Petri net-based path-execution collaborative control method for multiple mobile transfer robots

Abstract

This invention belongs to the field of flexible manufacturing systems technology and discloses a path-execution cooperative control method for multiple mobile transport robots based on Petri nets. The method includes: acquiring current system state information; determining the effective starting point and first effective target based on the current task sequence information of each mobile transport robot; updating the current position index by combining historical path cache, and reusing or re-determining discrete path sequences according to preset conditions; injecting the path sequences as spatial constraints into the Petri net execution model, and constructing an executable transition binding instance set by combining the current system state, robot execution state, and resource enabling relationships; selecting and advancing compatible transition binding instances from the instance set to form execution fragments, and updating the system state accordingly, thereby achieving rolling closed-loop cooperative control of the multiple mobile transport robot system. This invention is beneficial for improving the consistency of path planning and execution processes and the efficiency of system resource coordination.

Description

Technical Field

[0001] This invention relates to the field of flexible manufacturing systems technology, and in particular to a path-execution cooperative control method for multiple mobile handling robots based on Petri nets. Background Technology

[0002] In a flexible workshop environment, multi-mobile transport robot systems not only need to determine target stations based on transport tasks, but also need to achieve coordinated operation under the constraints of shared road networks, shared stations, and shared processing resources. Unlike static path planning scenarios, the actual operating status of mobile transport robots, workpiece processing status, loading and unloading status, and resource occupancy relationships in a workshop continuously change over time. Therefore, the control problem of multi-mobile transport robots involves not only path determination, but also state feedback during execution, resource contention analysis, and rolling update control.

[0003] In existing technologies, various control methods have been proposed for path planning problems in multi-mobile transport robot systems, especially AGV systems. For example, Okumura et al. proposed the PIBT method based on priority inheritance and backtracking in their paper "Priority inheritance with backtracking for iterative multi-agent path finding" for online path planning in multi-AGV systems. This type of method can assign the next action to multiple agents at each time step based on local priority rules, and has good real-time performance in large-scale online path coordination scenarios. However, this type of method mainly focuses on path conflict resolution itself, and does not adequately consider the connection between the path result and the manufacturing execution state, resource competition relationships, and subsequent execution control processes.

[0004] On the other hand, Petri nets, especially timed Petri nets, colored Petri nets, and timed colored Petri nets, have been used for modeling and controlling automated manufacturing systems and mobile transport robot systems. For example, Dotoli and Fanti proposed a real-time control model for AGVs based on timed colored Petri nets in their paper "Coloured timed Petri net model for real-time control of automated guided vehicle systems"; Hsieh and Kang presented a method for constructing AGV control models based on Petri nets in their paper "Developing AGVS Petri net control models from flowpath nets"; and He et al. used timed Petri nets to model the time-constrained AGV path planning problem in their paper "Path planning for automated guided vehicle systems with time constraints using timed Petri nets". In addition, existing patent literature also discloses dynamic path planning and scheduling for AGVs and multi-AGV path planning and obstacle avoidance methods based on Petri net models. For example, Chinese invention application CN110503260A discloses an AGV scheduling method based on dynamic path planning, and Chinese invention application CN114740834A discloses a multi-AGV path planning and obstacle avoidance method based on Petri net models. These solutions demonstrate that Petri net models can describe concurrent behavior, resource sharing, and temporal evolution processes in mobile transport robot systems, and can be used to model path constraints and execute control logic.

[0005] However, existing technologies still have the following shortcomings: First, path planning methods and execution control methods are usually designed separately, making it difficult to integrate path constraints, the execution state of the mobile transport robot, and resource enabling relationships into the same control framework; Second, in dynamic operation scenarios, if the entire path is replanned in each decision round, it is easy to cause redundant calculations and path instability problems, and existing solutions do not adequately consider the combination of historical path caching and reuse with on-demand replanning; Third, although existing Petri net execution schemes can describe the evolution of system state, they usually lack a closed-loop collaborative control mechanism that directly injects path sequences into the execution model and promotes the formation of execution segments through compatible transition binding.

[0006] Therefore, there is an urgent need for a path-execution cooperative control method for multiple mobile handling robots based on Petri nets, which can extract effective targets based on the current tasks to be executed provided by the upper-level task allocation module or the external scheduling module, reuse cached paths when historical paths meet the conditions, trigger replanning when necessary, and integrate path constraints, execution status and resource competition into the closed-loop control process. Summary of the Invention

[0007] The purpose of this invention is to solve the problems of separation between path planning and execution control, redundant path calculation, and difficulty in uniformly handling resource competition at the execution layer in the prior art.

[0008] The technical solution adopted by this invention to solve its technical problem is: to provide a path-execution cooperative control method for multiple mobile handling robots based on Petri nets, including the following steps:

[0009] Obtain system status information at the current decision-making moment. The system status information includes at least: the current position, execution status, resource occupancy status, and system time information of each mobile handling robot.

[0010] Based on the current task sequence information of each mobile handling robot, determine the effective starting point and first effective target of each mobile handling robot at the current decision time; for mobile handling robots that have no tasks to perform, their current position is taken as the target position;

[0011] Based on the historical path sequence and current actual position cached by each mobile handling robot at the previous decision time, update the current position index of each mobile handling robot in the historical path sequence so that it corresponds to the current actual position.

[0012] Determine whether the conditions for reusing historical paths are met; if they are met, then the historical path sequence is used; if not, then a new discrete path sequence is determined for each mobile transport robot, and the new discrete path sequence is cached.

[0013] Discrete path sequences are injected as spatial constraints into the Petri net execution model, and an executable set of transition binding instances is constructed by combining the current system state, the execution state of the mobile transport robot, and resource enabling relationships.

[0014] Select and advance compatible transition binding instances from the set of executable transition binding instances to form execution fragments, and update the system state accordingly.

[0015] Preferably, the system status information also includes: site occupancy status, loading status, unloading status, workpiece processing status, input buffer status, and shared machine resource status.

[0016] Preferably, the effective starting point is determined in the following way:

[0017] When the mobile handling robot is at a station location, the current station is determined as a valid starting point;

[0018] When the mobile transport robot is moving along a path, the station that it is about to reach along the current direction of movement will be determined as the valid starting point.

[0019] Preferably, updating the current position index of each mobile handling robot in the historical path sequence based on the historical path sequence cached at the previous decision time and the current actual position, so that it corresponds to the current actual position, includes the following steps:

[0020] Based on the current actual location, push the current location index in the historical path sequence forward to match the actual location;

[0021] If the historical path sequence still cannot be aligned with the actual location after the current location index is updated, then the historical path sequence is determined to be mismatched.

[0022] Preferably, satisfying the historical path reuse condition means satisfying at least all of the following conditions:

[0023] There are available historical path sequences;

[0024] The primary effective target of the corresponding mobile transport robot has not changed;

[0025] After the current location index is updated, the historical path sequence remains consistent with the current actual location;

[0026] The preset replanning flag was not triggered.

[0027] Preferably, determining the new discrete path sequence specifically involves: using a conflict-free path generation module to generate a new discrete path sequence for each mobile transport robot, wherein the conflict-free path generation module is used to at least handle site occupancy conflicts, reciprocal exchange conflicts, and shared target conflicts among multiple mobile transport robots.

[0028] Preferably, the step of injecting the discrete path sequence as a spatial constraint into the Petri net execution model includes:

[0029] Each mobile transport robot is restricted to performing only the movement or dwell behavior corresponding to the current path node during the current execution phase.

[0030] The enabling conditions for corresponding transition binding instances are determined based on the path sequence, the execution state of the mobile handling robot, and the resource occupancy state.

[0031] Preferably, the step of selecting and advancing mutually compatible transition binding instances from the set of executable transition binding instances specifically involves:

[0032] In the case of competition for shared machine resources, the transition binding instance to be promoted is determined based on the remaining total processing time, waiting time and preset balance disturbance term of the workpiece corresponding to the candidate transition binding instance.

[0033] Preferably, the execution segment consists of a set of mutually compatible transition binding instances selected and advanced from the current decision moment to the next decision moment. The execution segment simultaneously induces system time advancement and system state updates. Within the same execution segment, any mobile transport robot may advance a movement or dwell behavior corresponding to the current path node at most once.

[0034] Preferably, it also includes: triggering target extraction, path determination and execution control at the next decision moment based on the updated system state of the execution segment, so as to realize the rolling closed-loop collaborative control of the multi-mobile handling robot system.

[0035] The present invention has the following beneficial effects:

[0036] (1) By using a reuse mechanism based on historical path caching, repeated path planning can be avoided when conditions are met, thereby reducing computational overhead and improving path stability;

[0037] (2) By establishing a constraint relationship between the path sequence and execution control, the path planning results can directly guide the execution process, thus achieving consistency between the path and the execution.

[0038] (3) By promoting mutually compatible transition binding instances under the Petri net model, unified processing of resource contention, execution state and time evolution is achieved, thereby improving the consistency of resource conflict handling and the coordination and reliability of system execution;

[0039] (4) By rolling up the path and execution process, the multi-mobile handling robot system can achieve continuous closed-loop control in a dynamic environment.

[0040] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments, but the present invention is not limited to the embodiments. Attached Figure Description

[0041] Figure 1 This is a flowchart illustrating the method steps of the Petri net-based path-execution cooperative control method for multiple mobile handling robots according to an embodiment of the present invention.

[0042] Figure 2 This is a schematic diagram of the closed-loop process of the Petri net-based path-execution cooperative control method for multiple mobile handling robots according to an embodiment of the present invention;

[0043] Figure 3 This is a schematic diagram of the multi-AGV workshop road network and execution scenario structure in an embodiment of the present invention;

[0044] Figure 4 This is a schematic diagram illustrating the historical path cache reuse and replanning determination process in an embodiment of the present invention;

[0045] Figure 5 This is a schematic diagram of the constraint relationships in the Petri net execution model for path sequence injection in an embodiment of the present invention;

[0046] Figure 6 This is a schematic diagram of the fragment formation process in an embodiment of the present invention;

[0047] Figure 7 This is a schematic diagram illustrating the shared machine resource contention resolution process in an embodiment of the present invention. Detailed Implementation

[0048] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0049] This invention provides a path-execution cooperative control method for multiple mobile handling robots based on Petri nets, comprising the following steps:

[0050] S101, Obtain the system status information at the current decision moment. The system status information includes at least: the current position, execution status, resource occupancy status, and system time information of each mobile handling robot.

[0051] S102, based on the current task sequence information of each mobile handling robot, determine the effective starting point and first effective target of each mobile handling robot at the current decision time; for mobile handling robots that do not have any tasks to be executed, take their current position as the target position;

[0052] S103, based on the historical path sequence and current actual position cached by each mobile handling robot at the previous decision time, update the current position index of each mobile handling robot in the historical path sequence so that it corresponds to the current actual position;

[0053] S104, determine whether the historical path reuse condition is met; if it is met, then the historical path sequence is used; if it is not met, then a new discrete path sequence is determined for each mobile transport robot, and the new discrete path sequence is cached.

[0054] S105, the discrete path sequence is injected into the Petri net execution model as a spatial constraint, and combined with the current system state, the execution state of the mobile transport robot and the resource enabling relationship, an executable set of transition binding instances is constructed.

[0055] S106, Select and advance mutually compatible transition binding instances from the set of executable transition binding instances to form an execution fragment, and update the system state accordingly.

[0056] In this embodiment, a multi-AGV system in a flexible workshop is used as a specific implementation of the multi-mobile transport robot system to illustrate the present invention. AGVs are a preferred implementation of the mobile transport robot described in this invention. Other autonomous mobile robots (AMRs) and mobile transport vehicles capable of performing transportation, waiting, loading, unloading, or transfer tasks in discrete road networks can also be applied to the path-execution cooperative control framework of this invention.

[0057] In this embodiment, a flexible workshop scenario with multiple AGVs operating collaboratively is constructed. The workshop includes an input station, an output station, multiple processing devices, several service stations, robotic arm resources, and multiple automated guided vehicles (AGVs). Each AGV operates within a discrete road network within the workshop and performs behaviors such as transportation, waiting, loading, unloading, and obstacle avoidance under the constraints of shared stations, shared paths, and shared processing resources. Figure 3 The diagram illustrates the multi-AGV workshop road network and execution scenario structure used in this embodiment.

[0058] In this embodiment, the workshop transportation space is modeled using a discrete graph, denoted as:

[0059] ;

[0060] in, This represents the set of accessible stations. This represents the set of edges connecting stations; if two stations are spatially adjacent and AGVs are allowed to pass each other, then an edge is established between the corresponding stations. In any discrete execution phase, the AGV is only allowed to perform two basic behaviors: staying in place or moving to an adjacent station.

[0061] Let the collection of mobile transport robots be denoted as In this embodiment, the mobile handling robot is specifically an AGV, therefore set A can also be understood as a set of AGVs. Let the first... The system state at each decision moment is:

[0062] ;

[0063] in, Indicates the current system status information. This indicates the current system time. This includes at least the location, execution status, station occupancy status, loading status, unloading status, workpiece processing status, input buffer status, and shared machine resource occupancy status of each AGV. In other words, the... This can be understood as a set of current system state information related to path-execution collaborative control. This invention does not limit the specific underlying implementation of the Petri net execution model; the Petri net execution model can be a time-assigned Petri net, a colored Petri net, or a time-assigned colored Petri net. In this embodiment, a time-assigned colored Petri net execution model is preferred. The system state information should be able to support subsequent path cache reuse, path constraint injection, transition enablement judgment, and execution segment advancement.

[0064] The current sequence of tasks to be executed can be provided by the upper-level task allocation module, the external scheduling module, or other transportation task generation modules; the present invention does not limit its generation method. As an optional implementation, the current sequence of tasks to be executed can be generated using a distance-first rule. Specifically, at any decision moment, the upper-level task allocation module obtains the current set of transportation tasks to be served, which may include unloading tasks and picking tasks. For a mobile handling robot already loaded with goods, an unloading task can be generated based on the candidate unloading stations of the workpiece it carries; for an AGV that is empty or can continue executing tasks, a task to be served can be selected and a corresponding sequence of tasks to be executed can be generated based on the path distance between its available location and the target station of the task to be served.

[0065] In one example, if at the decision moment AGV Carrying workpiece And the unloading site is selected according to the distance priority rule. This will generate a sequence of tasks to be executed. If AGV The vehicle is currently unloaded and its current location is far from the pickup point. Recently, a sequence of tasks to be executed can be generated. If AGV If a transportation task is not assigned at the current decision-making time, it can be generated. In this embodiment, only the sequence of tasks to be executed is used as the input for path-execution coordinated control. For any AGV Record its decision-making moment The current sequence of tasks to be executed is The sequence of tasks to be executed is used to characterize the first target to be served by the AGV at the current decision moment and its subsequent targets to be executed. Let AGV be... The historical path sequence cached at the previous decision time is:

[0066] ;

[0067] in, Indicates AGV In the previous round of path planning results, the first Dispersed sites This represents the corresponding path length. Further, let the current position index after alignment of this historical path sequence be denoted as... .

[0068] The method of this invention performs path-execution coordinated control once based on the current system state at each decision-making moment, and triggers the next round of decision-making based on the updated system state after the execution segment, thereby forming a rolling closed-loop operation mechanism. In this embodiment, this process applies to the multi-AGV system. Its overall flow is as follows: Figure 2 As shown.

[0069] In step S101, the system state information at the current decision-making moment is obtained. Specifically, from the current system state... The system extracts the current position, execution status, resource occupancy status, and system time information of each AGV; it also extracts the station occupancy status, loading status, unloading status, workpiece processing status, input buffer status, and shared machine resource status. This system status information serves as input for subsequent extraction of valid starting points and the first valid target, and also as the state basis for historical path cache alignment, path reuse determination, and time-assigned colored Petri net execution control. In this embodiment, the Petri net execution control is preferably implemented using a time-assigned colored Petri net.

[0070] In step S102, based on the current task sequence information of each AGV, the effective starting point and first effective target of each AGV at the current decision moment are determined. For any AGV When the AGV is at a station location, the current station is determined as the valid starting point; when it is moving along a path, the station it is about to reach along the current direction of movement is determined as the valid starting point. At the moment of decision The effective starting point is The primary effective target is .

[0071] Furthermore, in this embodiment, the corresponding target type can also be recorded, denoted as... .when At that time, by The first task extracts the corresponding target site, and can further extract the corresponding target type; when season It can also record its target type as idle target type, which means that the AGV has no new target driving task in the current round and needs to maintain its current position to wait for subsequent decisions.

[0072] In step S103, based on the historical path sequence cached by each AGV at the previous decision time and its current actual position, the current position index of each AGV in the historical path sequence is updated to correspond to its current actual position. For any AGV If a usable historical path sequence exists Then, based on its current actual location or current valid starting point... The current position index in the historical path sequence is advanced to the path node that matches the actual position. Preferably, it can be updated as follows:

[0073] ;

[0074] If a node corresponding to the current actual position can be found in the historical path sequence after the update, the historical path sequence is considered to be aligned; if it still cannot be aligned with the current actual position after the update, the historical path sequence is determined to be mismatched, and replanning is triggered in subsequent steps. Figure 4 The process of reusing and replanning historical path caches is shown.

[0075] In step S104, it is determined whether the conditions for historical path reuse are met. For any AGV A historical path sequence can be used when the following conditions are met simultaneously: First, an available historical path sequence exists; second, the first valid target of the corresponding AGV has not changed; third, the historical path sequence is consistent with the current actual position after the current position index is updated; and fourth, a preset replanning flag has not been triggered. The preset replanning flag may include a blockage recovery flag, a mismatch recovery flag, a deadlock release flag, or other flags that require path regeneration.

[0076] If the above path reuse conditions are not met, a new discrete path sequence is determined for the corresponding AGV, and the new discrete path sequence is cached. The path sequence obtained after weighted planning is as follows:

[0077] ;

[0078] in, The effective starting point corresponding to the current decision-making moment , Corresponding to the first effective target .

[0079] The new discrete path sequence can be determined by a conflict-free path generation module. In this embodiment, the conflict-free path generation module can adopt a multi-AGV path generation method based on priority inheritance, combined with a multi-step pre-simulation and shared target processing mechanism, to generate discrete path sequences for each AGV that satisfy station occupancy constraints and opposing conflict constraints. It should be noted that the above-mentioned path generation module is only one implementation method for determining discrete path sequences in this invention. The core protection points of this invention lie in the path cache reuse and on-demand replanning mechanism, as well as the collaborative closed loop of path sequence and Petri net execution control, and are not limited to a specific path algorithm itself. In this embodiment, the Petri net execution control is preferably implemented using a time-assigned colored Petri net.

[0080] In step S105, the path sequence is used to constrain the movement behavior of the AGV, and an executable set of transition binding instances is constructed in the Petri net execution model. In this embodiment, the Petri net execution model is preferably a time-assigned colored Petri net execution model. Specifically, for any AGV... During the current execution phase, the AGV is only allowed to perform movement or dwell actions corresponding to the current node in the path sequence. In other words, if the AGV... The currently aligned path nodes are Then its permitted behaviors only include those by Move to The AGV can either maintain consistent movement along its path or wait at the current node. Therefore, the path planning results can be injected as spatial constraints into the Petri net execution model to limit the range of executable actions of the AGV.

[0081] Furthermore, in the Petri net execution model, based on the current system state, AGV execution state, path constraints, and resource enabling relationships, a set of executable transition binding instances is constructed, denoted as... .

[0082] A transition binding instance is included in the executable set only if it simultaneously meets the following conditions: First, the corresponding input library has an identifier that meets the conditions; second, the required resources are not occupied by other execution behaviors; third, if it is an AGV movement type transition, its movement direction is consistent with the current path sequence constraint; fourth, when the Petri net execution model is a timed Petri net or a timed colored Petri net, the corresponding time enablement condition is also met. Figure 5 The constraints of the path sequence injection Petri net execution model are shown.

[0083] In step S106, the executable transition binding instance set is... Select and advance compatible transition binding instances to form an execution fragment from the current decision moment to the next decision moment. And update the system status accordingly.

[0084] In this embodiment, the execution fragment From the executable transition binding instance collection The selected set of mutually compatible transition binding instances can be represented as:

[0085] ;

[0086] in, Indicates the first Each transition binding instance is pushed forward. Mutual compatibility means that corresponding transition binding instances will not contend for the same exclusive resource within the same execution segment, will not cause conflicting actions within the same AGV, and will not disrupt the consistency of current path constraints and state evolution. Furthermore, within the same execution segment, for any AGV, at most one movement or dwell action corresponding to the current path node will be executed.

[0087] Execution fragment Make the system state change Advance to , can be represented as:

[0088] ;

[0089] Among them, execution fragments Not only induce system state information Evolved to It also synchronously induces the system time to change from Advance to .

[0090] Furthermore, in the case of shared machine resource contention, if multiple candidate transition binding instances simultaneously request the same shared machine resource, the instance is determined to be included in the current execution segment based on the remaining total processing time, waiting time, and preset balancing disturbance term of the workpiece corresponding to the candidate transition binding instance. The transition binding instance. Preferably, the following priority criterion can be constructed:

[0091] ;

[0092] in, Indicates that the candidate transition is bound to an instance. This indicates the total remaining processing time of the workpiece corresponding to the instance bound to this transition at the current moment. This indicates the cumulative waiting time of the workpiece at the current moment. This indicates a preset balance perturbation term. Multiple candidate transitions are bound to instances according to... Perform lexicographical comparison and select first. Smaller Transitions with larger and balanced perturbations that satisfy preset rules are bound to instances for advancement. This rule is consistent with the overall idea of ​​"selecting compatible instances from executable instances to form execution fragments based on the current system state, path constraints, and resource usage." Figure 7 The process of resolving shared machine resource contention is illustrated.

[0093] Figure 6 The process of forming an execution fragment is illustrated. Specifically, at the current decision moment, based on path constraints and resource enablement relationships, a set of executable transition binding instances is first constructed. Then, compatible instances are selected to form execution fragments. After advancing the execution segment, update the system state and time information to obtain the state at the next decision moment. . Figure 6 The execution segment in the text does not mean that the same mobile handling robot performs multiple path advancement actions consecutively within the same decision segment; in this embodiment, it does not mean that the same AGV performs multiple path advancement actions consecutively.

[0094] In this embodiment, the current system state information is reacquired at each decision-making moment, and steps S2 to S6 are repeated. That is, after each execution segment is completed, the system adjusts its state according to the new state. The system re-extracts valid starting points and first valid targets, aligns historical path caches, determines path reuse or replanning, injects path constraints, and generates execution fragments, thereby forming a rolling closed-loop collaborative control mechanism that unifies and couples path determination and execution control.

[0095] Through the above implementation methods, the present invention can integrate path sequence constraints, Petri net execution control, resource contention analysis, and system time progression into a unified closed-loop framework in a dynamic manufacturing environment. In this embodiment, the Petri net execution control is preferably implemented using time-assigned colored Petri nets. On the one hand, by using historical path caching and reuse mechanisms and on-demand replanning, unnecessary repeated path calculations can be reduced, improving path stability. On the other hand, by promoting mutually compatible transition binding instances in the Petri net execution model, unified coordination and control of concurrent operation of multiple AGVs, loading and unloading execution, resource contention, and time evolution can be achieved, thereby improving the execution consistency, resource coordination efficiency, and overall operational reliability of the multi-mobile handling robot system. In this embodiment, this is specifically reflected in improving the operational reliability of the multi-AGV system.

[0096] In this embodiment, the workshop transportation space is discretized into 120 passable stations. The input and output stations are located at S61 and S74, respectively. Nine processing devices are arranged at S101, S109, S117, S78, S86, S94, S4, S12, and S20. The AGV takes 2 seconds to move between adjacent stations, and the loading and unloading times are both 4 seconds. The workpiece processing time is preset according to different workpiece types and processes, and each process can only select one device from the candidate machine set for processing. Preferably, the historical path reuse determination satisfies the following conditions: the first valid target has not changed, the current position index is aligned, and no replanning flag is triggered. When target changes, path mismatches, blockage recovery, or other replanning situations occur, a discrete path sequence is regenerated. Preferably, during shared machine resource competition, the transition binding instance to be advanced is selected according to the priority order of remaining total processing time, waiting time, and balancing disturbance items to ensure compatibility and system stability within the execution segment.

[0097] To further verify the application effect of the present invention in the above-mentioned multi-AGV path-execution collaborative control scenario, the present invention will be further illustrated below with a set of simulation verification examples and comparative examples. It should be noted that the following simulation verification examples and comparative examples are only used to illustrate the effectiveness of the present invention and are not intended to limit the scope of protection of the present invention.

[0098] In a simulation verification example, the following method is used. Figure 3 The diagram shows a flexible workshop with multiple AGVs operating in operation. The system includes 120 accessible stations, 1 input station, 1 output station, 9 processing devices, and 11 robotic arms. The AGV service stations corresponding to the input and output stations are as follows: and The nine processing machines are respectively labeled as... Their corresponding service sites are respectively The AGV takes 2 seconds to complete one movement between adjacent stations, and the loading and unloading operations each take 4 seconds. The process routes, candidate machine sets, and corresponding processing time settings for the three types of workpieces are shown in Table 1.

[0099] Table 1 - Production process steps for three types of workpieces:

[0100] To establish a fixed verification scenario, in this simulation verification example, the initial quantities of three workpiece types 1, 2, and 3 are set to 12, 12, and 12 respectively, totaling 36 workpieces. Initially, all workpieces are located at the input station awaiting entry into the first process. Furthermore, the number of AGVs is set to 8, each with a carrying capacity of 1, all initially in an unloaded state. The initial positions of the AGVs are determined by the location of the attached... Figure 3The scenario shown is given. The system starts from the initial state and runs in a rolling fashion until all workpieces have completed all processing steps and are transported to the output station.

[0101] In this simulation verification example, to ensure the reproducibility of the verification process, the main parameters related to path-execution collaborative control are further set as follows: In the path cache reuse determination, when the first valid target of the AGV has not changed, the historical path is still consistent with the current actual position after being updated by the current position index, and no replanning flag is triggered, the historical path sequence is used; when the first valid target of the AGV changes, the historical path and the current position are mismatched, a blockage recovery requirement occurs, or other replanning flags are triggered, a new discrete path sequence is regenerated. The path generation module adopts a multi-AGV path generation method based on priority inheritance, and the planning mode adopts a multi-step pre-simulation mode. For the scheme of this invention, the path sequence is used as a hard constraint condition for the AGV movement behavior in the execution layer, that is, the AGV is only allowed to perform forward or stop actions according to the corresponding node of the current path sequence.

[0102] Furthermore, in the Petri net execution control process of the present invention, a time-assigned colored Petri net execution model is preferably adopted; for transition binding instances that meet the conditions identified by the input library, have no resource conflicts, consistent path constraints, and valid time enable conditions, they are included in the candidate executable set. In the case of shared machine resource contention, the remaining total processing time of the workpiece corresponding to the candidate transition binding instance is used as the basis. Waiting time and balance disturbance term Priority comparisons are performed to determine the transition binding instances to be advanced. Preferably, selection is made according to the principle of "prioritizing those with shorter remaining total processing time, longer waiting time, and those that meet the balancing perturbation rules," thereby forming mutually compatible execution fragments.

[0103] Under the same scenario and parameter conditions described above, the following comparison example is set up.

[0104] Comparative Example 1: Except for not reusing the historical path cache, the system layout, workpiece configuration, number of AGVs, path constraint rules, execution control rules, and parameter settings are all the same as the present invention. Specifically, at each decision moment, a discrete path sequence is regenerated based on the current valid starting point and the first valid target of each AGV, without determining whether the historical path cached at the previous decision moment can continue to be used.

[0105] Comparative Example 2: Except for not using the path sequence as a hard constraint on the execution layer's movement behavior, the rest of the system layout, workpiece configuration, AGV quantity, path cache update rules, path generation rules, and parameter settings are the same as the solution of this invention. Specifically, after obtaining the discrete path sequence, the execution layer only uses the path sequence as a preference reference for action selection, and does not force the AGV to strictly follow the currently recommended path nodes to perform movement behavior.

[0106] In this simulation verification example, the following metrics were used to compare the various schemes: average decision time, path cache reuse rate, AGV cumulative waiting steps, and system completion time. Average decision time reflects the average computational cost of path-execution collaborative control in a single decision round; path cache reuse rate reflects the degree to which the historical path caching mechanism is effectively utilized; AGV cumulative waiting steps reflect the cumulative dwell time of the AGV during rolling operation due to path coordination, execution waiting, or resource occupation; and system completion time reflects the overall completion efficiency of all workpieces being processed and transported to the output station. The statistical results after each scheme was run under the same initial conditions are shown in Table 2.

[0107] Table 2 - Comparison of the execution results of different paths and coordinated control schemes:

[0108] As shown in Table 2, compared with Comparative Example 1, the present invention, after introducing historical path cache reuse and on-demand replanning mechanisms, reduces the average decision time from 24.042ms to 22.360ms, increases the path cache reuse rate from 0 to 0.6124, reduces the cumulative waiting steps of the AGV from 1321 steps to 1009 steps, and shortens the system completion time from 892s to 886s. This indicates that in dynamic operation scenarios, reusing historical paths that meet the conditions can effectively reduce repeated path calculations, reduce the computational overhead in rolling control, and reduce the additional coordination waiting caused by frequent replanning, thereby improving the overall system execution efficiency.

[0109] Furthermore, as shown in Table 2, compared with Comparative Example 2, the average decision time of the present invention decreased from 26.782ms to 22.360ms, the path cache reuse rate increased from 0.4496 to 0.6124, the cumulative waiting steps of the AGV decreased from 1132 steps to 1009 steps, and the system completion time was significantly shortened from 1028s to 886s. This indicates that by using the path sequence as a hard constraint on the movement behavior of the execution layer, and by uniformly constructing a set of executable transition binding instances under the Petri net model, preferably a time-assigned colored Petri net model, and promoting mutually compatible transition binding instances, the consistency between the path planning results and the execution process can be enhanced, reducing execution deviations, repeated movements, and the resulting subsequent coordination costs, thereby significantly improving the overall operating performance of the multi-mobile transport robot system; in this simulation verification example, this is specifically reflected in the improved performance of the multi-AGV system.

[0110] Therefore, based on the above simulation verification results, it can be shown that the present invention reduces repeated path calculations and improves path stability through historical path caching and reuse and on-demand replanning mechanisms. At the same time, by incorporating path constraints, execution feedback and resource competition into the closed-loop control process under the Petri net model, preferably the time-assigned colored Petri net model, the execution consistency, resource coordination efficiency and overall operating performance of the multi-mobile handling robot system in the dynamic manufacturing environment can be improved. In this simulation verification example, this is specifically reflected in the improvement of the performance of the multi-AGV system.

[0111] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A path-execution cooperative control method for multiple mobile transport robots based on Petri nets, characterized in that, Includes the following steps: Obtain system status information at the current decision-making moment. The system status information includes at least: the current position, execution status, resource occupancy status, and system time information of each mobile handling robot. Based on the current task sequence information of each mobile handling robot, determine the effective starting point and first effective target of each mobile handling robot at the current decision time; for mobile handling robots that have no tasks to perform, their current position is taken as the target position; Based on the historical path sequence and current actual position cached by each mobile handling robot at the previous decision time, update the current position index of each mobile handling robot in the historical path sequence so that it corresponds to the current actual position. Determine whether the conditions for reusing historical paths are met; if they are met, then the historical path sequence is used; if not, then a new discrete path sequence is determined for each mobile transport robot, and the new discrete path sequence is cached. Discrete path sequences are injected as spatial constraints into the Petri net execution model, and an executable set of transition binding instances is constructed by combining the current system state, the execution state of the mobile transport robot, and resource enabling relationships. Select and advance compatible transition binding instances from the set of executable transition binding instances to form execution fragments, and update the system state accordingly.

2. The path-execution cooperative control method for multiple mobile transport robots based on Petri nets according to claim 1, characterized in that, The system status information also includes: site occupancy status, loading status, unloading status, workpiece processing status, input buffer status, and shared machine resource status.

3. The path-execution cooperative control method for multiple mobile transport robots based on Petri nets according to claim 1, characterized in that, The valid starting point is determined in the following way: When the mobile handling robot is at a station location, the current station is determined as a valid starting point; When the mobile transport robot is moving along a path, the station that it is about to reach along the current direction of movement will be determined as the valid starting point.

4. The path-execution cooperative control method for multiple mobile transport robots based on Petri nets according to claim 1, characterized in that, The step of updating the current position index of each mobile handling robot in the historical path sequence, based on the historical path sequence cached by each mobile handling robot at the previous decision time and its current actual position, to make it correspond to the current actual position, includes the following steps: Based on the current actual location, push the current location index in the historical path sequence forward to match the actual location; If the historical path sequence still cannot be aligned with the actual location after the current location index is updated, then the historical path sequence is determined to be mismatched.

5. The path-execution cooperative control method for multiple mobile transport robots based on Petri nets according to claim 1, characterized in that, The condition for satisfying historical path reuse means that at least all of the following conditions must be met: There are available historical path sequences; The primary effective target of the corresponding mobile transport robot has not changed; After the current location index is updated, the historical path sequence remains consistent with the current actual location; The preset replanning flag was not triggered.

6. The path-execution cooperative control method for multiple mobile transport robots based on Petri nets according to claim 1, characterized in that, The determination of the new discrete path sequence specifically involves: using a conflict-free path generation module to generate a new discrete path sequence for each mobile transport robot. The conflict-free path generation module is used to handle at least the site occupancy conflict, the reciprocal exchange conflict, and the shared target conflict among multiple mobile transport robots.

7. The path-execution cooperative control method for multiple mobile transport robots based on Petri nets according to claim 1, characterized in that, The discrete path sequence is injected as a spatial constraint into the Petri net execution model, and the spatial constraints include: Each mobile transport robot is restricted to performing only the movement or dwell behavior corresponding to the current path node during the current execution phase. The enabling conditions for corresponding transition binding instances are determined based on the path sequence, the execution state of the mobile handling robot, and the resource occupancy state.

8. The path-execution cooperative control method for multiple mobile transport robots based on Petri nets according to claim 1, characterized in that, The step of selecting and advancing mutually compatible transition binding instances from the set of executable transition binding instances specifically involves: In the case of competition for shared machine resources, the transition binding instance to be promoted is determined based on the remaining total processing time, waiting time and preset balance disturbance term of the workpiece corresponding to the candidate transition binding instance.

9. The path-execution cooperative control method for multiple mobile transport robots based on Petri nets according to claim 1, characterized in that, The execution segment consists of a set of mutually compatible transition binding instances selected and advanced from the current decision moment to the next decision moment. The execution segment simultaneously induces system time advancement and system state updates. Within the same execution segment, any mobile transport robot can advance a movement or dwell behavior corresponding to the current path node at most once.

10. The path-execution cooperative control method for multiple mobile transport robots based on Petri nets according to claim 1, characterized in that, It also includes: target extraction, path determination and execution control at the next decision moment triggered by the updated system state of the execution segment, so as to realize the rolling closed-loop collaborative control of the multi-mobile handling robot system.

Images