to step S300, task priorities are set according to the urgency of goods delivery and the quantified inventory storage requirements. , the value range is set to [1, N], where N is the number of priority levels divided according to actual operations, and is dynamically adjusted through the real-time inventory system and order delivery time monitoring data; based on the coordinate information obtained by the laser navigation system, the Euclidean distance formula is used to calculate that in the two-dimensional plane of the distribution center, if the current coordinate of the AGV robot j is , the starting coordinates of task i are , then the distance between the AGV robot and the starting point of the task is , in meters, distance data is updated in real time; comprehensive multi-dimensional operation status monitoring data is quantified to set the AGV operation status score ,in For electrical system health, Is the integrity of the mechanical structure, Indicates the reliability of the braking and steering systems. is the weight of the electrical system health, is the weight of the mechanical structure integrity, is the weight of the braking and steering system reliability. When the AGV is in full operation state ; Estimated cargo weight for task i AGV robot rated load weight , then the rated freight weight adaptability , with a value between [0,1]; based on the historical driving trajectory big data analysis combined with real-time road condition monitoring, the congestion probability and duration of each section on the driving trajectory of AGV robot j to the starting point of task i are counted, and the congestion index is set , where m is the number of driving track sections, is the congestion probability of road section k, Estimate additional time for congestion; Initialize Task Set , and AGV robot , sort the tasks in descending order by priority, where n is the total number of new tasks and m is the number of robots that can be deployed; For each task in the task set , traverse in priority order and calculate each AGV robot For tasks Comprehensive adaptation index , the formula is: ； in is the weight coefficient of task priority, is the weight coefficient of the distance between the AGV robot and the starting point of the task, is the weight coefficient of AGV operation status score, is the weight coefficient of the rated freight weight adaptability, is the weight coefficient of the congestion index; Selection The most valuable AGV robot Assigning tasks , update the robot's task execution data and task set.

7. A method for monitoring working data of an AGV robot according to claim 6, characterized in that: According to step S300, a set of initial task allocation plans are randomly generated, each plan covers the allocation of various types of cargo handling tasks to different AGV robots, and specifies the driving trajectory; based on the task execution data of the AGV robots, the average time spent on different types of tasks by each robot in the past is counted, and the execution time of a single task is estimated in combination with the characteristics of the current task; when evaluating the allocation plan, the estimated time of all tasks is accumulated, and the time spent on the connection between tasks is considered to obtain the total task execution time of the overall plan; the shorter the plan time, the higher the score in the execution efficiency dimension; In the task allocation scheme evaluation, the cumulative load weight of the tasks assigned to each robot is analyzed to measure whether the load distribution is balanced; the smaller the standard deviation, the more balanced the load of each robot; for the driving trajectory selected by each robot in each allocation scheme, the probability of encounter within the same road section and time window is analyzed, and historical collision events and congestion point data are reviewed. Combined with real-time location tracking, points are deducted for high-risk driving trajectory combinations; After several rounds of simulated annealing, until the number of iterations reaches the target, the optimal solution is sorted and converted into a task assignment list.

8. The method for monitoring working data of an AGV robot according to claim 1, characterized in that: According to step S400, task instructions are customized for each AGV robot involved based on the task allocation list. The task instructions are presented in the form of digital data packets, clearly specifying the starting coordinates and the end coordinates of the task; the weight threshold that the robot needs to carry is set in the instruction in combination with the average weight of similar goods in the past counted by the pressure sensor of the cargo platform and the estimated weight of the current order; at the monitoring end, when the task instruction is successfully issued to the AGV robot, the supporting monitoring software system synchronously triggers the task start state marking action, and on the distribution center map interface presented in two-dimensional form, the icon representing the corresponding AGV robot will change, and a pop-up window will be displayed next to the icon to display the task information.

9. A method for monitoring working data of an AGV robot according to claim 8, characterized in that: According to step S400, in the time dimension, the start time and end time of the task are recorded to obtain the actual time taken by the task; the actual driving trajectory of the robot is obtained, and the operating status data is synchronously summarized; in the time dimension comparison, the estimated task execution time is compared with the actual time taken, and the path data is compared to verify the predicted optimal driving trajectory and the actual driving trajectory, and analyze the reasons for the deviation; in terms of operating status comparison, the estimated battery power consumption and motor temperature range are obtained. Differences with actual data; re-execute steps S100-S300 to trigger the model self-learning update mechanism.

10. A monitoring system for AGV robot working data, using a monitoring method for AGV robot working data according to any one of claims 1 to 9, characterized in that: include: Data collection and integration module: including: task execution data collection unit and robot work data set integration unit; the task execution data collection unit uses laser navigation to obtain the real-time position of the AGV robot in the logistics distribution center, records its driving trajectory, monitors the weight of each shipment through the pressure sensor installed on the AGV cargo platform, monitors the operating status of the AGV robot, and records the task execution data of the AGV robot; the robot work data set integration unit pre-processes and integrates the driving trajectory, cargo weight, operating status and task execution data of the AGV robot to form a robot work data set; Trajectory analysis and path planning module: including: trajectory clustering analysis unit and dynamic path planning unit; among them, the trajectory clustering analysis unit is based on the robot work data set, and uses clustering analysis based on the driving trajectories of all robots to cluster similar trajectories into one category, and obtains a cargo handling trajectory requirement set containing several start-end pairs; the dynamic path planning unit is based on the cargo handling trajectory requirement set, and uses the A* algorithm to build a dynamic driving trajectory planning model, which integrates the driving trajectory distance, real-time road conditions and avoidance rules to generate several candidate optimal driving trajectories for each task combination; Task scheduling and allocation module: including: task classification and priority setting unit, greedy algorithm allocation unit and simulated annealing optimization unit; among them, the task classification and priority setting unit classifies the newly added cargo handling tasks of the logistics distribution center in the current period, sets the task priority according to the urgency of cargo delivery and inventory storage requirements, builds a task requirement list, and sorts out the logical relationship of task sequence; the greedy algorithm allocation unit uses a greedy algorithm to preferentially allocate high-priority tasks to AGV robots that are close, in normal operation, with suitable rated cargo weight and no congestion in the driving trajectory; the simulated annealing optimization unit uses a simulated annealing mechanism to evaluate the comprehensive task execution efficiency, robot load balance and collision risk of different candidate optimal driving trajectories, and generate a task allocation list; Instruction issuing and monitoring feedback module: including: instruction issuing unit, monitoring and tracking unit and data comparison and model updating unit; among them, the instruction issuing unit issues task instructions to the corresponding AGV robot control system based on the task allocation list, starts the task execution process, and the monitoring and tracking unit marks the task start status at the monitoring end and starts full-process tracking; the data comparison and model updating unit collects the actual task execution data after the AGV robot completes a single task and a round of task sequence, compares and analyzes it with the predicted data in the task allocation stage, and triggers the model self-learning update mechanism.

Images