IGV and outer container truck intelligent interactive mixed scheduling method based on large language model

Abstract

The invention relates to an IGV and outer container truck intelligent interactive mixed scheduling method based on a large language model. The method comprises the following steps: S1, arranging a container automatic storage yard area; s2, determining mixed traffic flow of storage yard operation; s3, formulating a traffic control rule based on multi-agent interaction, including formulating multi-agent interaction communication and formulating a traffic control rule based on a large language model; and S4, designing an IGV and outer container truck mixed scheduling process. According to the horizontal transportation system, the scheme that in-port horizontal transportation vehicles and outer container trucks are physically separated can be omitted, the effective area of a storage yard is enlarged, mixed traffic of the IGV and the outer container trucks is achieved, the operation process is optimized, the collection and distribution efficiency and the port area horizontal transportation capacity are improved, meanwhile, the equipment loading and unloading operation efficiency is improved, and the equipment demand quantity is reduced. And the construction cost of port equipment is reduced, and the method is suitable for new construction or reconstruction and extension of automatic container wharfs with horizontally-arranged storage yards.

Description

Technical Field

[0001] The present invention relates to the technical field of container terminal transportation, and more specifically to an intelligent interactive mixed dispatching method of IGVs and external container trucks based on a large language model. Background Art

[0002] In recent years, the global trade volume has increased significantly. Due to the advantages of large volume, economic efficiency and high efficiency of container transportation, the proportion of container transportation has increased year by year. The sudden increase in container throughput has put forward higher requirements on the production and operation of terminals, among which the horizontal transportation of container terminal yards is an important link affecting the efficiency of container loading and unloading.

[0003] like Figure 1 As shown in the figure, most traditional manual container terminals adopt a horizontal layout, with the container area and the shoreline kept horizontal. After the automation upgrade, the horizontal transport equipment will be mixed when operating in the yard area. In the early stage of the construction of the automated terminal, since the horizontal transport equipment in the port does not have the mixed traffic technology, most automated terminals adopt the physical separation scheme of the horizontal transport vehicles in the port and the external container trucks. When the land area in the port is certain, the physical separation scheme needs to occupy part of the space, and the effective area of ​​the yard is reduced, which greatly affects the port operation efficiency. With the continuous development of the intelligent level of ports, emerging technologies such as satellite navigation and automatic driving are applied to unmanned container trucks, providing technical support for the mixed operation of external container trucks and unmanned container trucks. When the traditional manual terminal is upgraded and renovated, the original layout can be basically maintained without setting up a physical isolation area, which greatly reduces the transformation cost. Therefore, how to achieve the mixed operation of external container trucks and unmanned container trucks in the horizontal layout container terminal yard and improve the efficiency of container truck transportation has become a current research hotspot.

[0004] Traditional manual container terminals mostly adopt a horizontal layout, with the container area and the shoreline kept horizontal. After the introduction of automated horizontal transport equipment, unmanned container trucks and external container trucks are mixed when operating in the yard area. Some automated terminals have adopted the method of adjusting the yard layout and rebuilt it into a vertical layout yard, which increases the cost of port construction; some automated terminals have adopted a physical isolation solution to separate the horizontal transport vehicles from the external container trucks, set up a physical isolation area to occupy part of the space, and reduce the effective area of ​​the yard. Summary of the invention

[0005] The technical problem to be solved by the present invention is to provide an intelligent interactive mixed scheduling method for IGVs and foreign container trucks based on a large language model, which expands the effective area of ​​the yard, realizes the mixed traffic of IGVs and foreign container trucks, optimizes the operation process, improves the collection and distribution efficiency and the horizontal transportation capacity of the port area, reduces the demand for equipment, and reduces the cost of port equipment construction. It is suitable for the new construction or renovation and expansion of automated container terminals with horizontal yard layout.

[0006] The technical solution adopted by the present invention to solve the technical problem is: constructing an intelligent interactive mixed dispatching method of IGV and external container truck based on a large language model, comprising the following steps:

[0007] S1. Arrange the automated container yard area;

[0008] S2. Determine the mixed traffic flow for yard operations;

[0009] S3, Formulate traffic control rules based on multi-agent interaction, including: rule formulation multi-agent interactive communication and formulation of traffic control rules based on large language models;

[0010] S4. Design the mixed dispatching process of IGV and external container trucks.

[0011] According to the above scheme, in step S1, the automated container yard area includes a yard container area, a rubber-tyred gantry crane and a road in the yard; the yard container area is arranged parallel to the dock shoreline, the IGV drives counterclockwise in the seaside lane to enter the yard container area, and the external container truck drives clockwise in the landside lane to enter the yard container area.

[0012] According to the above scheme, in step S2, the method for confirming the mixed traffic flow for yard operations is as follows: all roads in the port area are one-way lanes. When the IGV and the external container truck have traffic intersections at the yard overtaking lane and the main road, they travel along the mixed route to the rubber-tyred gantry crane to carry out container loading and unloading operations; when the IGV and the external container truck enter the same section of road, they produce following driving; when the IGV and the external container truck pass the same intersection, intersection interaction occurs.

[0013] According to the above scheme, when the IGV and the external container truck interact at the intersection, there is a conflict in the right of way and the safe time distance of the intersection, resulting in a mixed traffic conflict. The IGV and the external container truck both go straight through their own lanes; when the time difference between the IGV and the external container truck when they reach the intersection of the lanes is less than the safe time distance, the two vehicles collide; the IGV and the external container truck both go straight through the same lane, and the two vehicles are in a following driving. If the time difference between the front vehicle and the rear vehicle passing the same position of the section is less than the safe time distance required for following, the rear vehicle cannot emergency brake and a conflict occurs.

[0014] According to the above scheme, in step S3, the rules for formulating multi-agent interactive communication are specifically as follows: the vehicle management agent obtains the geographic location and transportation task status data of all IGVs and external container trucks in real time, determines the time when the vehicles enter and leave the road section, and generates the initial path time window of each vehicle;

[0015] The road network agent interacts with the vehicle management agent. The road network agent detects mixed traffic conflict information in the road network, uses the traffic control strategy based on the large language model to adjust the path time window, and passes it to the vehicle management agent for processing;

[0016] When IGV and external container trucks are traveling at the same intersection, the road network agent will uniformly macro-control the opening of the intersection according to vehicle priority, right of way, and safe time interval, and intelligently control the passage of IGV and external container trucks.

[0017] According to the above scheme, in step S3, formulating traffic control rules based on a large language model includes the following steps:

[0018] S301, data collection: To solve the mixed traffic problem of IGV and foreign container trucks at horizontal container terminals, obtain mixed traffic control data, divide training samples, and obtain training set, validation set and test set;

[0019] S302, Model architecture selection: Select a large language model based on Transformer to understand natural language text, process structured data, and generate traffic control strategies;

[0020] S303, model training: input the training set data into the model, generate the predicted traffic control strategy through forward propagation, calculate the loss function between the predicted and the actual results, update the model parameters through the back propagation algorithm, and iterate the training until the model achieves satisfactory performance on the validation set;

[0021] S304, fine-tuning and iterative optimization: Based on the meta-optimizer, use the mixed row conflict data to fine-tune the model with a passed verification result to improve its accuracy on the task;

[0022] S305, model evaluation and deployment: Use the test set to evaluate the performance of the model. When the evaluation passes, the model is deployed to the actual application to handle mixed row conflict problems in real time.

[0023] According to the above scheme, the mixed traffic control data includes the time and location of the mixed traffic conflict, the specific information of the conflicting vehicles, the environmental conditions of the conflict and the traffic control strategy to avoid the mixed traffic conflict; the specific information of the conflicting vehicles includes the vehicle type, speed and driving direction; the environmental conditions of the conflict include weather, visibility and yard layout; the traffic control strategy to avoid the mixed traffic conflict includes that when the IGV and the external container truck interact at the intersection, the external container truck has priority to pass, when the same type of vehicles interact, the right turn becomes the left turn, and the left turn becomes the straight go, when two vehicles of the same type pass through the intersection in the same way and conflict occurs, the container truck on the main road goes first, and the time difference between any two vehicles passing the same position point should be greater than the safe time distance for passage.

[0024] According to the above scheme, in step S4, the mixed dispatching process of IGV and external container truck is designed, including the following steps:

[0025] S401, the terminal operation management system dispatches horizontal transportation tasks;

[0026] S402, without considering the time window conflict, the vehicle management agent dispatches the IGV and the external container truck to perform horizontal transportation operations, determines the vehicle driving path, and generates the initial path time window;

[0027] When the kth vehicle (k∈[1, K]) performs horizontal transport task n(n∈[1, N]), from the moment the vehicle enters the road section s At the beginning, the moment when the vehicle leaves the road section s End, the time window t when vehicle k performs horizontal transportation task on road section s kns for When vehicle k performs horizontal transport task n without passing through road section s or horizontal transport task n does not use vehicle k, t kns is [0, 0]; the time window set of all vehicles on road section s is T(s) = {t 11s , t 21s , ..., t K1s , ..., t 1Ns , t 2Ns , ..., t KNs}, t kns The order of indicates the order in which vehicles pass through road section s;

[0028] S403, determining whether the time interval between the IGV and the external container truck arriving at the same location node in the road network has a conflict that does not meet the safety time interval;

[0029] If the card K from Enter the initial node i of section s at time During the time period, the usability of node i is occupied by vehicle k. When the current vehicle enters, other vehicles cannot pass through node i. When the current vehicle leaves node i, the next vehicle k′ cannot enter node i at the same time. k′ needs to maintain a certain safety time distance with k. To avoid mixed traffic conflicts, the time difference between two vehicles arriving at the same node in the road network is greater than the sum of the travel time of the safety distance and the travel time of the vehicle at the node:

[0030]

[0031] The detection time t and the vehicle node pointer p are introduced to detect mixed traffic conflicts; starting from the initial time t=0, the mixed traffic conflicts are detected step by step along the time axis to determine whether there are the same road section nodes or the sum of the travel time of the node with a time difference less than the safe spacing and the vehicle's travel time at the node among the nodes pointed to by the pointer p;

[0032] S404, if there is a conflict, the vehicle management agent issues a stop and wait instruction to the conflicting vehicle;

[0033] For vehicles whose pointers p point to the same node or whose node time difference is less than the safe time headway, two-to-two judgments are made, and the conflicting nodes, conflicting vehicles at the nodes, and the types of the corresponding mixed traffic conflicts are recorded;

[0034] S405, the road network agent gives a traffic plan for the IGV and the external container truck according to the traffic control strategy based on the large language model, and adjusts the path time window;

[0035] Conflict handling for mixed traffic conflict nodes: According to the traffic control strategy based on the large language model, the traffic plan for IGV and external container truck is given, and the vehicle with higher priority in the conflict node is selected, the time window is kept unchanged, and the time window of other vehicles with lower priority is moved to the right until all conflicts at the detection time t are resolved; when the time when a vehicle pointer p points to a node is less than or equal to the detection time t, it is determined whether there is a section node after the node pointed by the vehicle pointer. If so, the vehicle pointer is moved back one position. If there is no subsequent node, the vehicle completes the mixed traffic conflict detection, deletes the current vehicle pointer p, and the subsequent mixed traffic conflict judgment does not include the vehicle;

[0036] S406: If there is no conflict, the IGV and the external container truck complete the horizontal transportation task according to the specified route.

[0037] The implementation of the large language model-based intelligent interactive mixed dispatching method of IGV and external container trucks of the present invention has the following beneficial effects:

[0038] 1. In the present invention, the scheme of physically separating the horizontal transport vehicles in the port from the external container trucks is cancelled, the effective area of ​​the yard is expanded, the mixed traffic of IGV and external container trucks is realized, the operation process is optimized, and the collection and distribution efficiency and the horizontal transportation capacity of the port area are improved;

[0039] 2. The horizontal transportation system of the present invention improves the efficiency of equipment loading and unloading operations, reduces the equipment demand, and reduces the cost of port equipment construction. It is suitable for the construction or expansion of automated container terminals with horizontal yard layout;

[0040] 3. The mixed traffic dispatching method of IGV and external container trucks proposed in the present invention is a major innovation in the horizontal transportation system of automated container terminals. It has broad application prospects in the construction of automated container terminals and the upgrading and transformation of traditional terminals, and has good economic value and social benefits. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] The present invention will be further described below with reference to the accompanying drawings and embodiments, in which:

[0042] Figure 1 It is a structural diagram of the existing automated container yard area;

[0043] Figure 2 It is a structural schematic diagram of the automated container yard area of ​​the present invention;

[0044] Figure 3 It is a schematic diagram of the conflict between the IGV and the external container truck of the present invention;

[0045] Figure 4 It is a flow chart of traffic control based on multi-agent interaction of the present invention;

[0046] Figure 5 It is a traffic control flow chart based on a large language model of the present invention;

[0047] Figure 6 This is a flow chart of the mixed dispatching of IGVs and external container trucks according to the present invention. DETAILED DESCRIPTION

[0048] In order to have a clearer understanding of the technical features, purposes and effects of the present invention, specific embodiments of the present invention are now described in detail with reference to the accompanying drawings.

[0049] The large language model-based intelligent interactive mixed traffic scheduling method of IGVs and foreign container trucks of the present invention specifically includes the following steps: S1, arranging the automated container yard area; S2, determining the mixed traffic flow of yard operations; S3, formulating traffic control rules based on multi-agent interaction, including: formulating multi-agent interactive communication rules and formulating traffic control rules based on the large language model; S4, designing the mixed traffic scheduling process of IGVs and foreign container trucks.

[0050] The container area of ​​the yard is arranged parallel to the dock shoreline. Both IGVs and external container trucks need to enter the yard area to cooperate with the rubber-tyred gantry crane to complete the loading and unloading of containers. The mixed traffic flow of yard operations occurs in the overtaking lane and the main road of the yard. IGVs and external container trucks have mixed traffic conflicts on the main road of the yard, the overtaking lane and the intersection. The formulation of traffic control rules based on multi-agent interaction includes the formulation of multi-agent interactive communication and the formulation of traffic control rules based on large language models. The mixed traffic scheduling process of IGVs and external container trucks includes the terminal operation management system distributing horizontal operation tasks, the vehicle management agent allocating tasks, and the road network agent coordinating the mixed transportation of IGVs and external container trucks.

[0051] like Figure 2 As shown in the figure, the automated container yard area includes the yard container area, rubber-tyred gantry cranes and roads within the yard. The yard container area is arranged parallel to the dock shoreline. IGVs drive counterclockwise on the seaside lane to enter the yard container area, and foreign container trucks drive clockwise on the landside lane to enter the yard container area. The yard operation has mixed traffic flow, and the roads within the port area are all one-way lanes. IGVs and foreign container trucks have traffic intersections on the yard overtaking lane and the main road, and drive along the mixed route to the rubber-tyred gantry crane to carry out container loading and unloading operations. When IGVs and foreign container trucks enter the same section of road, they produce following driving; when passing the same intersection, intersection interaction occurs.

[0052] like Figure 3 As shown in the figure, mixed traffic conflicts occur when vehicles are mixed. When IGV and external trucks interact at the intersection, the two vehicles have conflicts in right of way and safe crossing time at the intersection. Figure 3 No. 1, both vehicles are driving straight through their own lanes. If the time difference between the two vehicles reaching the intersection of the lanes is less than the safe time distance, the two vehicles will collide. When the IGV and the external container truck are driving on the same road section, there is a conflict in the safe time distance between the two vehicles. Figure 3 Sequence number 14, two vehicles are driving straight through the same lane. If the time difference between the front vehicle and the rear vehicle passing the same position on the road section is less than the safe time distance required for following, the rear vehicle cannot brake urgently and a conflict will occur.

[0053] like Figure 4 As shown in the figure, the multi-agent interactive communication rules are formulated, and the vehicle management agent obtains the geographic location, transportation task and other status data of all IGVs and external container trucks in real time, determines the time when the vehicle enters and leaves the road section, and generates the initial path time window of each vehicle. At the same time, the road network agent interacts with the vehicle management agent, and the road network agent detects mixed traffic conflict information in the road network, uses the traffic control strategy based on the large language model to adjust the path time window, and passes it to the vehicle management agent for processing. Figure 3In case of mixed traffic conflicts, when IGV and external container trucks are traveling at the same intersection, the road network agent will uniformly control the opening of intersections based on information such as vehicle priority, right of way, and safe time interval, and intelligently control the passage of IGV and external container trucks, effectively improving the passage efficiency of container trucks and shortening the overall operation time.

[0054] The definitions and functions of each agent are as follows:

[0055] (1) Vehicle Management Agent. The vehicle management agent manages all IGVs and external container trucks in the system, obtains and summarizes the status data of all vehicles, such as numbers, locations, transportation tasks, fault information, etc., in real time, and updates them in a timely manner; the vehicle management agent communicates and interacts with the road network agent. The vehicle management agent receives the route time window adjustment instruction transmitted by the road network agent and transmits it to the corresponding vehicle agent for execution. At the same time, the vehicle management agent transmits the driving position and route time window of the IGV and external container truck to the road network agent for real-time monitoring and adjustment.

[0056] (2) IGV Agent. IGV Agent is the mapping of IGV in the system. Each IGV corresponds to an IGV Agent to complete the specified horizontal transportation task. IGV Agent is a hybrid intelligent structure that can plan autonomous behavior and respond to events. IGV Agent senses the external environment, interacts with other agents, and updates IGV position, driving path and other status data. At the same time, IGV Agent transmits its own status information such as IGV position and path time window to the vehicle management agent in a timely manner. The vehicle management agent provides the IGV Agent with road network information for path adjustment.

[0057] (3) External container card Agent. External container card Agent is the mapping of external container cards in the system, and its function is the same as IGV.

[0058] (4) Road Network Agent. The road network agent is a map of the overall path network in the system. It detects path time window conflict information in the road network and updates the status data in a timely manner. It interacts with the vehicle management agent and uses intelligent traffic control strategies to adjust the path time window for conflict information in the road network and transmits it to the vehicle management agent for processing. At the same time, it obtains and updates the location and path time window of the IGV and external container truck from the vehicle management agent in real time for real-time monitoring and adjustment.

[0059] like Figure 5 As shown, formulating traffic control rules based on a large language model includes the following steps:

[0060] S301, data collection: To solve the mixed traffic problem of IGV and foreign container trucks at horizontal container terminals, obtain mixed traffic control data, divide training samples, and obtain training set, validation set and test set;

[0061] S302, Model architecture selection: Select a large language model based on Transformer to effectively understand natural language text, process structured data, and generate traffic control strategies;

[0062] S303, model training: input the training set data into the model, generate the predicted traffic control strategy through forward propagation, calculate the loss function between the predicted and the actual results, update the model parameters through the back propagation algorithm, and iterate the training until the model achieves satisfactory performance on the validation set;

[0063] S304, fine-tuning and iterative optimization: Based on the meta-optimizer, use the mixed row conflict data to fine-tune the model with a passed verification result to improve its accuracy on the task;

[0064] S305, Model evaluation and deployment: Use the test set to evaluate the performance of the model to ensure that it has sufficient accuracy and generalization ability in practical applications. Once the evaluation passes, the model is deployed to the actual application to handle mixed row conflicts in real time.

[0065] Mixed traffic control data mainly include the time and location of mixed traffic conflicts; specific information of conflicting vehicles, such as vehicle type, speed, and driving direction; environmental conditions of the conflict, such as weather, visibility, and yard layout; traffic control strategies to avoid mixed traffic conflicts, such as when IGVs and external container trucks interact at intersections, external container trucks have priority, and when vehicles of the same type interact, right turns become left turns, and left turns become straight. When two vehicles of the same type pass through an intersection in the same way and conflict occurs, the container truck on the main road goes first, and the time difference between any two vehicles passing the same position point should be greater than the safe time distance for passage.

[0066] like Figure 6 As shown in the figure, the mixed dispatching process of IGV and external container truck is designed, including the following steps:

[0067] S401, the terminal operation management system dispatches horizontal transportation tasks;

[0068] S402, without considering the time window conflict, the vehicle management agent dispatches the IGV and the external container truck to perform horizontal transportation operations, determines the vehicle driving path, and generates the initial path time window;

[0069] When the kth vehicle (k∈[1, K]) performs horizontal transport task n(n∈[1, N]), from the moment the vehicle enters the road section s At the beginning, the moment when the vehicle leaves the road section s End, the time window t when vehicle k performs horizontal transportation task on road section s kns for When vehicle k performs horizontal transport task n without passing through road section s or horizontal transport task n does not use vehicle k, t kns = [0, 0]. The time window set of all vehicles on road section s is T(s) = {t 11s , t 21s , ..., t K1s , ..., t 1Ns , t 2Ns , ..., t KNs}, t kns The order indicates the order in which vehicles pass through road section s.

[0070] S403, determining whether the time interval between the IGV and the external container truck arriving at the same location node in the road network has a conflict that does not meet the safety time interval;

[0071] If the card K from Enter the initial node i of section s at time During the time period, the usable right of node i is occupied by vehicle k. When the current vehicle enters, other vehicles cannot pass through node i. When the current vehicle leaves node i, the next vehicle k′ cannot enter node i at the same time. k′ needs to maintain a certain safety time distance with k. To avoid mixed traffic conflicts, the time difference between two vehicles arriving at the same node in the road network should be greater than the sum of the travel time of the safety distance and the travel time of the vehicle at the node.

[0072]

[0073] The detection time t and the vehicle node pointer p are introduced to detect mixed traffic conflicts. Starting from the initial time t=0, the mixed traffic conflicts are detected step by step along the time axis. It is determined whether there are the same road section nodes or the sum of the travel time of the node with a time difference less than the safe spacing and the vehicle's travel time at the node among the nodes pointed to by the pointer p.

[0074] S404, if there is a conflict, the vehicle management agent issues a stop and wait instruction to the conflicting vehicle;

[0075] For vehicles whose pointers p point to the same node or whose node time difference is less than the safe time distance, pairwise judgment is performed, and the node where the conflict occurs, the vehicles that conflict at the node, and the type of the corresponding mixed traffic conflict are recorded.

[0076] S405, the road network agent gives a traffic plan for the IGV and the external container truck according to the traffic control strategy based on the large language model, and adjusts the path time window;

[0077] Conflict handling for mixed traffic conflict nodes: According to the traffic control strategy based on the large language model, the traffic plan for IGV and external container trucks is given, and the vehicle with higher priority in the conflict node is selected, the time window is kept unchanged, and the time window of other vehicles with lower priority is moved to the right until all conflicts at the detection time t are resolved. When the time when a vehicle pointer p points to a node is less than or equal to the detection time t, it is determined whether there is a section node after the node pointed by the vehicle pointer. If so, the vehicle pointer is moved back one position. If there is no subsequent node, the vehicle completes the mixed traffic conflict detection, deletes the current vehicle pointer p, and the subsequent mixed traffic conflict judgment does not include the vehicle.

[0078] S406: If there is no conflict, the IGV and the external container truck complete the horizontal transportation task according to the specified route.

[0079] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation modes, which are merely illustrative rather than restrictive. Under the guidance of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the present invention and the claims, all of which are within the protection of the present invention.

Claims

1. An intelligent interactive mixed traffic scheduling method for IGVs and external container trucks based on a large language model, characterized in that: The following steps are involved: S1. Arrange the automated container yard area; S2. Determine the mixed traffic flow for yard operations; S3, Formulate traffic control rules based on multi-agent interaction, including: rule formulation multi-agent interactive communication and formulation of traffic control rules based on large language models; S4. Design the mixed dispatching process of IGV and external container trucks.

2. The intelligent interactive mixed traffic scheduling method of IGV and external container truck based on large language model according to claim 1 is characterized in that: In step S1, the automated container yard area includes a yard container area, a rubber-tyred gantry crane and a road in the yard; the yard container area is arranged parallel to the wharf shoreline, the IGV drives counterclockwise in the seaside lane to enter the yard container area, and the external container truck drives clockwise in the landside lane to enter the yard container area.

3. The intelligent interactive mixed traffic scheduling method of IGV and external container truck based on large language model according to claim 1 is characterized in that: In step S2, the method for confirming the mixed traffic flow for yard operations is as follows: all roads in the port area are one-way lanes. When the IGV and the external container truck have traffic intersections in the yard overtaking lane and the main road, they travel along the mixed route to the rubber-tyred gantry crane to carry out container loading and unloading operations; when the IGV and the external container truck enter the same section of road, they produce following driving; when the IGV and the external container truck pass the same intersection, intersection interaction occurs.

4. The intelligent interactive mixed traffic scheduling method of IGV and external container truck based on large language model according to claim 3 is characterized in that: When the IGV and the external container truck interact at the intersection, there is a conflict in the right of way and the safe time distance of the intersection, resulting in a mixed traffic conflict. The IGV and the external container truck both go straight through their own lanes; when the time difference between the IGV and the external container truck when they reach the intersection of the lanes is less than the safe time distance, the two vehicles collide; the IGV and the external container truck both go straight through the same lane, and the two vehicles are in a following driving situation. If the time difference between the front vehicle and the rear vehicle passing the same position of the road section is less than the safe time distance required for following, the rear vehicle cannot brake urgently and a conflict occurs.

5. The intelligent interactive mixed traffic scheduling method of IGV and external container truck based on large language model according to claim 1 is characterized in that: In step S3, the rules for multi-agent interactive communication are formulated as follows: the vehicle management agent obtains the geographic location and transportation task status data of all IGVs and external container trucks in real time, determines the time when the vehicles enter and leave the road section, and generates the initial path time window of each vehicle; The road network agent interacts with the vehicle management agent. The road network agent detects mixed traffic conflict information in the road network, uses the traffic control strategy based on the large language model to adjust the path time window, and passes it to the vehicle management agent for processing; When IGV and external container trucks are traveling at the same intersection, the road network agent will uniformly macro-control the opening of the intersection according to vehicle priority, right of way, and safe time interval, and intelligently control the passage of IGV and external container trucks.

6. The method for intelligent interactive mixed traffic scheduling of IGVs and external container trucks based on a large language model according to claim 1 is characterized in that: In step S3, traffic control rules based on the large language model are formulated, including the following steps: S301, data collection: To solve the mixed traffic problem of IGV and foreign container trucks at horizontal container terminals, obtain mixed traffic control data, divide training samples, and obtain training set, validation set and test set; S302, Model architecture selection: Select a large language model based on Transformer to understand natural language text, process structured data, and generate traffic control strategies; S303, model training: input the training set data into the model, generate the predicted traffic control strategy through forward propagation, calculate the loss function between the predicted and the actual results, update the model parameters through the back propagation algorithm, and iterate the training until the model achieves satisfactory performance on the validation set; S304, fine-tuning and iterative optimization: Based on the meta-optimizer, use the mixed row conflict data to fine-tune the model with a passed verification result to improve its accuracy on the task; S305, model evaluation and deployment: Use the test set to evaluate the performance of the model. When the evaluation passes, the model is deployed to the actual application to handle mixed row conflict problems in real time.

7. The intelligent interactive mixed traffic scheduling method of IGV and external container truck based on large language model according to claim 6 is characterized in that: The mixed traffic control data includes the time location of the mixed traffic conflict, the specific information of the conflicting vehicles, the environmental conditions of the conflict and the traffic control strategy to avoid the mixed traffic conflict; the specific information of the conflicting vehicles includes the vehicle type, speed and driving direction; the environmental conditions of the conflict include weather, visibility and yard layout; the traffic control strategy to avoid the mixed traffic conflict includes that when the IGV and the external container truck interact at the intersection, the external container truck has priority to pass, when the same type of vehicles interact, right turn becomes left turn, left turn becomes straight, when two vehicles of the same type pass through the intersection in the same way and conflict occurs, the container truck on the main road goes first, and the time difference between any two vehicles passing the same position point should be greater than the safe time distance for passage.

8. The method for intelligent interactive mixed traffic scheduling of IGVs and external container trucks based on a large language model according to claim 1 is characterized in that: In step S4, a mixed dispatching process of IGV and external container trucks is designed, including the following steps: S401, the terminal operation management system dispatches horizontal transportation tasks; S402, without considering the time window conflict, the vehicle management agent dispatches the IGV and the external container truck to perform horizontal transportation operations, determines the vehicle driving path, and generates the initial path time window; When the kth vehicle (k∈[1, K]) performs horizontal transport task n(n∈[1, N]), from the moment the vehicle enters the road section s At the beginning, the moment when the vehicle leaves the road section s End, the time window t when vehicle k performs horizontal transportation task on road section s kns for When vehicle k performs horizontal transport task n without passing through road section s or horizontal transport task n does not use vehicle k, t kns is [0, 0]; the time window set of all vehicles on road section s is T(s) = {t 11s , t 21s , ..., t K1s , ..., t 1Ns , t 2Ns , ..., t KNs }, t kns The order of indicates the order in which vehicles pass through road section s; S403, determining whether the time interval between the IGV and the external container truck arriving at the same location node in the road network has a conflict that does not meet the safety time interval; If the card K from Enter the initial node i of section s at time During the time period, the usability of node i is occupied by vehicle k. When the current vehicle enters, other vehicles cannot pass through node i. When the current vehicle leaves node i, the next vehicle k′ cannot enter node i at the same time. k′ needs to maintain a certain safety time distance with k. To avoid mixed traffic conflicts, the time difference between two vehicles arriving at the same node in the road network is greater than the sum of the travel time of the safety distance and the travel time of the vehicle at the node: The detection time t and the vehicle node pointer p are introduced to detect mixed traffic conflicts; starting from the initial time t=0, the mixed traffic conflicts are detected step by step along the time axis to determine whether there are the same road section nodes or the sum of the travel time of the node with a time difference less than the safe spacing and the vehicle's travel time at the node among the nodes pointed to by the pointer p; S404, if there is a conflict, the vehicle management agent issues a stop and wait instruction to the conflicting vehicle; For vehicles whose pointers p point to the same node or whose node time difference is less than the safe time headway, two-to-two judgments are made, and the conflicting nodes, conflicting vehicles at the nodes, and the types of the corresponding mixed traffic conflicts are recorded; S405, the road network agent gives a traffic plan for the IGV and the external container truck according to the traffic control strategy based on the large language model, and adjusts the path time window; Conflict handling for mixed traffic conflict nodes: According to the traffic control strategy based on the large language model, the traffic plan for IGV and external container truck is given, and the vehicle with higher priority in the conflict node is selected, the time window is kept unchanged, and the time window of other vehicles with lower priority is moved to the right until all conflicts at the detection time t are resolved; When the time when a vehicle pointer p points to a node is less than or equal to the detection time t, determine whether there is a road section node after the node pointed to by the vehicle pointer. If so, move the vehicle pointer back one position. If there is no subsequent node, the vehicle completes mixed traffic conflict detection, deletes the current vehicle pointer p, and does not include the vehicle in subsequent mixed traffic conflict judgments; S406: If there is no conflict, the IGV and the external container truck complete the horizontal transportation task according to the specified route.

Images