Production workshop intersection passing method
By building a model and a quick pass knowledge decision model in the production workshop, the problems of slow AGV path planning speed and poor cross-pass decision-making ability are solved, and the rapid planning and transportation plan of AGV routes are realized, which improves the transportation efficiency of the production workshop and Level of informatization and intelligence.
Patent Information
- Application Number
- CN202411042013.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-31
- Publication Date
- 2024-11-15
AI Technical Summary
In the prior art, the AGV path planning speed in the production workshop is slow, the intersection decision-making ability is poor, and it is prone to deadlocks and congestion, which affects transportation efficiency and production efficiency.
Build a production workshop model, plan the driving path through the starting point and end point, obtain vehicle information at the intersection, input it into the state space model to obtain spatial state parameters, combine it with the fast-pass knowledge decision model, formulate action plans and combine them with road driving rules to form an intersection The traffic plan for intersections can realize the rapid planning of AGV routes and decision-making of intersections.
It improves the effectiveness of AGV scheduling plans, reduces deadlocks and congestion, improves the transportation efficiency of the production workshop, and promotes the development of informatization and intelligence.
Smart Images

Figure CN118968798A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of vehicle dispatching, and in particular to a method for passing through a production workshop intersection. Background Art
[0002] AGV scheduling in production workshops is an important aspect of current production management. Battery cell production has the characteristics of complex process routes and high degree of automation. The AGV cluster is responsible for the transportation of materials in the workshop. The problem of AGV scheduling in production workshops is how to distribute the transportation tasks to each AGV in a balanced and reasonable manner, arrange the order of task execution reasonably, plan the optimal driving path for the AGV, monitor and control the driving status of the AGV, and avoid path conflicts between AGVs.
[0003] In an environment where multiple AGVs operate collaboratively, conflicts and deadlocks that occur when AGVs pass through intersections are important factors affecting transportation efficiency. Traditional scheduling methods avoid conflicts by setting simple driving rules, but such fixed driving rules cannot make full use of workshop operation data, which can easily lead to road congestion, reduce production efficiency and corporate competitiveness, and are not conducive to the informatization and intelligent development of production workshops. Summary of the invention
[0004] In view of this, the purpose of the present invention is to provide a method for passing through an intersection in a production workshop, which solves the shortcomings of the prior art in which the AGV path planning in the production workshop is slow, the decision-making ability for intersection passage is poor, and deadlock and congestion are prone to occur, thereby improving the effectiveness of the AGV scheduling plan in the production workshop and the production efficiency of the enterprise, and at the same time promoting the informatization and intelligent development of the production workshop.
[0005] The present invention provides a method for passing through a production workshop intersection, constructing a production workshop model, and planning a driving path on the production workshop model through a starting point and an end point, wherein the production workshop model includes a plurality of intersections formed by the intersection of a plurality of lanes;
[0006] Obtain relevant information of vehicles at the intersection and input it into the state space model to obtain space state parameters;
[0007] Input the spatial state parameters into the fast-pass knowledge decision model to obtain an action plan;
[0008] Combine the action plan with the road driving rules to obtain the traffic plan for the intersection.
[0009] In one embodiment, planning a driving route through a starting point and an end point further includes:
[0010] Constructing a production workshop model according to the production workshop;
[0011] Marking a starting point, an end point and several nodes on the production workshop model;
[0012] Initialize OP table and CL table;
[0013] Searching for nodes adjacent to the starting point according to the starting point;
[0014] Store the adjacent nodes into the OP table;
[0015] Calculate the node cost of all nodes in the OP table according to the heuristic function;
[0016] Select the corresponding node according to the node cost and store it in the CL table;
[0017] Repeat the above steps until the end point exists among the adjacent nodes being searched;
[0018] Retrieve all nodes in the CL table to generate a driving path.
[0019] In one embodiment, selecting a corresponding node according to the node cost and storing it in the CL table further includes:
[0020] When several nodes have different costs, select the node with the smallest cost and store it in the CL table;
[0021] When there are several nodes with the same cost, and the same node cost value is the smallest, the node that allows the vehicle to go straight is selected from the nodes with the same node cost and stored in the CL table. When there is no node that allows the vehicle to go straight, one of the nodes with the same node cost is randomly selected and stored in the CL table.
[0022] In one embodiment, the driving rules further include:
[0023] Obtain the driving conditions of vehicles and determine whether there is a conflict between vehicles;
[0024] In the case of a conflict, calculating the order in which the conflicting vehicles arrive at the intersection;
[0025] The vehicle that arrives at the intersection first passes first, and the vehicle that arrives at the intersection later stops first;
[0026] After the vehicle that arrives at the intersection first passes, the vehicle that arrives at the intersection later continues to pass;
[0027] Repeat the above steps until all vehicles pass through the intersection;
[0028] In the absence of conflict, all vehicles pass through the intersection according to the corresponding planned driving paths.
[0029] In one embodiment, the traffic plan further includes:
[0030] Execute the action plan, in the case of a conflict between vehicles, the vehicle that arrives at the intersection first passes first, and the vehicle that arrives at the intersection later stops first;
[0031] After the vehicle that arrives at the intersection first passes, the vehicle that arrives at the intersection later continues to pass until the execution of the action plan is completed.
[0032] In one embodiment, the training process of the fast-pass knowledge decision model further includes:
[0033] Obtain relevant information about the vehicle and input it into the constructed state space model, action model and reward function model to obtain the corresponding output parameters;
[0034] The corresponding parameters of the output are input into the initial decision model for training to obtain a fast-pass knowledge decision model.
[0035] In one embodiment, the state space model is expressed as:
[0036]
[0037] Among them, NUM t Represented as a matrix of the number of vehicles on the lanes in the four directions, where num lt 、num dt 、num rt 、num ut They represent the number of vehicles in the left lane, the lower lane, the right lane, and the upper lane, respectively. It is represented as the driving direction matrix of each vehicle after arriving at the intersection, where max means that when obtaining the driving direction, only the max number of vehicles close to the intersection on each lane is considered, L1 t , D2 t , R3 t , U4 t It is expressed as the driving direction of the maximum number of vehicles in the left lane, the lower lane, the right lane, and the upper lane after arriving at the intersection, POR t is the length ratio matrix, which indicates the ratio of the length of the stationary AGV to the length of the lane. t =[P lt , P dt , P rt , P ut ] is expressed as the ratio of the length of the stationary AGV vehicles in the left lane, the lower lane, the right lane and the upper lane to the lane length.
[0038] In one embodiment, the action model can be expressed as:
[0039] a t ∈{0, 1}
[0040] Among them, 0 and 1 represent the two passable directions of the intersection respectively.
[0041] In one embodiment, the reward function model can be expressed as:
[0042]
[0043] Among them, w ik is the waiting time of the kth stationary car on lane i, n i is the number of vehicles waiting at rest on road i.
[0044] The production workshop intersection passage method provided by the present invention can realize the rapid planning of AGV routes, the rapid formulation of intersection passage decision plans, and then realize real-time scheduling of AGVs in production workshops, which solves the shortcomings of slow AGV path planning in production workshops, poor intersection passage decision-making ability, and easy deadlock and congestion in the prior art, improves the effectiveness of the production workshop AGV scheduling plan and the enterprise's production efficiency, and at the same time promotes the informatization and intelligent development of production workshops. BRIEF DESCRIPTION OF THE DRAWINGS
[0045] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.
[0046] Figure 1 A schematic flow chart of the production workshop intersection passage method provided by the present invention.
[0047] Figure 2 A schematic diagram of the process flow of driving route planning provided by the present invention.
[0048] Figure 3 A schematic diagram of the road driving rules provided by the present invention.
[0049] Figure 4 This is a schematic diagram of the training of the common knowledge decision model provided by the present invention. DETAILED DESCRIPTION
[0050] The specific embodiments of the present invention will be described in detail below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the description of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0051] In the description of the present invention, unless otherwise clearly specified and limited, the terms "set", "install", "connection" and the like should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium. For ordinary technicians in this field, the specific meanings of the above terms can be understood according to specific circumstances.
[0052] The directions or positional relationships indicated by terms such as “upper”, “lower”, “left”, “right”, “front”, “back”, “top”, “bottom”, “inside” and “outside” are based on the directions or positional relationships shown in the accompanying drawings, or are the directions or positional relationships in which the inventive product is usually placed when used. They are only for the convenience of description and simplified description, and do not indicate or imply that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore should not be understood as a limitation on the present invention.
[0053] The terms "first", "second", "third" and the like are merely used to distinguish elements of similar nature, and do not indicate or imply relative importance or a particular order.
[0054] The terms "comprises," "comprising," or any other variations thereof, are intended to cover a non-exclusive inclusion of the elements listed and may also include additional elements not expressly listed.
[0055] See also Figure 1 The production workshop intersection passing method provided by the present invention comprises the following steps:
[0056] S1, constructing a production workshop model, and planning a driving path on the production workshop model through a starting point and an end point, wherein the production workshop model includes a plurality of intersections formed by the intersection of a plurality of lanes.
[0057] See also Figure 2 It is understandable that the above steps may further include:
[0058] S101, constructing a production workshop model according to the production workshop.
[0059] It is understandable that the workshop model can be constructed 1:1 according to the prototype of the production workshop.
[0060] S102, mark the starting point and end point on the production workshop model (equivalent to Figure 2 The target position in () and several nodes.
[0061] It can be understood that the starting point can be the starting point in the corresponding production workshop, the end point can be the end point in the corresponding production workshop, and the node can be a representation point set on the lane, which can be located on the lane or at the intersection formed between lanes. The distance between adjacent nodes can be different.
[0062] S103, initializing the OP table and the CL table;
[0063] S104, searching for nodes adjacent to the starting point according to the starting point.
[0064] It can be understood that the number of nodes adjacent to the starting point may be one or more.
[0065] S105, storing the adjacent nodes into the OP table;
[0066] S106, calculating the node costs of all nodes in the OP table according to the heuristic function.
[0067] It is understandable that the node cost can be calculated using the following formula:
[0068] f(P)=g(P)+h(P)+d(P)
[0069] Where f(P) is the total cost, g(P) is the actual cost from the starting point to node P, h(P) is the estimated cost from node P to the end point, and d(P) is the vehicle turning penalty cost from the starting point to node P. The estimated cost h(P) is the Manhattan distance between node P and the end point. d(P) = C num_r (P)×l turn_r +C num_l (P)×l turn_l , where C num_r (P) and C num_l (P) are the number of right turns and left turns from the starting point to the node P, respectively. turn_r and l turn_l are the penalty costs for each right turn and left turn respectively.
[0070] S107, selecting corresponding nodes according to the node cost and storing them in the CL table.
[0071] It is understandable that the above steps may further include:
[0072] S107a, when several node costs are different, select the node with the smallest node cost and store it in the CL table;
[0073] It is understandable that before storing it in the CL table, it will be determined whether the node is the end point. If it is, it does not need to be stored in the CL table again. The driving path can be planned directly through the starting point and the end point. The node information in the CL table includes coordinate information, node cost and other parameters.
[0074] S107b, when there are several nodes with the same cost, and the same node cost value is the smallest, select the node that allows the vehicle to go straight from the nodes with the same node cost and store it in the CL table; when there is no node that allows the vehicle to go straight, randomly select one of the nodes with the same node cost and store it in the CL table.
[0075] It can be understood that, when there are multiple nodes that allow the vehicle to go straight, a node can be randomly selected from these adjacent nodes and stored in the CL table.
[0076] S108, repeat the above steps until an end point is found among the adjacent nodes being searched.
[0077] It is understandable that when there is an end point between adjacent nodes, the above steps can be terminated and the end point may not be put into the CL table.
[0078] S109, retrieve all nodes in the CL table to generate a driving path.
[0079] It can be understood that, by combining the starting point and the end point and connecting the nodes in sequence to form a driving path, the above driving path generation method can achieve rapid generation of the driving path.
[0080] S2, obtain relevant information of vehicles at the intersection and input it into the state space model to obtain space state parameters.
[0081] It can be understood that each lane includes two one-way roads, and the vehicles on the two one-way roads are moving in opposite directions. The intersection can be formed by the intersection of four lanes. The intersection can act as an intelligent entity, constantly interacting with the environment in the workshop model to collect vehicle-related information. Vehicles in the intersection area can actively send corresponding relevant information to the intersection. The vehicle-related information may include state space data.
[0082] S3, obtaining relevant information of vehicles at the intersection and inputting it into the state space model to obtain space state parameters.
[0083] S4, input the spatial state parameters into the fast-pass knowledge decision model to obtain an action plan.
[0084] Understandably, see Figure 4 , the training process of the state space model can further include:
[0085] The corresponding parameters output by the state space model, action model and reward function model are input into the initial decision model for training to obtain a fast-pass knowledge decision model.
[0086] It can be understood that the initial decision model can be formed by building a target network and an estimation network, initializing the network parameters and experience pool of the target network and the estimation network, and using a deep reinforcement learning algorithm to train based on the output parameters of the established state space model, action model, and reward function model. The network parameters are optimized to obtain a fast-pass knowledge decision model fitted with a DQN network, and a traffic plan for the intersection is generated based on the fast-pass knowledge decision model to maximize the set reward function. The execution plan can be to allow vehicles in the north-south lane to travel. In this case, vehicles in the east-west lane direction, if they need to enter If the vehicle is currently at an intersection formed by the north-south lane and the east-west lane, the vehicle needs to stop, while vehicles that have left the intersection can continue to drive. When the execution plan is to allow vehicles on the east-west lane to drive, in this case, vehicles located in the southeast north direction need to enter the intersection formed by the current north-south lane and the east-west lane, and the vehicle needs to stop, while vehicles that have left the intersection can continue to drive. It can be understood that the knowledge decision-making model will output an action plan at predetermined intervals, and the knowledge decision-making model will transmit the action plan to the intersection, and the intersection will interact with the vehicle to execute the action plan.
[0087] It can be understood that the above state space model, action model and reward function model can be expressed as:
[0088] The state space model is expressed as:
[0089]
[0090] Among them, NUM t Represented as a matrix of the number of vehicles on the lanes in the four directions, where num lt 、num dt 、num rt 、num ut They represent the number of vehicles in the left lane, the lower lane, the right lane, and the upper lane, respectively. It is represented as the driving direction matrix of each vehicle after arriving at the intersection, where max means that when obtaining the driving direction, only the max number of vehicles close to the intersection on each lane is considered, L1 t , D2 t , R3 t , U4 t It is expressed as the driving direction of the maximum number of vehicles in the left lane, the lower lane, the right lane, and the upper lane after arriving at the intersection, POR tis the length ratio matrix, which indicates the ratio of the length of the stationary AGV to the length of the lane. t =[P lt , P dt , P rt , P ut ] is expressed as the ratio of the length of the stationary AGV vehicles in the left lane, the lower lane, the right lane and the upper lane to the lane length.
[0091] The action model can be expressed as:
[0092] a t ∈{0, 1}
[0093] Among them, 0 and 1 respectively represent the two passable directions of the intersection. It can be known that the two directions can be the east-west direction and the north-south direction respectively.
[0094] The reward function model can be expressed as:
[0095]
[0096] Among them, w ik is the waiting time of the kth stationary car on lane i, n i is the number of vehicles waiting at rest on road i.
[0097] S4, combining the action plan with the road driving rules to obtain a traffic plan for the intersection.
[0098] See also Figure 3 It is understood that the road driving rules may further include:
[0099] Obtain the driving conditions of vehicles and determine whether there is a conflict between vehicles;
[0100] In the case of conflicts, the order in which conflicting vehicles arrive at the intersection is calculated;
[0101] The vehicles that arrive at the intersection first will pass first, and the vehicles that arrive at the intersection later will stop first;
[0102] After the vehicles that arrived at the intersection first have passed, the vehicles that arrived at the intersection later can continue to pass;
[0103] Repeat the above steps until all vehicles have passed the intersection;
[0104] In the absence of conflicts, all vehicles pass through the intersection according to the corresponding planned driving paths.
[0105] It can be understood that the judgment of the conflict situation can be understood as calculating the distance of the vehicle from the intersection, the driving direction of the vehicle at the intersection, and the driving conditions of other vehicles at the intersection when the vehicle is at the intersection to determine whether there will be a conflict between vehicles.
[0106] The driving speeds of the above vehicles can be consistent, and the order of arriving at the intersection can be the distance between the intersection and the intersection. The distance to the intersection can be calculated by setting a mark point at the center of the intersection and calculating the distance to the mark point. The smaller the distance, the earlier the arrival order.
[0107] It is understandable that in the case of the vehicle that passes first, any vehicle that does not conflict with the vehicle can pass, and the vehicles behind the vehicle that conflicts with the vehicle will also stop driving.
[0108] After the vehicles that arrive at the intersection first have passed, the vehicles that arrive at the intersection later can continue to pass.
[0109] It is understandable that when there is no conflict, vehicles do not affect each other and can pass through the intersection according to their own planned driving paths.
[0110] Based on the above road driving rules, the road driving rules are combined with the action plan to obtain a traffic plan. The traffic plan may further include:
[0111] Execute the action plan. In the case of a conflict between vehicles, the vehicle that arrives at the intersection first will pass first, and the vehicle that arrives at the intersection later will stop first;
[0112] After the vehicles that arrive at the intersection first pass through, the vehicles that arrive at the intersection later continue to pass until the execution of the action plan is completed.
[0113] It is understandable that the action plan has the highest priority. When executing the action plan, in case of conflict, the road driving rules will be combined to obtain the final traffic plan.
[0114] S5, the above-mentioned traffic plan for the intersection can also be simulated in the production workshop model.
[0115] It is understandable that the production workshop model can be a digital twin workshop.
[0116] From the above description, it can be known that the production workshop intersection passage method provided by the present invention can realize the rapid planning of AGV routes, the rapid formulation of intersection passage decision plans, and then realize real-time scheduling of AGVs in production workshops, which solves the shortcomings of slow path planning of AGVs in production workshops, poor intersection passage decision-making ability, and easy deadlock and congestion in the prior art, improves the effectiveness of the production workshop AGV scheduling plan and the production efficiency of the enterprise, and at the same time promotes the informatization and intelligent development of production workshops.
[0117] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed by the present invention should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the attached claims.
Claims
1. A method for passing through a production workshop intersection, It is characterized in that The steps include: Constructing a production workshop model, and planning a driving path on the production workshop model through a starting point and an end point, wherein the production workshop model includes a plurality of intersections formed by the intersection of a plurality of lanes; Obtain relevant information of vehicles at the intersection and input it into the state space model to obtain space state parameters; Input the spatial state parameters into the fast-pass knowledge decision model to obtain an action plan; Combine the action plan with the road driving rules to obtain the traffic plan for the intersection.
2. The production workshop intersection passing method according to claim 1, It is characterized in that Constructing a production workshop model, and planning a driving path on the production workshop model through a starting point and an end point, further comprising: Constructing a production workshop model according to the production workshop; Marking a starting point, an end point and several nodes on the production workshop model; Initialize OP table and CL table; Searching for nodes adjacent to the starting point according to the starting point; Store the adjacent nodes into the OP table; Calculate the node cost of all nodes in the OP table according to the heuristic function; Select the corresponding node according to the node cost and store it in the CL table; Repeat the above steps until the end point exists among the adjacent nodes being searched; Retrieve all nodes in the CL table to generate a driving path.
3. The production workshop intersection passing method as claimed in claim 2, It is characterized in that The selecting corresponding nodes according to the node costs and storing them in the CL table further includes: When several nodes have different costs, select the node with the smallest cost and store it in the CL table; When there are several nodes with the same cost, and the same node cost value is the smallest, the node that allows the vehicle to go straight is selected from the nodes with the same node cost and stored in the CL table. When there is no node that allows the vehicle to go straight, one of the nodes with the same node cost is randomly selected and stored in the CL table.
4. The production workshop intersection passing method according to claim 1, It is characterized in that The driving rules further include: Obtain the driving conditions of vehicles and determine whether there is a conflict between vehicles; In the case of a conflict, calculating the order in which the conflicting vehicles arrive at the intersection; The vehicle that arrives at the intersection first passes first, and the vehicle that arrives at the intersection later stops first; After the vehicle that arrives at the intersection first passes, the vehicle that arrives at the intersection later continues to pass; Repeat the above steps until all vehicles pass through the intersection; In the absence of conflict, all vehicles pass through the intersection according to the corresponding planned driving paths.
5. The production workshop intersection passing method as claimed in claim 4, It is characterized in that The passage scheme further includes: Execute the action plan, in the case of a conflict between vehicles, the vehicle that arrives at the intersection first passes first, and the vehicle that arrives at the intersection later stops first; After the vehicle that arrives at the intersection first passes, the vehicle that arrives at the intersection later continues to pass until the execution of the action plan is completed.
6. The production workshop intersection passing method according to claim 1, It is characterized in that The training process of the fast-pass knowledge decision model further includes: Obtain relevant information about the vehicle and input it into the constructed state space model, action model and reward function model to obtain the corresponding output parameters; The corresponding parameters of the output are input into the initial decision model for training to obtain a fast-pass knowledge decision model.
7. The production workshop intersection passing method according to claim 6, It is characterized in that The state space model is expressed as: Among them, NUM t Represented as a matrix of the number of vehicles on the lanes in the four directions, where num lt 、num dt 、num rt 、num ut They represent the number of vehicles in the left lane, the lower lane, the right lane, and the upper lane, respectively. It is represented as the driving direction matrix of each vehicle after arriving at the intersection, where max means that when obtaining the driving direction, only the max number of vehicles close to the intersection on each lane is considered, L1 t , D2 t , R3 t , U4 t It is expressed as the driving direction of the maximum number of vehicles in the left lane, the lower lane, the right lane, and the upper lane after arriving at the intersection, POR t is the length ratio matrix, which indicates the ratio of the length of the stationary AGV to the length of the lane. t =[P lt , P dt , P rt , P ut ] is expressed as the ratio of the length of the stationary AGV vehicles in the left lane, the lower lane, the right lane and the upper lane to the lane length.
8. The method for passing through a production workshop intersection as claimed in claim 6, It is characterized in that The action model can be expressed as: a t ∈{0,1} Among them, 0 and 1 represent the two passable directions of the intersection respectively.
9. The production workshop intersection passing method according to claim 6, It is characterized in that The reward function model can be expressed as: Among them, w ik is the waiting time of the kth stationary car on lane i, n i is the number of vehicles waiting at rest on road i.
