A workshop integrated dynamic scheduling method considering path planning and disturbance response

Abstract

This invention discloses a workshop integrated dynamic scheduling method considering path planning and disturbance response. Specifically, it involves: First, collision-free path planning based on information interaction and decision-making between AGV agents and node agents, including AGV path selection based on node congestion and research on resolution strategies for three common types of conflicts during operation; Second, an integrated scheduling hybrid simulation model under path planning and disturbance response strategies is developed. Renegotiation response strategies are designed for disturbance events including emergency order insertion, AGV failure, and machine tool failure. This results in a flexible production workshop machine tool and AGV dynamic integrated scheduling system based on the hybrid simulation model, incorporating collision-free path planning and disturbance response strategies. This invention significantly reduces the number of collisions between AGVs, demonstrating good collision avoidance performance. In the event of equipment failure or emergency order insertion, the robustness of the entire system exhibits minimal fluctuation, and the system demonstrates good disturbance response capability.

Description

Technical Field

[0001] This invention belongs to the field of workshop production scheduling applications, and in particular relates to a workshop integrated dynamic scheduling method that considers path planning and disturbance response. Background Technology

[0002] In flexible production workshops using multiple AGVs for material feeding, conflicts between AGVs are prone to occur, as are dynamic emergencies such as order insertions and equipment failures. These can disrupt the original scheduling plan and even affect the normal operation of the entire manufacturing system. Therefore, when developing a scheduling plan, it is necessary to consider AGV path planning and conflict resolution, as well as dynamic disturbance event response strategies, to perform dynamic integrated scheduling of multiple machine tools and multiple AGVs. The scheduling results will be more in line with actual production conditions.

[0003] Traditional integrated scheduling of machine tools and AGVs in flexible manufacturing workshops is primarily a static integrated scheduling problem involving AGV transport time. The integrated AGV transport time is the shortest time under static conditions and a fixed single path, without considering dynamic events such as AGV path selection, conflict resolution, and interference from emergency order insertions or equipment failures. Current methods mainly include intelligent optimization algorithms and simulation methods. For example, Saidi et al., when establishing an integrated scheduling model, considered conflict-free AGV path selection and proposed a two-stage ant colony algorithm to solve the integrated scheduling problem of machine tools and AGVs in the job shop. The first stage allocates AGVs for transport tasks, and the second stage selects conflict-free paths for the AGVs. They also studied the impact of the number of AGVs on the maximum completion time. However, they did not consider the situation where multiple machine tools are available for processing each workpiece in the workshop; that is, the machine tool for each process is unique. He Changzheng et al. proposed a hybrid genetic algorithm based on time windows and Dijkstra's algorithm. To integrate machine tool and AGV scheduling, they designed a three-chain encoding structure and crossover and mutation operators for the AGV encoding chain. Furthermore, in the decoding operation of the genetic algorithm, they combined Dijkstra's algorithm with the time window principle to accurately plan a collision-free and conflict-free shortest path for the task vehicle. However, they did not consider the occupancy conflict problem when the vehicle was stationary in the AGV path planning. Deng Xi et al. proposed a hybrid genetic algorithm based on time schedules and the A* (A-Star) algorithm. First, a three-segment encoding structure was used to obtain the scheduling results of the machine tool and AGV. Second, the A* algorithm, which includes a conflict resolution strategy, was used to obtain the AGV path planning results. These results were then integrated into the decoding stage to finally obtain the start and end times of processing and transportation for each process, forming the final scheduling scheme.

[0004] Simulation methods include Discrete Event Modeling and Simulation (DEMS) and Agent-Based Modeling and Simulation (ABMS). DEMS abstracts the real system into a series of events, driving the simulation model by controlling the order of these events. Due to its speed and accuracy in modeling, many scholars have used this method to solve shop floor scheduling problems. For example, Gao Keting used DEMS to recreate the shop floor operation process for the hybrid flow shop scheduling problem and combined it with a genetic algorithm to optimize the machine tool scheduling scheme, reducing the system runtime. Multi-Agent technology is a new method developed from the field of artificial intelligence. Using ABMS, it can well reflect intelligent behavior in the shop floor. ABMS is considered the method with the greatest simulation potential, application prospects, and practical value, and has been well applied to the integrated scheduling problem of flexible manufacturing shops. For example, Komma et al. formulated the machine tool and AGV scheduling problem in a flexible manufacturing system as a multi-Agent system, and thus implemented agent-based shop floor simulation based on the Java Agent Development Framework (JADE). Erol et al. proposed a multi-agent method for machine tool and AGV scheduling in manufacturing systems, consisting of a management agent, an AGV agent, a task agent, and a machine tool agent. These agents operate in real-time and generate feasible schedules through negotiation mechanisms. However, the machine tools studied lack flexibility. While single simulation methods can solve some problems in production systems, they cannot fully reflect all system characteristics. Hybrid simulation using two or more methods from DEMS, ABMS, and SD can compensate for the shortcomings of single simulation methods and more completely represent the characteristics and essence of manufacturing systems. Rodrigue et al. applied a hybrid simulation method combining DEMS and ABMS to the production scheduling problem in a flexible work shop, realizing the allocation and sequencing of process tasks on machine tools, but did not consider the impact of AGV scheduling.

[0005] In summary, the shortcomings of relevant scholars in solving the problem include: (1) Most studies have only solved part of the problems of machine tool selection, AGV selection, and the sequencing of process tasks on machine tools and AGVs. There is still little research on the integrated scheduling optimization problem of AGV path planning. When there are multiple AGVs in the actual factory, path selection and conflict problems are difficult to avoid. Different path selections and conflict resolution strategies will inevitably cause different transportation times, thus affecting the scheduling results. (2) Most studies rarely consider the impact of dynamic emergencies such as order insertion and equipment failure in the manufacturing system. The generation of other dynamic events in the system will also disrupt the original production scheduling plan. (3) A single simulation method is insufficient to describe the operation process of AGVs and machine tools in the complex flexible production workshop. It is necessary to use a hybrid simulation of multiple simulations to describe its operation essence. At present, research on hybrid simulation in this problem is far from sufficient.

[0006] Therefore, considering the actual situation, the production scheduling of flexible production workshops urgently needs to consider multi-AGV collision-free path planning and AGV and machine tool integrated scheduling methods that respond to dynamic disturbance events. Summary of the Invention

[0007] When multiple AGVs exist in a real factory, path selection and conflict issues are unavoidable. Different path choices and conflict resolution strategies will inevitably result in different transportation times, thus affecting the scheduling results. At the same time, the occurrence of other dynamic events in the system can also disrupt the original scheduling plan. To address this, this invention provides a workshop-integrated dynamic scheduling method that considers path planning and disturbance response.

[0008] The present invention provides a workshop integrated dynamic scheduling method that considers path planning and disturbance response, specifically as follows:

[0009] A. Collision-free path planning based on information interaction and decision-making between AGV Agent and node Agent.

[0010] When there are pending transport tasks in the AGV Agent task buffer pool, the task is submitted to the path planning module. After comparing the current node position information of the AGV and the transport task, path planning is performed. That is, when entering the monitoring range of the node Agent, the AGV interacts with information by calling the node Agent interface to initially determine the next travel node. It also detects and resolves conflicts such as head-on collisions and contention for the same node. Finally, it determines the node with no collisions and the shortest distance as the actual next travel node. The AGV then travels to the next travel node, and during the journey, it detects and resolves chasing conflicts. The above process is repeated until the target node is reached.

[0011] (1) Plan the shortest collision-free path for AGV based on the interaction between node Agent and AGV Agent.

[0012] 1) Node congestion calculation:

[0013] Each AGV in the system is considered as an ant agent, carrying a certain pheromone λ when entering the road network. Real-time control of the AGVs employs a real-time global update strategy, updating the congestion level of each node based on the positions of all AGVs within the transportation network at each time step. At time t, the congestion level at any node (i,j) within the transportation network is as follows:

[0014]

[0015] In the formula, K is the number of AGVs arriving at the AGV set from the node in the transportation network at time t; M is the number of AGVs expected to arrive at the AGV set from the node in the transportation network at time t; and N is the number of AGVs departing from the AGV set from the node in the transportation network at time t. Let α, β, and ε be the straight-line distances between the kth AGV arriving in the AGV set at time t, the mth AGV scheduled to arrive in the AGV set, and the nth AGV departing from the AGV set, respectively, and the node; α, β, and ε are importance coefficients, and α > β > ε.

[0016] 2) Next travel node selection:

[0017] When the AGV reaches a new node position (i,j), it needs to select the next travel node from among many nodes. This process is repeated, and the next travel node is continuously updated until the target node position is reached, at which point the path planning ends. The set of the next travel nodes after node (i,j) is (I',J')={(i',j')|i'≠i orj'≠j}, where Let (i,j) represent the k-th AGV. The higher the congestion at node (i',j'), the greater the likelihood of collisions and conflicts between AGVs at that node. Therefore, it's necessary to avoid overly congested nodes. Based on the node conditions, nodes with lower collision probabilities can be further divided from (i',j') to form [a more efficient network]. The set of drivable nodes, denoted as allowed. k ;like illustrate There are no suitable drivable nodes. Parking and waiting; if but Select the next driving node from it.

[0018] Introducing a crowding threshold q to divide allowed k If τ (i',j') If (t) < q, then the probability of a collision occurring with (i',j') as the next travel node is relatively small and will not affect the normal operation of the AGV system. (i',j') ∈ allowed k If τ(i',j') If (t)≥q, then the probability of a collision occurring at (i',j') as the next travel node is relatively high, which will affect the normal operation of the system.

[0019] From allowed k The next driving node determined in the process needs to satisfy two conditions: minimizing the probability of collision and minimizing the path. These conditions are respectively limited by the attractiveness and distance heuristic functions of the candidate nodes. The attractiveness, i.e., the congestion threshold q, is related to the congestion τ at node (i',j') at time t. (i',j') The difference between (t) and the attraction is calculated using the following formula:

[0020]

[0021] The shortest path requires ensuring that the next node (i', j') is within the range of the AGV's current node (i, j) and the target point (e). x ,e y The shortest distance between the two classes is found; therefore, a heuristic function for the two-class distance is established. and These represent the visibility between the current node (i,j) of the AGV and the next node (i',j'), and the visibility between the next node (i',j') and the target point (e), respectively. x ,e y The visibility between two types of distances is calculated using the following formulas for the two types of distance heuristic functions:

[0022]

[0023]

[0024] Therefore, the transition probability of the AGV moving to each node in the set of drivable nodes is:

[0025]

[0026] Among them, σ, δ, and γ are the importance coefficients of attraction and the two types of visibility.

[0027] Ultimately, the node with the highest probability of node transfer is selected as the next travel node for the AGV.

[0028] (2) Conflict detection and resolution during driving.

[0029] 1) Common Collision and Conflict Types

[0030] Collisions or deadlocks in AGVs during actual operation can be categorized into the following three types:

[0031] ① Heading-to-head conflict: At the same time, two or more AGVs travel towards each other from different nodes to the same road segment, or there are AGVs traveling towards each other or parked AGVs on the next travel segment of a certain node.

[0032] ②Conflict over the same node: When two or more AGVs depart from different nodes or road sections or compete for the right to drive at the same node.

[0033] ③ Chase-and-go collision: A situation where two AGVs traveling in the same direction on a certain road segment collide because the latter is traveling faster than the former, or the former stops in the latter's only path.

[0034] 2) Conflict detection mechanism

[0035] Conflict detection between AGVs is achieved through information exchange between the node Agent and the AGV Agent. Different detection methods are set for different conflict types, as follows:

[0036] ① Heading-to-head conflict: When an AGV enters the monitoring range of a certain node, the corresponding AGV Agent makes a preliminary decision on its next travel node by the next travel node decision submodule in the path planning module and submits it to the node Agent that it has arrived at at this moment; then the node Agent traverses its set of pre-arrived vehicles and set of arrived vehicles to search for whether there are any vehicles traveling on the road segment between the current node and the next travel node of the AGV. If there are, then the vehicle must be a heading-to-head conflict vehicle.

[0037] ② Conflicts over the same node: When an AGV enters the monitoring range of a certain node, the corresponding AGV Agent, after ensuring that there are no conflicts over the next driving node by the opposing conflict detection and resolution submodule, enters the same node conflict detection and resolution submodule. This module submits the AGV information to the next driving node Agent. The node Agent traverses its pre-arrival vehicle set and arrival vehicle set to search for whether there are conflicting vehicles competing with it for the right to drive at the same node. The existence of such vehicles is determined by equations (6) to (7).

[0038]

[0039]

[0040] In the formula, The estimated arrival time for an AGV that is currently making a conflict judgment at the same node to reach the next travel node; is the estimated arrival time of the conflicting vehicle to the next travel node; r is the monitoring radius of the node; v m The speed at which the collision vehicle travels; This indicates that the conflicting vehicle currently occupies the conflict node. The car involved in the conflict stopped at this point.

[0041] ③ Chasing and Conflict Detection: This is achieved mainly through the AGV Agent's own distance monitoring module, which monitors the distance between two vehicles traveling in the same direction on the same road segment in real time.

[0042] 3) Conflict resolution mechanism

[0043] ① Resolving conflicts in opposite directions

[0044] a) Remove the next driving node with conflicting directions from the set of feasible nodes. If the set is empty after removal, proceed to step e. Otherwise, continue to select the node with the highest transfer probability from the set as the new next driving node and proceed to step b.

[0045] b. Based on formula (8), determine the time loss of the vehicle taking the route-changing driving strategy compared to the stationary waiting strategy. If there is no time loss, directly take the newly decided driving node as the next driving node and execute step d to make a judgment; otherwise, proceed to step c.

[0046]

[0047] In the formula: tTn is the time it takes for the vehicle to reach the replacement node from its current position; nTg is the shortest time it takes for the vehicle to reach the target node via the replacement node; tTo is the time it takes for the vehicle to reach the conflict node from its current position; oTg is the shortest time it takes for the vehicle to reach the target node from the conflict node; and wt is the parking waiting time.

[0048] c. Changing routes incurs time loss. Based on the road conditions at the current node where the AGV is located, strategies such as stopping in place, changing sides, or forced route change will be considered. The road conditions corresponding to different strategies are as follows:

[0049] Stop in place: The AGV Agent calls the conflict vehicle one-step prediction interface of the Agent of the current node. The encapsulated function in it predicts the next driving node after the conflict vehicle arrives at this node. After obtaining the prediction result, the AGV continues to determine if the predicted node is not in the direction of the current road segment. If so, it is determined that the conflict vehicle will not collide with the AGV after it arrives at this node, and the AGV will implement the stop in place waiting strategy.

[0050] Side-switching parking: If the predicted node is in the direction of the current road segment of this vehicle, and it is determined that the conflicting small vehicles will collide after reaching the node of this vehicle, then the AGV Agent will interact with the Agent of the current node to see if there is an empty road segment near this node. If there is, the AGV will drive to the empty road segment and stop to wait, which is called side-switching parking.

[0051] Forced route change: If none of the above conditions are met, the AGV must change routes, and the node determined in step a will be the next travel node.

[0052] d. If the AGV changes route, it is necessary to continue to detect the node-to-node conflict on the road segment where the new node is located. If the conflict still exists, repeat steps a to c. Otherwise, the AGV will move to the new node.

[0053] e. When the current car has no way to change lanes or can not stop and wait, a collision is inevitable. In this case, it is necessary to interact with the conflicting car to make it reverse and give way, so that the current car can occupy the right of way in the conflicting section.

[0054] ② Resolution of conflicts involving the same node

[0055] The same node contention conflict is identified and resolved by the same node contention conflict detection and resolution module in the AGV Agent. After such a conflict is detected, the conflict resolution strategy encapsulated in this module is mainly implemented by comparing the priority of the vehicle and the conflicting vehicle. The vehicle with higher priority obtains the right to drive at the conflicting node, while the vehicle with lower priority changes routes. If there is no alternative node, the vehicle stops and waits.

[0056] The priority is determined through the following steps:

[0057] a. Vehicles that are in a conflicting situation at this moment or at the previous driving node should have a higher priority to avoid collisions with vehicles in the conflicting direction.

[0058] b. Compare the congestion levels of the nodes where the conflicting AGVs are currently located. The AGVs with higher congestion levels are given higher priority so that they can leave the node as soon as possible and reduce the congestion level of the node.

[0059] c. If the current nodes have the same level of congestion, compare which AGV is more attractive to the conflicting nodes, and the one with the higher attraction will be given a higher priority.

[0060] ③ Catching up and resolving conflicts

[0061] When the AGV Agent's distance monitoring module detects that the distance between the two workshops is less than the safe distance, the chasing vehicle will stop and wait. The chasing vehicle will only continue to move when the distance between the two workshops reaches the safe distance.

[0062] B. Consider disturbance events, formulate response strategies, and form a hybrid simulation model.

[0063] The disturbances here include emergency order insertion, AGV malfunction, and machine tool malfunction.

[0064] (1) Emergency order insertion

[0065] When an urgent order arrives, the system agent generates a corresponding workpiece agent and process agent based on the order information and registers their status as urgent. Accordingly, when the workpiece agent enters the production system from the "order entry" object in the production process sub-model, the urgent workpiece is identified and given the highest priority, allowing it to be negotiated first. When entering the contract negotiation stage of the "negotiation" process, the urgent process agent marks the tender as urgent during the bidding process. Based on this, when calculating the bid content, the AGV and machine tool agents postpone all unexecuted tasks in the buffer pool and bid based on the earliest start time of handling. After winning the bid, the urgent task is inserted in front of all non-urgent tasks in the buffer pool, allowing it to be executed first.

[0066] (2) AGV malfunction

[0067] When an AGV malfunctions, all tasks in the AGV task buffer pool are renegotiationd. At this time, the tasks in the buffer pool have two states: in progress and waiting to be moved. In progress is further divided into two states: empty and loaded. When a malfunction occurs, tasks in the empty or waiting state only need to release the corresponding workpiece Agent from the "Waiting and Executing Transportation" module in the production process sub-model to the "Negotiation" module for renegotiation of moving the AGV and the processing machine tool. The workpiece position does not change. In the loaded state, since the workpiece is moving in the road network with the AGV, when the AGV malfunctions, the carried workpiece will be stuck at the node position closest to the AGV at this moment. During renegotiation, the distance information is calculated based on the actual stuck position of the workpiece.

[0068] Regardless of the status of the tasks in the buffer pool, after the "Waiting and Executing Transportation" module releases them, they must flow into the "Judging AGV Fault" object. This module will then divert the workpieces that need to be renegotiation. Workpieces that need to be renegotiation are returned to the "Negotiation" module for renegotiation; otherwise, they flow into the processing module for processing. In addition, before all task renegotiation begins, the task buffer pool of the faulty AGV must be cleared, and the processing tasks of all renegotiation workpieces must be deleted from the task buffer pool of their respective machine tools. After that, the faulty AGV will participate in the bidding again based on the estimated fault end time.

[0069] (3) Machine tool failure

[0070] The response strategy for machine tool failure is to renegotiate all pending tasks in the task buffer pool of the failed machine tool. At this time, the tasks in the buffer pool have four states: processing, waiting for processing in the buffer station, being transported, and waiting for transport. Workpieces being processed are not renegotiated and will continue processing after the failure ends. Tasks waiting for processing and waiting for transport in the buffer station need to be released from the "processing buffer" and "waiting and executing transportation" modules of the production process sub-model to the "negotiation" module for renegotiation of the transport AGV and the processing machine tool. Tasks in the transport state only need to be renegotied for processing machine tools. After obtaining the new processing machine tool position, the path planning module of the contracted AGV can directly change the destination to the new position.

[0071] Workpieces released from the "Processing Buffer" and "Waiting and Executing Transport" modules flow into the "Judgment AGV Fault" and "Judgment Machine Tool Fault" objects, respectively. These modules then divert workpieces requiring renegotiation, returning them to the "Negotiation" module for further negotiation. Otherwise, the workpieces flow into the processing module for continued processing. Similarly, before any renegotiation task begins, all pending tasks in the faulty machine tool's task buffer pool must be cleared, and the transport tasks of all renegotiation workpieces waiting to be transported must be removed from their respective transport AGV's task buffer pool. Afterward, the faulty machine tool participates in the bidding process based on its estimated fault end time. If there are workpieces currently being processed, the estimated end time of processing for those workpieces after the fault ends will be used for bidding.

[0072] The beneficial technical effects of this invention are as follows:

[0073] In the event of equipment failure, this invention improves both system completion time and robustness under a renegotiation response strategy. Under the emergency order insertion strategy, the completion time of emergency workpieces is shortened, and the overall system robustness exhibits minimal fluctuation, averaging only 5%, demonstrating good system disturbance response capability and proving the effectiveness of the disturbance event response strategy presented in this chapter. Attached Figure Description

[0074] Figure 1 This is a flowchart of the collision-free path planning execution process of the present invention.

[0075] Figure 2 This diagram illustrates common conflict types in this invention.

[0076] Figure 3 This diagram illustrates strategies for resolving conflicting situations.

[0077] Figure 4 This invention considers an integrated scheduling hybrid simulation model for disturbance response. Detailed Implementation

[0078] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0079] The present invention provides a workshop integrated dynamic scheduling method that considers path planning and disturbance response, specifically as follows:

[0080] A. Collision-free path planning based on information interaction and decision-making between AGV Agent and node Agent, with a specific flowchart as follows: Figure 1 As shown.

[0081] When there are pending transport tasks in the AGV Agent task buffer pool, the task is submitted to the path planning module. After comparing the current node position information of the AGV and the transport task, path planning is performed. That is, when entering the monitoring range of the node Agent, the AGV interacts with information by calling the node Agent interface to initially determine the next travel node. It also detects and resolves conflicts such as head-on collisions and contention for the same node. Finally, it determines the node with no collisions and the shortest distance as the actual next travel node. The AGV then travels to the next travel node, and during the journey, it detects and resolves chasing conflicts. The above process is repeated until the target node is reached.

[0082] 1. Calculate the shortest collision-free path for AGVs based on the interaction between node Agent and AGV Agent.

[0083] (1) Detailed description of the path planning functions of node agents and AGV agents

[0084] When planning a collision-free path for an AGV, the decision-making process mainly relies on the information exchange between the node Agent and the AGV Agent. The specific functions of the corresponding modules for the two types of Agents that implement path planning are detailed below:

[0085] Node Agent

[0086] Each node agent possesses the following knowledge and functions:

[0087] 1) Distance knowledge to all non-adjacent nodes in the road network.

[0088] 2) Knowledge of the distance to all adjacent nodes in the road network and whether the adjacent node is drivable.

[0089] 3) It has three types of AGV sets for statistical analysis and querying of AGV information near the node. That is, the pre-arrival AGV set: stores the AGV information whose next travel node is this node; the departure AGV set: stores the AGV information whose previous travel node is this node; and the arrival AGV set: stores the AGV information currently occupying this node.

[0090] 4) Get your own congestion level interface: You can get the congestion level of this node in real time.

[0091] 5) Conflict AGV One-Step Prediction Interface: Predicts the next travel node after an AGV with a conflict arrives at this node.

[0092] 6) Monitoring radius: Ensure that a judgment is made when the AGV enters the node monitoring range, thereby avoiding the AGV from colliding or stopping at the intersection of the path, which would cause the node to become unusable.

[0093] AGV Agent

[0094] During path planning, the path planning module in the AGV Agent implements collision-free path planning for the AGV. The path planning module includes the following sub-modules:

[0095] 1) Next travel node decision module: When the AGV enters the node monitoring range, it interacts with the node Agent to make a preliminary decision on the next travel node of the AGV.

[0096] 2) Conflict Detection and Resolution Module: Using the next driving node obtained in 1) and the current node of the vehicle as detection objects, determine whether there is an AGV that has a conflict with the vehicle in the road segment between the two nodes, and provide a solution.

[0097] 3) Same Node Contention Conflict Detection and Resolution Module: Determines whether there are other AGVs competing with this vehicle for the right to drive at the same node, and provides a solution.

[0098] 4) Distance monitoring module: Monitors and controls the distance between AGVs traveling in the same direction during AGV operation.

[0099] (2) Collision-free path planning based on node congestion

[0100] 1) Node congestion calculation:

[0101] Drawing inspiration from ants foraging for food, each AGV in the system is considered an ant agent, carrying a certain pheromone λ when entering the road network. Traditional ant colony algorithms use a heuristic function η. ij =1 / d ij The visibility between two points is represented by the pheromone concentration perceived by the node, which decreases as the distance to the AGV increases. Based on this, Lan Peizhen et al. proposed the concept of node congestion to achieve real-time control of AGVs using a real-time global update strategy. The congestion level of each node is updated based on the positions of all AGVs in the transportation network at each time step. At time t, the congestion level at any node (i,j) in the transportation network is as follows:

[0102]

[0103] In reality, not all AGVs in the road network affect the congestion level of a particular node, and the impact of different AGVs' different driving states relative to a node on the congestion level varies. Therefore, this invention only considers AGVs near the node when calculating congestion level (AGVs on the road segment formed by the calculation node and each adjacent node are considered nearby AGVs). As mentioned above, this AGV information is stored in three types of AGV sets of the node Agent. It is easy to see that the three types of AGV sets correspond to three driving states of AGVs relative to the node. Among them, AGVs arriving in the AGV set cause the greatest congestion level to the node because they are currently occupying the node or within the node's monitoring range; secondly, AGVs in the AGV set about to arrive are gradually approaching the node and also contribute to the node's congestion level; AGVs in the AGV set departing from the node are gradually moving away from the node and have the least impact on the node's congestion level. Therefore, this invention improves the congestion level calculation formula based on Lan Peizhen's work as follows:

[0104]

[0105] In the formula, K is the number of AGVs arriving at the AGV set from the node in the transportation network at time t; M is the number of AGVs expected to arrive at the AGV set from the node in the transportation network at time t; and N is the number of AGVs departing from the AGV set from the node in the transportation network at time t. Let α, β, and ε be the straight-line distances between the kth AGV arriving in the AGV set at time t, the mth AGV scheduled to arrive in the AGV set, and the nth AGV departing from the AGV set, respectively, and the node; α, β, and ε are importance coefficients, and α > β > ε.

[0106] 2) Next travel node selection:

[0107] When the AGV reaches a new node position (i,j), it needs to select the next travel node from among many nodes. This process is repeated, and the next travel node is continuously updated until the target node position is reached, at which point the path planning ends. The set of the next travel nodes after node (i,j) is (I',J')={(i',j')|i'≠i orj'≠j}, where Let (i,j) represent the k-th AGV. The higher the congestion at node (i',j'), the greater the likelihood of collisions and conflicts between AGVs at that node. Therefore, it's necessary to avoid overly congested nodes. Based on the node conditions, nodes with lower collision probabilities can be further divided from (i',j') to form [a more efficient network]. The set of drivable nodes, denoted as allowed. k ;like illustrate There are no suitable drivable nodes. Parking and waiting; if but Select the next driving node from it.

[0108] Introducing a crowding threshold q to divide allowed k If τ (i',j') If (t) < q, then the probability of a collision occurring with (i',j') as the next travel node is relatively small and will not affect the normal operation of the AGV system. (i',j') ∈ allowed k If τ (i',j') If (t)≥q, then the probability of a collision occurring at (i',j') as the next travel node is relatively high, which will affect the normal operation of the system.

[0109] From allowed k The next driving node determined in the process needs to satisfy two conditions: minimizing the probability of collision and minimizing the path. These conditions are respectively limited by the attractiveness and distance heuristic functions of the candidate nodes. The attractiveness, i.e., the congestion threshold q, is related to the congestion τ at node (i',j') at time t. (i',j') The difference between (t) and the attraction is calculated using the following formula:

[0110]

[0111] The shortest path requires ensuring that the next node (i', j') is within the range of the AGV's current node (i, j) and the target point (e). x ,e y The shortest distance between the two classes is found; therefore, a heuristic function for the two-class distance is established. and These represent the visibility between the current node (i,j) of the AGV and the next node (i',j'), and the visibility between the next node (i',j') and the target point (e), respectively. x ,e y The visibility between two types of distances is calculated using the following formulas for the two types of distance heuristic functions:

[0112]

[0113]

[0114] Therefore, the transition probability of the AGV moving to each node in the set of drivable nodes is:

[0115]

[0116] Among them, σ, δ, and γ are the importance coefficients of attraction and the two types of visibility.

[0117] Ultimately, the node with the highest probability of node transfer is selected as the next travel node for the AGV.

[0118] 2. Detect and resolve conflicts during driving.

[0119] Multi-AGV path planning and conflict resolution is an extremely complex problem. Path planning based on node congestion can only minimize the chances of AGVs traveling to overly congested nodes, thus avoiding potential collisions, congestion, and deadlocks at those nodes. In actual operation, such conflicts are still highly likely to occur, necessitating continued conflict detection and resolution during operation.

[0120] 1) Common Collision and Conflict Types

[0121] Collisions or deadlocks in AGVs during actual operation can be categorized into the following three types:

[0122] ① Conflicting directions: such as Figure 2 As shown in (a) to (c), during the same time period, two or more AGVs travel towards each other from different nodes to the same road segment, or there are AGVs traveling towards each other or parked AGVs on the next travel segment of a certain node. This is the most serious type of conflict and is difficult to avoid directly through simple measures such as deceleration or braking.

[0123] ②Conflicts over the same node: such as Figure 2 As shown in (d) to (f), two or more AGVs depart from different nodes or road segments or compete for the right to drive at the same node.

[0124] ③ Chase and conflict: such as Figure 2 As shown in (g) to (h), a rear-end collision occurs when two AGVs traveling in the same direction on a certain road segment have different speeds, such as the latter being faster than the former, or the former stopping on the latter's only path.

[0125] 2) Conflict detection mechanism

[0126] Conflict detection between AGVs is achieved through information exchange between the node Agent and the AGV Agent. Different detection methods are set for different conflict types, as follows:

[0127] ① Heading-to-head conflict: When an AGV enters the monitoring range of a certain node, the corresponding AGV Agent makes a preliminary decision on its next travel node by the next travel node decision submodule in the path planning module and submits it to the node Agent that it has arrived at at this moment; then the node Agent traverses its set of pre-arrived vehicles and set of arrived vehicles to search for whether there are any vehicles traveling on the road segment between the current node and the next travel node of the AGV. If there are, then the vehicle must be a heading-to-head conflict vehicle.

[0128] ② Conflicts at the same node: When an AGV enters the monitoring range of a node, the corresponding AGV Agent, after ensuring there are no conflicts at the next travel node by the opposing conflict detection and resolution submodule, enters the same node conflict detection and resolution submodule. This module submits the AGV information to the next travel node Agent. The node Agent traverses its pre-arrival set and arrival set of vehicles to search for any conflicting vehicles competing for the right to travel at the same node. The existence of such vehicles is determined by equations (6) to (7). Equation (6) is for... Figure 2 In cases (e) and (f), equation (7) is for case (g).

[0129]

[0130]

[0131] In the formula, The estimated arrival time for an AGV that is currently making a conflict judgment at the same node to reach the next travel node; is the estimated arrival time of the conflicting vehicle to the next travel node; r is the monitoring radius of the node; v m The speed at which the collision vehicle travels; This indicates that the conflicting vehicle currently occupies the conflict node. The car involved in the conflict stopped at this point.

[0132] ③ Chasing and Conflict Detection: This is achieved mainly through the AGV Agent's own distance monitoring module, which monitors the distance between two vehicles traveling in the same direction on the same road segment in real time.

[0133] 3) Conflict resolution mechanism

[0134] ① Resolving conflicts in opposite directions

[0135] Heading-to-head conflicts are complex. When the AGV Agent identifies such a conflict through the heading-to-head conflict detection and resolution module, the heading-to-head conflict resolution strategy encapsulated in this module resolves the conflict through the following steps:

[0136] a) Remove the next driving node with conflicting directions from the set of feasible nodes. If the set is empty after removal, proceed to step e. Otherwise, continue to select the node with the highest transfer probability from the set as the new next driving node and proceed to step b.

[0137] b. Based on formula (8), determine the time loss of the vehicle taking the route-changing driving strategy compared to the stationary waiting strategy. If there is no time loss, directly take the newly decided driving node as the next driving node and execute step d to make a judgment; otherwise, proceed to step c.

[0138]

[0139] In the formula: tTn is the time it takes for the vehicle to reach the replacement node from its current position; nTg is the shortest time it takes for the vehicle to reach the target node via the replacement node; tTo is the time it takes for the vehicle to reach the conflict node from its current position; oTg is the shortest time it takes for the vehicle to reach the target node from the conflict node; and wt is the parking waiting time.

[0140] c. Changing routes incurs time loss. Based on the road conditions at the current node where the AGV is located, strategies such as stopping in place, changing sides, or forced route change will be considered. The road conditions corresponding to different strategies are as follows:

[0141] In-situ parking: The AGV Agent calls the conflict vehicle one-step prediction interface of the Agent at the current node. The encapsulated function within this interface predicts the next travel node for the conflict vehicle after it reaches this node. After obtaining the prediction result, the AGV further determines whether the predicted node is in the direction of its current road segment. If so, it concludes that the conflict vehicle will not collide with the AGV after reaching this node, and implements an in-situ parking and waiting strategy. Figure 3 As shown in (a).

[0142] Switching sides for parking: If the predicted node is in the direction of the vehicle's current road segment, and a collision is determined after the conflicting vehicles reach the vehicle's current node, the AGV Agent interacts with the current node Agent to check if there is an available road segment near this node. If so, the AGV travels to the available road segment and stops to wait, i.e., switching sides for parking. Figure 3 As shown in (b).

[0143] Forced route change: If none of the above conditions are met, the AGV must change routes, using the node determined in step a as the next travel node. Figure 3 As shown in (c).

[0144] d. If the AGV changes route, it is necessary to continue to detect the node-to-node conflict on the road segment where the new node is located. If the conflict still exists, repeat steps a to c. Otherwise, the AGV will move to the new node.

[0145] e. If the car in question has no alternative route and cannot stop to wait, a collision is inevitable. In this situation, it is necessary to interact with the other car to make it reverse and give way, thus allowing the current car to gain the right of way over the conflicting section. Figure 3 As shown in (d).

[0146] ② Resolution of conflicts involving the same node

[0147] The same node contention conflict is identified and resolved by the same node contention conflict detection and resolution module in the AGV Agent. After such a conflict is detected, the conflict resolution strategy encapsulated in this module is mainly implemented by comparing the priority of the vehicle and the conflicting vehicle. The vehicle with higher priority obtains the right to drive at the conflicting node, while the vehicle with lower priority changes routes. If there is no alternative node, the vehicle stops and waits.

[0148] The priority is determined through the following steps:

[0149] a. Vehicles that are in a conflicting situation at this moment or at the previous driving node should have a higher priority to avoid collisions with vehicles in the conflicting direction.

[0150] b. Compare the congestion levels of the nodes where the conflicting AGVs are currently located. The AGVs with higher congestion levels are given higher priority so that they can leave the node as soon as possible and reduce the congestion level of the node.

[0151] c. If the current nodes have the same level of congestion, compare which AGV is more attractive to the conflicting nodes, and the one with the higher attraction will be given a higher priority.

[0152] ③ Catching up and resolving conflicts

[0153] When the AGV Agent's distance monitoring module detects that the distance between the two workshops is less than the safe distance, the chasing vehicle will stop and wait. The chasing vehicle will only continue to move when the distance between the two workshops reaches the safe distance.

[0154] B. Consider disturbance events, formulate response strategies, and form a hybrid simulation model.

[0155] Common disturbances in production systems include emergency order insertion, AGV malfunction, machine tool failure, and AGV collisions. Solutions for AGV collisions have already been proposed in path planning. Here, response strategies are formulated for the first three scenarios. The main execution flow of the resulting hybrid simulation model is as follows: Figure 4 As shown. Details are as follows:

[0156] (1) Emergency order insertion

[0157] When an urgent order arrives, the system agent generates a corresponding workpiece agent and process agent based on the order information and registers their status as urgent. Accordingly, when the workpiece agent enters the production system from the "order entry" object in the production process sub-model, the urgent workpiece is identified and given the highest priority, allowing it to be negotiated first. When entering the contract negotiation stage of the "negotiation" process, the urgent process agent marks the tender as urgent during the bidding process. Based on this, when calculating the bid content, the AGV and machine tool agents delay all unexecuted tasks in the buffer pool and bid based on the earliest start time of handling (i.e., the end time of the currently executing task). After winning the bid, the urgent task is inserted in front of all non-urgent tasks in the buffer pool, allowing it to be executed first.

[0158] (2) AGV malfunction

[0159] When an AGV malfunctions, all tasks in the AGV task buffer pool are renegotiationd. At this time, the tasks in the buffer pool have two states: in progress and waiting to be moved. In progress is further divided into two states: empty and loaded. When a malfunction occurs, tasks in the empty or waiting state only need to release the corresponding workpiece Agent from the "Waiting and Executing Transportation" module in the production process sub-model to the "Negotiation" module for renegotiation of moving the AGV and the processing machine tool. The workpiece position does not change. In the loaded state, since the workpiece is moving in the road network with the AGV, when the AGV malfunctions, the carried workpiece will be stuck at the node position closest to the AGV at this moment. During renegotiation, the distance information is calculated based on the actual stuck position of the workpiece.

[0160] Regardless of the status of the tasks in the buffer pool, after the "Waiting and Executing Transportation" module releases them, they must flow into the "Judging AGV Fault" object. This module will then divert the workpieces that need to be renegotiation. Workpieces that need to be renegotiation are returned to the "Negotiation" module for renegotiation; otherwise, they flow into the processing module for processing. In addition, before all task renegotiation begins, the task buffer pool of the faulty AGV must be cleared, and the processing tasks of all renegotiation workpieces must be deleted from the task buffer pool of their respective machine tools. After that, the faulty AGV will participate in the bidding again based on the estimated fault end time.

[0161] (3) Machine tool failure

[0162] The response strategy for machine tool failure is to renegotiate all pending tasks in the task buffer pool of the failed machine tool. At this time, the tasks in the buffer pool have four states: processing, waiting for processing in the buffer station, being transported, and waiting for transport. Workpieces being processed are not renegotiated and will continue processing after the failure ends. Tasks waiting for processing and waiting for transport in the buffer station need to be released from the "processing buffer" and "waiting and executing transportation" modules of the production process sub-model to the "negotiation" module for renegotiation of the transport AGV and the processing machine tool. Tasks in the transport state only need to be renegotied for processing machine tools. After obtaining the new processing machine tool position, the path planning module of the contracted AGV can directly change the destination to the new position.

[0163] Workpieces released from the "Processing Buffer" and "Waiting and Executing Transport" modules flow into the "Judgment AGV Fault" and "Judgment Machine Tool Fault" objects, respectively. These modules then divert workpieces requiring renegotiation, returning them to the "Negotiation" module for further negotiation. Otherwise, the workpieces flow into the processing module for continued processing. Similarly, before any renegotiation task begins, all pending tasks in the faulty machine tool's task buffer pool must be cleared, and the transport tasks of all renegotiation workpieces waiting to be transported must be removed from their respective transport AGV's task buffer pool. Afterward, the faulty machine tool participates in the bidding process based on its estimated fault end time. If there are workpieces currently being processed, the estimated end time of processing for those workpieces after the fault ends will be used for bidding.

[0164] Example:

[0165] To verify the feasibility and effectiveness of the integrated scheduling method considering path planning in dynamically handling disturbance events, this chapter tests the response strategies for three common disturbance events and compares them with the case of no response strategy. The comparison indicators include completion time (which reflects the overall scheduling execution result), robustness of the system against disturbances, and the number of collisions in AGV collision avoidance scenarios. During simulation, AGV and machine tool failures in all three scenarios occurred at the 30th second of the simulation time. The faulty devices were A0 and M3, and the failure duration was 10 seconds. Emergency order insertion occurred at the 10th second, with the inserted workpiece being the first workpiece in the baseline data. The comparison results are shown in Table 1, where the robustness calculation formula is:

[0166]

[0167] In the formula, a smaller RM indicates better robustness, and S r S represents the actual scheduling scheme. p Cmax represents the initial scheduling scheme; Cmax represents the maximum completion time of the scheduling scheme.

[0168] Table 1 Comparison of dynamic event response capabilities in different scenarios

[0169]

[0170]

[0171] As shown in Table 1, under the renegotiation response strategy, both system completion time and robustness are improved during equipment failures. There are two types of completion times for emergency order insertion: the time in parentheses represents the completion time of the emergency workpiece, and the other is the completion time of the last workpiece in the system. Under the emergency order insertion strategy, the completion times of the emergency workpieces are all advanced, and the overall system robustness fluctuates very little, averaging only 5% fluctuation. This indicates that the system has good disturbance response capability, proving the effectiveness of the disturbance event response strategy presented in this chapter.

[0172] This invention addresses the dynamic integrated scheduling problem of machine tools and AGVs considering AGV path planning, proposing a hybrid simulation model solution method that integrates path planning and disturbance response strategies. First, a multi-AGV collision-free path planning method based on the interaction between AGV agents and node agents is proposed, including AGV path selection based on node congestion, AGV conflict prediction, and collision avoidance strategy formulation. Second, an integrated scheduling hybrid simulation model under path planning and disturbance response strategies is developed. Renegotiation response strategies are designed for disturbance events including emergency order insertion, AGV failure, machine tool failure, and AGV collisions. This realizes a flexible production workshop machine tool and AGV dynamic integrated scheduling system based on the hybrid simulation model, incorporating collision-free path planning and disturbance response strategies. Finally, simulation examples are used to verify the method and compare it with other algorithms. The comparison results demonstrate the effectiveness and superiority of the proposed method.

Claims

1. A workshop integrated dynamic scheduling method considering path planning and disturbance response, characterized in that, Specifically: A. Collision-free path planning based on information interaction and decision-making between AGV Agent and node Agent; When there are pending transport tasks in the AGV Agent task buffer pool, the task is submitted to the path planning module. After comparing the current node position information of the AGV and the transport task, path planning is performed. That is, when entering the monitoring range of the node Agent, the AGV interacts with information by calling the node Agent interface to initially determine the next travel node. It also detects and resolves conflicts such as head-on collisions and contention for the same node. Finally, it determines the node with no collisions and the shortest distance as the actual next travel node. The AGV then travels to the next travel node, and during the travel, it detects and resolves chasing conflicts. The above process is repeated until the target node is reached. (1) Calculate the shortest collision-free path for AGV based on the interaction between node Agent and AGV Agent; 1) Node congestion calculation: Each AGV in the system is considered as an ant agent, carrying the pheromone λ into the road network. Real-time control of the AGVs employs a real-time global update strategy, updating the congestion level of each node based on the positions of all AGVs within the transportation network at each time step. At time t, the congestion level at any node (i,j) within the transportation network is as follows: In the formula, K is the number of AGVs arriving at the AGV set from the node in the transportation network at time t; M is the number of AGVs expected to arrive at the AGV set from the node in the transportation network at time t; and N is the number of AGVs departing from the AGV set from the node in the transportation network at time t. Let be the straight-line distances between the k-th AGV arriving at time t, the m-th AGV expected to arrive at time t, and the n-th AGV departing at time t, and the node; α, β, and ε are importance coefficients, and α > β > ε. 2) Next travel node selection: When the AGV reaches a new node position (i,j), it needs to select the next travel node from among many nodes. This process is repeated, and the next travel node is continuously updated until the target node position is reached, at which point the path planning ends. The set of the next travel nodes after node (i,j) is (I',J')={(i',j')|i'≠i or j'≠j}, where Let (i,j) represent the k-th AGV. The higher the congestion level at node (i',j'), the greater the possibility of AGVs colliding at that node. Therefore, it's necessary to avoid overly congested nodes. Based on the node conditions, nodes with a low probability of collision are further divided from (i',j') to form [the AGV's AGV configuration]. The set of drivable nodes, denoted as allowed. k ;like illustrate There are no suitable drivable nodes. Parking and waiting; if but Select the next driving node from the list; Introducing a crowding threshold q to divide allowed k If τ (i',j') If (t) < q, then the probability of a collision occurring with (i',j') as the next travel node is small and will not affect the normal operation of the AGV system. (i',j') ∈ allowed. k If τ (i',j') If (t)≥q, then the probability of a collision occurring at (i',j') as the next travel node is high, which will affect the normal operation of the system. From allowed k The next driving node determined in the process needs to satisfy two conditions: minimizing the probability of collision and minimizing the path. These conditions are respectively limited by the attractiveness and distance heuristic functions of the candidate nodes. The attractiveness, i.e., the congestion threshold q, is related to the congestion τ at node (i',j') at time t. (i',j') The difference between (t) and the attraction is calculated using the following formula: The shortest path requires ensuring that the next node (i', j') is within the range of the AGV's current node (i, j) and the target point (ex, e). y The shortest distance between the two classes is found; therefore, a heuristic function for the two-class distance is established. and These represent the visibility between the current node (i,j) of the AGV and the next node (i',j'), and the visibility between the next node (i',j') and the target point (e), respectively. x ,e y The visibility between two types of distances is calculated using the following formulas for the two types of distance heuristic functions: Therefore, the transition probability of the AGV moving to each node in the set of drivable nodes is: Among them, σ, δ, and γ are the importance coefficients of attraction and the two types of visibility; Ultimately, the node with the highest probability of node transfer is selected as the next travel node for the AGV; (2) Conflict detection and resolution during driving; 1) Common Collision and Conflict Types Collisions or deadlocks in AGVs during actual operation can be categorized into the following three types: ① Heading-to-head conflict: At the same time, two or more AGVs travel towards each other from different nodes to the same road segment, or there are AGVs traveling towards each other or parked AGVs on the next travel segment of a certain node. ②Conflict over the same node: When two or more AGVs depart from different nodes or road sections or compete for the right to drive at the same node. ③ Chase-and-go conflict: A rear-end collision occurs when two AGVs traveling in the same direction on a certain road segment have the latter traveling at a higher speed than the former, or when the former stops in the latter's only path. 2) Conflict detection mechanism Conflict detection between AGVs is achieved through information exchange between the node Agent and the AGV Agent. Different detection methods are set for different conflict types, as follows: ① Heading-to-head conflict: When an AGV enters the monitoring range of a certain node, the corresponding AGV Agent makes a preliminary decision on its next travel node by the next travel node decision submodule in the path planning module and submits it to the node Agent that it has arrived at at this moment; then the node Agent traverses its pre-arrived vehicle set and the arrived vehicle set to search for whether there is a vehicle in motion on the road segment between the current node and the next travel node of the AGV. If there is, the vehicle must be a heading-to-head conflict vehicle. ② Conflicts over the same node: When an AGV enters the monitoring range of a certain node, the corresponding AGV Agent is ensured by the opposing conflict detection and resolution submodule to ensure that there are no opposing conflicts at the next driving node. Then, it enters the same node conflict detection and resolution submodule. This module submits the AGV information to the next driving node Agent. The node Agent traverses its pre-arrival vehicle set and arrival vehicle set to search for whether there are conflicting vehicles competing with this vehicle for the right to drive at the same node. The existence is determined by equations (6) to (7). In the formula, The estimated arrival time for an AGV that is currently making a conflict judgment at the same node to reach the next travel node; is the estimated arrival time of the conflicting vehicle to the next travel node; r is the monitoring radius of the node; v m The speed at which the collision vehicle travels; This indicates that the conflicting vehicle currently occupies the conflict node. This indicates that the car involved in the conflict has stopped. ③ Chase-and-go conflict: This is mainly achieved through the AGV Agent's own distance monitoring module, which monitors the distance between two vehicles traveling in the same direction on the same road segment in real time; 3) Conflict resolution mechanism ① Resolving conflicts in opposite directions a. Remove the next driving node with conflicting directions from the set of feasible nodes. If the set is empty after removal, proceed to step e. Otherwise, continue to select the node with the highest transfer probability from the set as the new next driving node and proceed to step b. b. Based on formula (8), determine the time loss of the vehicle taking the route-changing driving strategy compared to the stationary waiting strategy. If there is no time loss, directly take the newly decided driving node as the next driving node and execute step d to make a judgment; otherwise, proceed to step c. Where: tTn is the time it takes for the vehicle to reach the replacement node from its current position; nTg is the shortest time it takes for the vehicle to reach the target node via the replacement node; tTo is the time it takes for the vehicle to reach the conflict node from its current position; oTg is the shortest time it takes for the vehicle to reach the target node from the conflict node; wt is the parking waiting time; c. Changing routes incurs time loss. Based on the road conditions at the current node where the AGV is located, strategies such as stopping in place, changing sides, or forced route change will be considered. The road conditions corresponding to different strategies are as follows: Stop in place: The AGV Agent calls the conflict vehicle one-step prediction interface of the Agent of the current node. The encapsulated function in it predicts the next driving node after the conflict vehicle arrives at this node. After obtaining the prediction result, the AGV continues to determine if the predicted node is not in the direction of the current road segment. If so, it is determined that the conflict vehicle will not collide with the AGV after it arrives at this node, and the stop in place waiting strategy is implemented. Switching sides parking: If the predicted node is in the direction of the current road segment of this vehicle, and it is determined that the conflicting small vehicles will collide after reaching the node of this vehicle, then the AGV Agent will interact with the Agent of the current node to see if there is an empty road segment near the current node. If there is, the AGV will drive to the empty road segment and stop to wait, which is called switching sides parking. Forced route change: If neither stopping in place nor changing sides can be satisfied, the AGV must change routes, and the node determined in step a will be the next driving node. d. If the AGV changes route, it is necessary to continue to detect the node-to-node conflict on the road segment where the new node is located. If the conflict still exists, repeat steps a to c; otherwise, the AGV will move to the new node. e. When the current car has no way to change lanes or can not stop and wait, a collision is inevitable. In this case, it is necessary to interact with the conflicting car to make it reverse and give way, so that the car can occupy the right of way in the conflicting section. ② Resolution of conflicts involving the same node Conflicts over the same node are identified and resolved by the same node conflict detection and resolution module in the AGV Agent. Upon detection, the conflict resolution strategy encapsulated in this module compares the priorities of the current vehicle and the conflicting vehicle. The vehicle with higher priority gains the right to drive at the conflicting node, while the vehicle with lower priority changes routes. If no alternative node is available, the vehicle stops and waits. The priority is determined through the following steps: a. Vehicles that are in a conflicting situation at this moment or at the previous driving node should have a higher priority to avoid collisions with vehicles in the conflicting direction; b. Compare the congestion levels of the nodes where the conflicting AGVs are currently located. The AGVs with higher congestion levels are given higher priority so that they can leave the node as soon as possible and reduce the congestion level of the node. c. If the current nodes have the same level of congestion, compare which AGV is more attractive to the conflicting nodes, and the one with the higher attraction will be given a higher priority. ③ Catching up and resolving conflicts When the AGV Agent's distance monitoring module detects that the distance between the two workshops is less than the safe distance, the chasing vehicle will stop and wait. The chasing vehicle will only continue to move when the distance between the two workshops is the safe distance. B. Consider disturbance events, formulate response strategies, and form a hybrid simulation model; The disturbances here include emergency order insertion, AGV malfunction, and machine tool malfunction; (1) Emergency order insertion When an urgent order arrives, the system agent generates a corresponding workpiece agent and process agent based on the order information and registers its status as urgent. Accordingly, when the workpiece agent enters the production system from the "order entry" object in the production process sub-model, the urgent workpiece is identified and given the highest priority, allowing it to be negotiated first. When entering the contract negotiation stage of the "negotiation" process, the urgent process agent marks the tender as urgent during the bidding process. Based on this, when calculating the bid content, the AGV and machine tool agents postpone all unexecuted tasks in the buffer pool and bid based on the earliest start time of handling. After winning the bid, the urgent task is inserted in front of all non-urgent tasks in the buffer pool, allowing it to be executed first. (2) AGV malfunction When an AGV malfunctions, all tasks in the AGV task buffer pool are renegotiationd. At this time, the tasks in the buffer pool have two states: in progress and waiting to be moved. In progress is further divided into two states: empty and loaded. When a malfunction occurs, tasks in the empty or waiting state only need to release the corresponding workpiece Agent from the "Waiting and Executing Transportation" module in the production process sub-model to the "Negotiation" module to renegotiation of moving the AGV and the machine tool. The workpiece position does not change. In the loaded state, since the workpiece is moving in the road network with the AGV, when the AGV malfunctions, the carried workpiece will be stuck at the node position closest to the AGV at this moment. During renegotiation, the distance information is calculated based on the actual stuck position of the workpiece. Regardless of the status of the tasks in the buffer pool, after the "Waiting and Executing Transportation" module releases them, they must flow into the "Judging AGV Fault" object. This module will then divert the workpieces that need to be renegotiation. Workpieces that need to be renegotiation are returned to the "Negotiation" module for renegotiation; otherwise, they flow into the processing module for processing. In addition, before all task renegotiation begins, the task buffer pool of the faulty AGV must be cleared, and the processing tasks of all renegotiation workpieces must be deleted from the task buffer pool of their respective machine tools. After that, the faulty AGV will participate in the bidding again based on the estimated fault end time. (3) Machine tool failure The response strategy for machine tool failure is to renegotiate all pending tasks in the task buffer pool of the failed machine tool. At this time, the tasks in the buffer pool have four states: processing, waiting for processing in the buffer station, being transported, and waiting for transport. Workpieces being processed are not renegotiated and will continue processing after the failure ends. Tasks waiting for processing and waiting for transport in the buffer station need to be released from the "processing buffer" and "waiting and executing transportation" modules of the production process sub-model to the "negotiation" module for renegotiation of transport AGVs and processing machine tools. Tasks being transported only need to be renegotied for processing machine tools. After obtaining the new processing machine tool position, the path planning module of the contracted AGV can directly change the destination to the new position. Workpieces released from the "Processing Buffer" and "Waiting and Executing Transport" modules flow into the "Judgment AGV Fault" and "Judgment Machine Tool Fault" objects, respectively. These modules then divert workpieces requiring renegotiation, returning them to the "Negotiation" module for further negotiation. Otherwise, the workpieces flow into the processing module for continued processing. Similarly, before any renegotiation task begins, all pending tasks in the faulty machine tool's task buffer pool must be cleared, and the transport tasks of all renegotiation workpieces waiting to be transported must be removed from their respective transport AGV's task buffer pool. Afterward, the faulty machine tool participates in the bidding process based on its estimated fault end time. If there are workpieces currently being processed, the estimated end time of processing for those workpieces after the fault ends will be used for bidding.

Images