Driving control system applied to multi-AGV cooperation

Abstract

The invention discloses a driving control system applied to multi-AGV cooperation, and relates to the technical field of AGV cooperation control. The system comprises a central control module and a plurality of AGVs in communication connection with the central control module, the dynamic environment adaptation capability is enhanced through multi-dimensional technology design, and the core is to construct a closed-loop mechanism of environmental perception-trajectory prediction-dynamic obstacle avoidance; according to the technical scheme, a sensor carried by an AGV is used for collecting environment data covering a working area, and the future moving track of an obstacle can be output in advance based on the environment data; in the process that the AGV goes to the task starting point, whether the optimal route is blocked or not is judged in combination with the predicted trajectory, if yes, an obstacle avoidance route is planned in real time, and it is ensured that the AGV arrives at the task starting point before the task starting moment; according to the method, real-time coping is converted into advanced avoidance, the collision risk caused by temporary obstacles is greatly reduced, delay caused by repeated path adjustment is reduced, and the safety and continuity of AGV cluster operation are effectively guaranteed.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of AGV cooperative control, and particularly relates to a driving control system applied to multi-AGV cooperation. BACKGROUND

[0002] In the field of AGV (Automatic Guided Vehicle) cooperative control, with the improvement of the requirement of intelligent manufacturing, logistics and warehousing and the like on material handling efficiency, multi-AGV cooperative operation has become a mainstream application direction. In the prior art, a multi-AGV cooperative system mostly relies on a preset static path planning scheme, that is, a work area map is drawn in advance, a fixed driving route is planned, and orderly scheduling of AGVs is realized.

[0003] The prior art cannot avoid dynamic obstacles in advance through static path planning and passive obstacle avoidance mode. AGVs often make emergency braking and repeatedly adjust paths due to sudden obstacles, which not only easily causes congestion or collision risks among AGVs, but also causes delay of material handling tasks due to time delay of path re-planning, and it is difficult to meet the safety and efficiency requirements of multi-AGV cooperation in a high dynamic scene.

[0004] The application provides a driving control system applied to multi-AGV cooperation to solve the above technical problems. SUMMARY

[0005] The application aims to at least solve one of the technical problems in the prior art. To this end, the application provides a driving control system applied to multi-AGV cooperation, which converts real-time response into advance avoidance, greatly reduces the collision risk caused by temporary obstacles, reduces the delay caused by repeated path adjustment, is particularly suitable for dynamic obstacle dense scenes such as workshops and warehouses, and effectively guarantees the safety and continuity of AGV cluster operation.

[0006] To achieve the above purpose, a first aspect of the application provides a driving control system applied to multi-AGV cooperation, comprising a central control module and a plurality of AGVs in communication connection with the central control module. The central control module is used for constructing an obstacle movement sequence according to environment data, and predicting a movement trajectory of an obstacle based on the obstacle movement sequence; wherein the environment data is used for identifying an obstacle state; and The central control module is used for determining a task start time of a plurality of work tasks in a task queue, and matching the plurality of work tasks with AGVs based on the movement trajectory and the task start time.

[0007] In a possible implementation manner, the construction of the obstacle movement sequence based on the environment data comprises: extracting environment data collected by the AGV in the process of executing the work task; wherein the environment data comprises image data and point cloud data; Identify the obstacle parameter in the environmental data, arrange the obstacle parameter according to the collection time of the environmental data, and obtain the obstacle movement sequence; wherein, the obstacle parameter includes type, position and size.

[0008] In a possible implementation, the moving track of the obstacle is predicted based on the obstacle movement sequence, comprising: Training a machine learning model based on the historical motion data set to obtain a track prediction model; wherein, the machine learning model is constructed based on an LSTM model; Constructing time sequence feature data based on the obstacle movement sequence, inputting the time sequence feature data into the track prediction model, and predicting the moving track of the obstacle.

[0009] In a possible implementation, the task start time of the work tasks in the task queue is determined, comprising: Extracting the task queue; identifying the task parameters of the work tasks in the task queue in turn, and extracting the task deadline from the task parameters; wherein, the task parameters include the task starting point, the task ending point, the task content and the task deadline; Calculating the task start time of the corresponding work task based on the task deadline and the moving track.

[0010] In a possible implementation, the AGV is matched for the work tasks based on the moving track and the task start time, comprising: Selecting the work tasks as target tasks in turn; matching a plurality of idle AGVs based on the moving track and the task start time of the target task; wherein, the idle AGV refers to the AGV in an idle state at the task start time; Determining the target AGV based on the state parameters of the plurality of idle AGVs, and associating the target AGV with the target task; wherein, the state parameters include power, load capacity and location.

[0011] In a possible implementation, the AGV is matched for the work tasks based on the moving track and the task start time, comprising: Selecting the AGV with normal working state as a candidate AGV; wherein, the candidate AGV includes the AGV without executing the work task and the AGV executing the work task; Judging whether the candidate AGV can reach the task starting point before the task start time in combination with the moving track; if yes, marking it as an idle AGV; if no, not marking it.

[0012] In a possible implementation, the candidate AGV is judged whether it can reach the task starting point before the task start time in combination with the moving track, comprising: Planning the optimal route of the candidate AGV to the task starting point; Judge whether the obstacle will block the optimal route according to the predicted moving track; if yes, plan an obstacle-avoiding route, and determine the moving route according to the optimal route; if no, take the optimal route as the moving route; If the alternative AGV can reach the starting point of the task before the starting time of the task along the moving route, the alternative AGV is marked as an idle AGV.

[0013] In a possible implementation, the target AGV is determined based on state parameters of the idle AGVs, including: Judge whether the idle AGVs can complete the target task based on the state parameters; if yes, mark the idle AGV as an alternative AGV; if no, do not mark; Select several alternative AGVs, which have the closest time to reach the starting point of the task and the starting time of the task, from the alternative AGVs as the target AGVs.

[0014] In a possible implementation, the target AGV is determined based on state parameters of the idle AGVs, including: Judge whether the idle AGVs can complete the target task based on the state parameters; if yes, mark the idle AGV as an alternative AGV; if no, do not mark; Determine the target AGV from the alternative AGVs based on the starting time of the task.

[0015] In a possible implementation, the target AGV is determined from the alternative AGVs based on the starting time of the task, including: Match the target AGV for the working task according to the starting time of the task in sequence, and lock the target AGV matched for the working task; When the alternative AGV matched for the subsequent working task is locked, judge whether the working task of locking the alternative AGV can replace the target AGV; if yes, replace the target AGV; if no, delay the subsequent working task.

[0016] Compared with the prior art, the present application has the following advantages: 1.The application enhances dynamic environment adaptation capability through multi-dimensional technology design, and the core is to build a closed-loop mechanism of "environment perception-trajectory prediction-dynamic obstacle avoidance"; the technical solution uses the sensors carried by the AGV to collect environmental data covering the working area, accurately identifies the type, position, size and other parameters of the obstacle, and arranges the obstacle movement sequence according to the collection time; at the same time, based on the historical motion data set containing the motion state, attribute and environmental context of the obstacle, the LSTM trajectory prediction model is trained to output the future movement trajectory of the obstacle in advance; during the AGV's journey to the task starting point, the predicted trajectory is combined to determine whether the optimal route is blocked, and if it is blocked, the obstacle avoidance route is planned in real time to ensure that the AGV arrives at the task starting point before the task starts; compared with the passive mode of traditional AGV relying on static path planning and real-time adjustment when encountering obstacles, the application changes real-time response to pre-avoidance, greatly reduces the collision risk caused by temporary obstacles, reduces the delay caused by repeated path adjustment, and is especially suitable for dynamic obstacle dense scenes such as workshops and warehouses, effectively ensuring the safety and continuity of AGV cluster operation.

[0017] 2.The application realizes accurate matching of AGV resources and task demand through a refined task allocation mechanism, and the core scheme includes task queue management-task time calculation-AGV multi-level screening-dynamic adjustment; first, a task queue is constructed according to task priority and ordering time, and the task start time of each work task is accurately calculated in combination with task deadline, task execution time and redundant time; then, the standby AGV is screened according to the standard of normal working state, and whether the standby AGV can arrive at the starting point before the task starts is judged in combination with the predicted trajectory of the obstacle, and is marked as an idle AGV; based on state parameters such as power and load capacity, the optional AGV that can complete the target task is selected, and finally the target AGV is determined according to the principle of "the arrival time being closest to the task start time" or "the task urgency priority", and dynamic replacement and urgent task priority protection are supported when AGV overlaps; this technical solution avoids the AGV resource idling or overload problem caused by the near allocation in traditional task allocation, fully excavates the potential of AGV resources, ensures the high adaptation of task and AGV capacity and time requirements, significantly improves the AGV resource utilization and the task completion rate on time, and is suitable for intelligent manufacturing and logistics warehouse scenes with multiple tasks and high dynamics. BRIEF DESCRIPTION OF DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the embodiments of the application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description only some embodiments of the application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of these drawings.

[0019] Figure 1A schematic diagram of a principle of a drive control system applied to multi-AGV cooperation in an embodiment of the present application; Figure 2 A schematic diagram of calculation of a task execution duration in an embodiment of the present application; Figure 3 A schematic diagram of an alternative AGV going to a task starting point when not affected by an obstacle in an embodiment of the present application; Figure 4 A schematic diagram of an alternative AGV going to a task starting point when affected by an obstacle in an embodiment of the present application. DETAILED DESCRIPTION

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

[0021] Please refer to Figure 1 The first aspect embodiment of the present application provides a drive control system applied to multi-AGV cooperation, comprising a central control module and a plurality of AGVs in communication connection with the central control module; the central control module is configured to construct an obstacle moving sequence according to environment data, predict a moving trajectory of an obstacle based on the obstacle moving sequence, wherein the environment data is used to identify an obstacle state; and determine a task start time of a plurality of work tasks in a task queue; and match the plurality of work tasks with AGVs based on the moving trajectory and the task start time.

[0022] In the embodiment of the present application, the central control module is used to cooperate with a plurality of AGVs in a work area, and is in communication connection with the plurality of AGVs. The central control module has data collection, processing and storage functions, and is configured to analyze and predict a moving trajectory of an obstacle according to environment data collected by the AGVs, combine the moving trajectory with subsequent cooperation scheduling of the plurality of AGVs, and improve the cooperation scheduling accuracy and scheduling efficiency of the plurality of AGVs.

[0023] The prediction of the moving trajectory of the obstacle is mainly realized through the environment data collected by the AGVs, i.e., the AGVs collect environment data by using various types of sensors deployed on the upper part thereof while moving along a task route during execution of a work task, and send the environment data in real time to the central control module. The central control module predicts the moving trajectory of the obstacle according to the environment data, and since the environment data returned by the AGVs can be updated in time as the work task is executed, the obstacle moving sequence is also updated in time, and thus the accuracy of the predicted moving trajectory of the obstacle is updated, which is conducive to improving the accuracy of the cooperation scheduling of the plurality of AGVs.

[0024] In a preferred embodiment, the obstacle movement sequence is constructed based on the environment data, including: extracting environment data collected by the AGV during execution of the work task; identifying obstacle parameters in the environment data, and arranging the obstacle parameters according to the collection time of the environment data to obtain the obstacle movement sequence.

[0025] The sensors deployed on the AGV include visual sensors, lidar, etc. The visual sensors mainly include monocular / binary camera, RGB-D camera, etc., which are used to collect environment images that can be used to identify obstacles; the lidar scans the surrounding environment by emitting laser beams, receives reflected light and calculates the time of flight (ToF) or phase difference of the light, thereby generating a three-dimensional point cloud model of the environment, and further identifying obstacles. One or more types of sensors can be deployed on the AGV, such as visual sensors or lidar. Of course, other sensors for identifying obstacles can also be deployed.

[0026] The obstacle parameters, i.e., the type, position and size of the obstacle, etc., can be identified from the environment data by an algorithm (e.g., a 2D / 3D target detection algorithm based on a neural network model or a point cloud algorithm based on a deep learning model). Here, the obstacle refers to an obstacle that may affect the AGV to pass along the planned path. For example, goods temporarily stacked in the passage, or workers walking in the passage. These obstacles cannot be foreseen when planning the path, and may affect the AGV to pass. However, the shelves and equipment in the work area are clear before planning the path, and therefore will not affect the AGV to move along the planned path.

[0027] It should be noted that the environment data is collected by the AGV that is executing the work task, and the obstacle parameters are identified immediately after the environment data is collected. By integrating the obstacle parameters identified in the environment data with the previously identified obstacle parameters, the obstacle movement sequence can be obtained. Due to the real-time nature of the environment data collection, the reliability and real-time nature of the obstacle movement sequence are ensured. Moreover, since multiple AGVs in the work area are executing the work task, not only can the environment data cover every corner of the work area, but also the same obstacle can be collected by multiple AGVs at the same time, which to some extent ensures the continuity of the movement trajectory of the obstacle. Of course, if the obstacle movement sequence is missing data for a certain period of time and is not conducive to predicting its movement trajectory, data can also be supplemented by sensors deployed within the work area.

[0028] In a preferred embodiment, the movement trajectory of the obstacle is predicted based on the obstacle movement sequence, including: training a machine learning model based on a historical motion data set to obtain a trajectory prediction model; constructing time series feature data based on the obstacle movement sequence, inputting the time series feature data into the trajectory prediction model, and predicting the movement trajectory of the obstacle.

[0029] The logic of the trajectory prediction model for predicting the trajectory of the obstacle is to take the motion state of the obstacle in the past N time stamps, including position, speed, direction, etc. as input features, and predict the position coordinates in the future M time stamps. The obstacle movement sequence includes the motion state of the obstacle collected by the AGV in the past period of time, and the obstacle movement sequence is converted into time sequence feature data. The time sequence feature data is input into the trajectory prediction model to predict the movement trajectory of the obstacle in the future period of time.

[0030] The historical motion data refers to a set of historical trajectory data of different obstacles collected in the working area. The historical motion data includes obstacle motion state data (such as position, speed, acceleration, etc.), obstacle attribute data (such as type, size, etc.), environmental context data (such as area type, channel width, etc.), and timestamp data. The time sequence feature data is constructed to ensure the operability of model training. The historical trajectory set is used to train a machine learning model, such as an LSTM model, to obtain a trajectory prediction model. The training process of the LSTM model is fully disclosed in the prior art, and will not be repeated here.

[0031] In some other preferred embodiments, the movement trajectory of the obstacle can also be predicted from the obstacle movement sequence by other algorithms, which will not be exemplified here.

[0032] In a preferred embodiment, the task start time of the plurality of working tasks in the task queue is determined, including: extracting the task queue; identifying the task parameters of the plurality of working tasks in the task queue in sequence, extracting the task deadline from the task parameters; and calculating the task start time of the corresponding working task based on the task deadline and the movement trajectory.

[0033] In order to provide the scheduling efficiency of the AGV, during the execution of the current working task, the task queue that needs to be scheduled by the AGV is determined, and the plurality of working tasks in the task queue are tasks that need to be processed in a timely manner next.

[0034] For example, the plurality of working tasks are sorted according to priority and order time, and the working tasks with high priority and early order time are sorted in front. According to the set number of tasks, the working tasks are selected in sequence to form a task queue, for example, the number of tasks is set to 10, and 10 working tasks are selected from the plurality of working tasks in sequence to form a task queue, and other working tasks are processed after the working tasks in the task queue are matched to enough AGVs. The number of tasks can be set based on the current total number of AGVs, for example, the number of tasks = current available AGV number x 1.2 (redundancy factor), to ensure that the queue size matches the AGV resources and avoid scheduling delay caused by too large queue.

[0035] It needs to be understood that the work tasks in the non-task queue can also be changed, such as task content change or priority change. As new work tasks are ordered, the ordering of work tasks can also change. For example, a work task with the highest priority is ordered first, and this work task can be placed in the front of all existing work tasks.

[0036] It is worth noting that if enough AGVs are matched for several work tasks in the task queue, and some AGVs have no work tasks to execute, the next task queue can be determined, and AGVs can be matched for several work tasks in the task queue, to avoid some AGVs being in an idle state for a long time.

[0037] The task parameters of several work tasks in the task queue are extracted in turn, including task start point, task end point, task content, and task deadline. The task deadline of the work task is extracted from the task parameters, and the task execution time of the work task is calculated according to the task content, task start point, and task end point. The task start time of the work task can be calculated from the task deadline, task execution time, and redundancy time. That is, the AGV needs to arrive at the task start point before the task start time, otherwise the corresponding work task cannot be completed within the task deadline.

[0038] It also needs to be understood that the redundancy time includes the preparation time of the AGV and the time for loading and unloading goods. The task execution time refers to the time spent by the AGV from the task start point to the task end point. If the AGV encounters an obstacle during movement, the task execution time will be longer.

[0039] Figure 2 A schematic diagram for calculating the task execution time is shown. In general, when the AGV moves from position Q to position M to perform a carrying task, the moving route is Y1, and the task execution time can be calculated according to the length of the moving route Y1 and the average speed of the AGV. However, if it is determined that the predicted obstacle movement trajectory intersects with the moving route Y1, the obstacle is likely to affect the task execution time of the AGV. For example, Figure 2 In the figure, the gray grid represents the position where the obstacle meets the AGV, and at this time the AGV needs to avoid the obstacle, so planning an obstacle avoidance path, executing the obstacle avoidance path, etc. will cause the task execution time to be prolonged. The time required to move along Y1 and pass Y2 to bypass the obstacle and then move along Y1 to position M can be calculated as time one, and the time required to plan the obstacle avoidance path can be calculated as time two. The task execution time is obtained by adding time one and time two.

[0040] For example, assume that work task A requires the AGV to carry goods from position Q to position M before 10 am, the task start point is position Q, the task end point is position M, the task content is carrying goods, and the task deadline is 10 am.Figure 2 As shown, if the task execution duration is 2 minutes and the redundancy duration is 1 minute, the task start time is 9:57 am.

[0041] In another preferred embodiment, the task execution duration can also be set according to empirical data, as long as the AGV can complete the work task on time.

[0042] In a preferred embodiment, the AGV is matched for a plurality of work tasks based on the movement trajectory and the task start time, including: selecting a work task as a target task in sequence; matching a plurality of idle AGVs based on the movement trajectory and the task start time of the target task; determining a target AGV based on the state parameters of the plurality of idle AGVs, and associating the target AGV with the target task.

[0043] The task start time refers to the time before which the AGV reaches the task starting point and can complete the work task. If the AGV does not reach the task starting point at the task start time, it certainly cannot select the corresponding AGV as the target AGV to complete the work task. When matching the idle AGVs based on the task start time, it is essentially to determine whether the AGV can reach the task starting point before the task start time. Not only the AGV that does not execute the work task is within the judgment range, but also the AGV that is executing the work task is within the judgment range. If the AGV does not execute the work task and it can reach the task starting point before the task start time, the AGV can be used as an idle AGV. Similarly, if the AGV that is executing the work task can also reach the task starting point before the task start time after completing the task, the AGV can also be used as an idle AGV.

[0044] Matching the AGV for the work task based on the task start time mainly includes two steps: First, selecting the AGVs that are idle before the task start time of the target task, i.e., these AGVs can be used to execute the target task. For example, the AGV that is always idle can certainly be used to execute the target task; the AGV that is currently executing the work task but can complete the task before the task start time is also idle before the task start time, and can also be used to execute the target task.

[0045] Second, selecting the AGVs from the idle AGVs that can reach the task starting point before the task start time and have the ability to complete the target task. If the AGV is far away from the task starting point and cannot reach the task starting point before the task start time, it cannot be used to execute the target task; if the AGV can reach the task starting point before the task start time, but its power and load capacity are insufficient to complete the target task, it also cannot be used to execute the target task.

[0046] It is worth noting that in the process of judging whether the AGV can reach the starting point of the task at the task starting time, the moving speed of the AGV should not exceed the highest speed limit (it can reach the starting point of the task at the highest speed limit), and the power should be sufficient to reach the starting point of the task from its current position (as a screening condition, it is not judged whether the power is sufficient to execute the target task).

[0047] In a preferred embodiment, matching a plurality of idle AGVs based on the moving trajectory and the task starting time of the target task includes: selecting AGVs in a normal working state as candidate AGVs; judging whether the candidate AGVs can reach the starting point of the task before the task starting time based on the moving trajectory; if yes, marking them as idle AGVs; if no, not marking them.

[0048] In matching idle AGVs for the target task, all AGVs in a normal state are first selected as candidate AGVs, with the purpose of eliminating AGVs that are in a fault state or cannot normally execute a working task due to other reasons. Of course, the candidate AGVs include AGVs that do not execute a working task and AGVs that are executing a working task.

[0049] Then, it is judged whether the candidate AGVs can reach the starting point of the task before the task starting time of the target task, and if they can reach the starting point of the task, they are marked as idle AGVs, that is, the idle AGVs can at least guarantee to be at the starting point of the task at the task starting time, so as to guarantee not to affect the execution of the subsequent target task.

[0050] In a preferred embodiment, judging whether the candidate AGVs can reach the starting point of the task before the task starting time based on the moving trajectory includes: planning an optimal route for the candidate AGVs to reach the starting point of the task; judging whether the optimal route will be blocked by obstacles based on the predicted moving trajectory; if yes, planning an obstacle-avoiding route and determining a moving route based on the optimal route; if no, taking the optimal route as the moving route; if the candidate AGVs can reach the starting point of the task before the task starting time along the moving route, marking the candidate AGVs as idle AGVs.

[0051] Based on the moving trajectory of the obstacles, it is judged whether the AGVs can reach the starting point of the task when moving normally or when bypassing the obstacles.

[0052] In judging whether the candidate AGVs can reach the starting point of the task before the task starting time, it is mainly judged whether the candidate AGVs have enough time to reach the starting point of the task. The time spent in reaching the starting point of the task is affected by the moving speed and the distance, and may also be affected by obstacles.

[0053] If not affected by the obstacle, first calculate the path distance between the position of the alternative AGV and the starting point of the task, and the time required to calculate the moving path distance at the average speed of the alternative AGV, and judge whether the alternative AGV can reach the starting point of the task before the task starts according to the time.

[0054] It is worth noting that the path distance corresponds to the distance of the optimal route between the position of the AGV and the starting point of the task (the optimal route is planned according to the environment of the working area, considering the impact of shelves, equipment, etc.). The position of the alternative AGV is divided into two cases: one is the alternative AGV that has not executed the work task, and the position of the alternative AGV is its current position; the other is the alternative AGV that is executing the work task, and the position of the alternative AGV is the task endpoint currently executing the work task.

[0055] Figure 3 The schematic diagram of the alternative AGV going to the starting point of the task without being affected by the obstacle. Figure 3 There are two alternative AGVs, B1 and B2, which are their current positions, B1 corresponds to the AGV that has not executed the work task, and the optimal route of B1 to reach the position Q is the "L" type route YB1. Since there is no obstacle, the time for B1 to reach the position Q according to the optimal route can be calculated directly. Assuming that the AGV at B1 can move immediately, whether it can reach the position Q before the task starts can be judged based on the current time and the time required according to the optimal route YB1. B2 is an alternative AGV that is executing a work task, and B2' is the position of the alternative AGV after the current task is completed. The optimal route for B2' to reach the position Q is a straight line YB2. The time required for B2' to reach the position Q according to the straight line YB2 can be calculated, and whether it can reach the position Q before the task starts can be judged in combination with the time to reach B2'.

[0056] Figure 4 The schematic diagram of the alternative AGV going to the starting point of the task affected by the obstacle. Figure 4 The gray grid in the figure is an obstacle, which will affect the alternative AGV at B1 to go to the position Q according to the predicted moving trajectory of the obstacle, and will also affect the alternative AGV at B2' to go to the position Q. At this time, the time required to plan the obstacle avoidance route and the time to go to the position Q according to the planned obstacle avoidance route need to be considered. Whether it can reach the position Q before the task starts can be judged by adding the two times and combining the departure time of the alternative AGV.

[0057] In a preferred embodiment, the target AGV is determined based on the state parameters of the plurality of idle AGVs, including: determining whether the plurality of idle AGVs can complete the target task based on the state parameters; if yes, marking the idle AGVs as optional AGVs; if no, not marking; and selecting a plurality of optional AGVs that arrive at the starting point of the target task at a time closest to the starting time of the target task as the target AGVs.

[0058] After determining the plurality of idle AGVs that can arrive at the starting point of the target task before the starting time of the target task, further screening is needed, specifically, screening whether the optional AGVs have the ability to complete the target task. According to the state parameters of the optional AGVs, i.e., whether the power and load capacity meet the task requirements of the target task, the optional AGVs that meet the task requirements are selected as the optional AGVs, and these optional AGVs can all meet the task requirements of the target task. If the target task needs to be completed by multiple AGVs, then a sufficient number of optional AGVs are selected according to the principle of arriving at the starting point of the target task at a time closest to the starting time of the target task, and these optional AGVs are the target AGVs for executing the target task. If the target task needs to be completed by one AGV, then one optional AGV is selected as the target AGV according to the principle of arriving at the starting point of the target task at a time closest to the starting time of the target task.

[0059] It should be noted that the order of screening the AGVs based on the state parameters and screening the AGVs based on the starting time of the task can be exchanged, i.e., the optional AGVs can be screened based on the state parameters first, and then the idle AGVs are screened from the optional AGVs based on the starting time of the task, and finally the target AGVs required are determined according to the principle of arriving at the starting point of the target task at a time closest to the starting time of the target task.

[0060] In a preferred embodiment, the target AGV is determined based on the state parameters of the plurality of idle AGVs, including: determining whether the plurality of idle AGVs can complete the target task based on the state parameters; if yes, marking the idle AGVs as optional AGVs; if no, not marking; and selecting a plurality of optional AGVs that arrive at the starting point of the target task at a time closest to the starting time of the target task as the target AGVs.

[0061] In this embodiment, the optional AGVs are not further screened according to the principle of "arriving at the starting point of the target task at a time closest to the starting time of the target task", so that each target task corresponds to multiple target AGVs, and the number is greater than the task requirements of the target task.

[0062] In a preferred embodiment, the target AGV is determined from the several optional AGVs based on the task start time, comprising: matching the target AGV for the work task in order of the task start time, and locking the target AGV matched for the work task; when the optional AGV matched for the subsequent work task is locked, determining whether the work task locking the optional AGV can change the target AGV; if yes, changing the target AGV; if no, delaying the subsequent work task.

[0063] It should be noted that the optional AGV corresponding to the work task in the task queue can complete the corresponding work task in time. For example, a work task corresponds to several optional AGVs, and it is assumed that one AGV can complete the work task, so any one of the several optional AGVs can complete the work task.

[0064] However, some work tasks may correspond to multiple optional AGVs, but some work tasks correspond to only one optional AGV, and since the AGVs corresponding to these work tasks may overlap, if the AGVs of the work tasks are determined in order according to the principle that the time of reaching the task start point is closest to the task start time, it will cause some work tasks to be unable to match the target AGV. For example, work task A1 corresponds to optional AGVs B1, B2 and B3, and work task A2 corresponds to only optional AGV B1; since work task A1 is in the front of the task queue, and B1 reaches the task start point at a time closest to the task start time of work task A1, B1 should be used as the target AGV of work task A1, and in this case, work task A2 cannot match the target AGV.

[0065] To solve this problem, the embodiment matches the target AGV based on the task start time. The earlier the task start time, the more urgent the corresponding work task time requirement, and the work tasks are extracted in order of the task start time, and the target AGV is determined according to the several optional AGVs corresponding to the work tasks. If the optional AGVs still overlap during the matching process, the target AGV of the previous work task is changed.

[0066] For example, the work task A1 corresponds to optional AGVs B1, B2 and B3, and the work task A2 corresponds to optional AGVs only B1. If the work task A2 has an earlier task start time, B1 is selected as the target AGV of the work task A2, and the target AGVs of the work task A1 are selected from B2 and B3; or, if the work task A1 has an earlier task start time, B1 is selected as the target AGV of the work task A1, and at this time, B1 is understood to be locked, and when the target AGV of the work task A2 is matched, it is found that B1 is already locked, and at this time, the target AGV of the work task A1 is changed, and B2 is selected as the target AGV of the work task A1. Also, if the work task A1 also only has B1, B1 is selected as the target AGV of the work task A1 (locked and not released), and the work task A2 is processed in the next task queue.

[0067] The above embodiments are only used to illustrate the technical method of the present application and not limit the present application. Although the present application is described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical method of the present application can be modified or replaced equivalently without departing from the spirit and scope of the technical method of the present application.

Claims

1. A drive control system for multi-AGV collaboration, characterized in that, It includes a central control module and several AGVs that communicate with it; Central control module: used to construct an obstacle movement sequence based on environmental data, and predict the movement trajectory of obstacles based on the obstacle movement sequence; wherein, environmental data is used to identify the obstacle state; and, Used to determine the start time of several work tasks in the task queue; AGVs are matched for several work tasks based on the movement trajectory and the task start time.

2. The drive control system for multi-AGV collaboration according to claim 1, characterized in that, Obstacle movement sequences were constructed based on environmental data, including: Extract environmental data collected by the AGV during the execution of its work tasks; the environmental data includes image data and point cloud data. Obstacle parameters in environmental data are identified, and the obstacle parameters are arranged according to the acquisition time of the environmental data to obtain an obstacle movement sequence; wherein, the obstacle parameters include type, location and size.

3. The drive control system for multi-AGV collaboration according to claim 2, characterized in that, Predicting the trajectory of an obstacle based on the obstacle movement sequence includes: A trajectory prediction model is obtained by training a machine learning model based on a historical motion dataset; the machine learning model is built based on an LSTM model. Based on the obstacle movement sequence, time-series feature data is constructed, and the time-series feature data is input into the trajectory prediction model to predict the movement trajectory of the obstacle.

4. A drive control system for multi-AGV collaboration according to any one of claims 1 to 3, characterized in that, Determine the start time of several tasks in the task queue, including: Extract the task queue; sequentially identify the task parameters of several work tasks in the task queue, and extract the task deadline from the task parameters; wherein, the task parameters include the task start point, task end point, task content, and task deadline; The task start time of the corresponding work task is calculated based on the task duration and the movement trajectory.

5. A drive control system for multi-AGV collaboration according to claim 1, characterized in that, Based on the movement trajectory and the task start time, several AGVs are matched for work tasks, including: Work tasks are selected sequentially as target tasks; several idle AGVs are matched based on the movement trajectory and the task start time of the target task; wherein, an idle AGV refers to an AGV that is in an idle state at the task start time; The target AGV is determined based on the status parameters of several idle AGVs, and the target AGV is associated with the target task; wherein, the status parameters include power, load capacity and location.

6. A drive control system for multi-AGV collaboration according to claim 5, characterized in that, Based on the movement trajectory and the task start time of the target task, several idle AGVs are matched, including: AGVs in normal working condition are used as backup AGVs; among them, backup AGVs include AGVs that are not performing work tasks and AGVs that are performing work tasks. Based on the movement trajectory, determine whether the candidate AGV can reach the task start point before the task starts; if yes, mark it as an idle AGV; otherwise, do not mark it.

7. A drive control system for multi-AGV collaboration according to claim 6, characterized in that, Determining whether the candidate AGV can reach the task start point before the task start time based on the movement trajectory includes: Plan the optimal route for the candidate AGVs to reach the task starting point; Based on the predicted movement trajectory, determine whether an obstacle will block the optimal route; if yes, plan an obstacle avoidance route and determine the movement route based on the optimal route; if no, use the optimal route as the movement route. If the candidate AGV can reach the task start point before the task start time along the movement route, then the candidate AGV is marked as an idle AGV.

8. A drive control system for multi-AGV collaboration according to claim 5, characterized in that, Determining the target AGV based on several state parameters of the idle AGVs includes: Based on the status parameters, determine whether several idle AGVs can complete the target task; if yes, mark the idle AGV as an optional AGV; otherwise, do not mark it. Select the target AGVs from the available AGVs whose arrival time at the task start point is closest to the task start time.

9. A drive control system for multi-AGV collaboration according to claim 5, characterized in that, Determining the target AGV based on several state parameters of the idle AGVs includes: Based on the status parameters, determine whether several idle AGVs can complete the target task; if yes, mark the idle AGV as an optional AGV; otherwise, do not mark it. The target AGV is determined from a number of selectable AGVs based on the start time of the task.

10. A drive control system for multi-AGV collaboration according to claim 9, characterized in that, Based on the task start time, the target AGV is determined from a plurality of selectable AGVs, including: The target AGVs for the work tasks are matched in order of their start times, and the target AGVs matched for the work tasks are locked. If the optional AGV matched in the subsequent work task has been locked, determine whether the target AGV can be changed in the work task that locked the optional AGV; if yes, change the target AGV; otherwise, postpone the processing of the subsequent work task.

Images