Railway and highway combined transport scheduling method and system based on digital twinning and multi-agent collaboration

Abstract

The application discloses a kind of based on digital twin and multi-agent collaborative highway-railway intermodal dispatching method and system, including the unique binding relationship of people, car, goods in transport order based on RFID binding algorithm is established, and it is synchronized to database;Real-time acquisition railway transport state data, highway transport state data and freight station operation data;The standardization processing of railway transport state data, highway transport state data and freight station operation data is carried out by multi-source data fusion center;Based on LSTM collaborative scheduling model, the connection time, route and stop position of highway vehicle are dynamically adjusted, and the highway vehicle rerouting and waiting time estimation scheme under the output of sudden situation are output;Railway carriage demand is predicted by LSTM, and the dynamic optimal matching of highway truck and railway carriage is dynamically matched based on mixed integer programming model, to generate the optimal matching scheme of highway truck and railway carriage, introduce RFID technology to realize the unique binding of people, car and goods, break information barrier, realize data real-time sharing and standardized interaction.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of logistics technology, and in particular to a highway-railway combined transport scheduling method and system based on digital twinning and multi-agent collaboration. BACKGROUND

[0002] The existing non-collaborative highway and railway independent transport scheduling system includes two independent systems of highway and railway. The highway system supports manual entry of the association information of goods and vehicles, tracks the vehicle position through GPS positioning and plans the transportation route, and the vehicle access to the station is manually verified by the staff to complete the access management. The railway system can formulate a train working diagram, coordinate the compartment allocation and loading and unloading plan of the freight station, and the information of goods arriving at the station is recorded and circulated in the internal system of the railway. The two systems handle the scheduling tasks of their respective fields: the highway system focuses on the vehicle monitoring and route arrangement of the highway transportation link, and the railway system focuses on the train operation and freight station operation management of the railway transportation, and the two systems are connected through manual information transmission. The information entry of the highway system depends on manual operation, and the information of goods arriving at the station in the railway system is only transmitted within the railway, and the verification of vehicle access to the station and the position guidance of goods connection are all completed through manual methods, and the scheduling adjustment needs to be coordinated by manual methods. In summary, the defects of the current technical solution are as follows:

[0003] 1) The matching reliability of people, vehicles and goods is insufficient, and the information sharing is lagging: the association information of goods and vehicles is manually entered, and there is no automatic verification mechanism, which is prone to problems such as "mismatch of goods and vehicles" and "invalid binding of driver and vehicle", like information errors when manually checking documents; at the same time, the data standards of railway and highway systems are not unified (the railway focuses on freight plan data and the highway focuses on vehicle dynamic data), which leads to the inability of real-time information exchange. For example, after the railway goods arrive at the station and complete unloading, the unloading status information needs to be manually transmitted to the highway system, and the highway carrier often waits for several hours due to the failure to obtain the information in time, which may cause misdelivery of goods or security loopholes.

[0004] 2) Dynamic scheduling response is delayed, and emergency collaboration capability is weak: in the face of railway train delay, highway traffic congestion and other conditions, the system lacks real-time information exchange channels across departments, and needs to be manually coordinated and adjusted. Due to the inability to quickly integrate railway capacity changes and highway vehicle distribution data, the average delay repair time of highway-railway combined transport caused by unexpected situations is longer, and the uncertainty of the binding relationship between drivers and vehicles further exacerbates the scheduling confusion.

[0005] 3) Resource matching is inefficient, and transport capacity is wasted: the supply and demand allocation of highway trucks and railway cars completely depends on manual estimation, and lacks intelligent matching mechanism. In the peak season, there is a shortage of railway carriages, and a large number of highway trucks return empty because there is no cargo to transport. In the off-season, there is a backlog of goods at the railway freight station, but the highway transport capacity is idle and cannot be transported in time, forming the phenomenon of "empty and pressure coexistence" of resource mismatch, and the operating cost is significantly increased. SUMMARY

[0006] The purpose of the present application is to provide a public and private joint transport scheduling method and system based on digital twin and multi-agent collaboration, which realizes the unique identification and automatic association of people, vehicles and goods by introducing RFID technology through digital twin model, establishes a real-time information interaction mechanism across systems; relying on the precise positioning algorithm to realize the precise matching of railway freight and highway vehicles, optimizing the automatic verification process of the field station; Construct a near-range monitoring and dynamic early warning system for people, vehicles and goods, and improve the collaborative response capability under unexpected conditions by combining intelligent scheduling models, so as to realize intelligent management and control of the whole process of public and private joint transport, improve transport efficiency, reduce operating costs, and strengthen safety protection.

[0007] The present application provides a public and private joint transport scheduling method based on digital twin and multi-agent collaboration, comprising

[0008] Receiving the transportation order submitted by the consignor, establishing the unique binding relationship of people, vehicles and goods in the transportation order based on the RFID binding algorithm and synchronizing to the database, and constructing the initial digital twin model including the transportation network, the freight station facility and the vehicle equipment;

[0009] Real-time collection of railway transportation state data, highway transportation state data and freight station operation data, the railway transportation state data including railway freight arrival time and railway car position, the highway transportation state data including real-time position and load state of highway vehicles, and the freight station operation data including vehicle parking position and loading and unloading equipment state in the freight station;

[0010] The railway transportation state data, highway transportation state data and freight station operation data are standardized by the multi-source data fusion center, breaking the information barrier between railway and highway, and generating real-time shared joint transport state information;

[0011] Based on the LSTM collaborative scheduling model, the joint transport state information is analyzed, when the railway delay, highway congestion, freight station operation abnormality and related unexpected conditions are detected, the connection time, route and parking position of the highway vehicle are dynamically adjusted, and the highway vehicle rerouting and waiting time estimation scheme under unexpected conditions is output;

[0012] The railway carriage demand is predicted through the LSTM, and the dynamic optimal matching of the road freight vehicles and the railway carriages is dynamically matched based on a mixed integer programming model, so that the shortest transportation distance, the highest time matching degree and the maximum loading rate are taken as objective functions to generate an optimal matching scheme of the road freight vehicles and the railway carriages.

[0013] As preferred, it further comprises:

[0014] A three-dimensional visual navigation system of the freight station is constructed, a centimeter-level digital twin model is established based on point cloud scanning to generate a centimeter-level accurate parking guide, and a partitioned dynamic traffic control strategy is combined to guide the vehicle to enter the designated parking position according to the preset route;

[0015] The position information of personnel, vehicles and goods is monitored in real time through high-precision positioning and safety monitoring network, and a safety distance real-time monitoring algorithm is combined, so that when it is detected that the personnel leave the post, the goods are separated from the safety range or the distance between the vehicle and the goods exceeds the threshold, a safety warning is triggered;

[0016] A cross-regional policy compliance evaluation model is constructed, core elements of the policy are extracted through NLP technology, and cross-regional transportation compliance is verified by combining a dynamic policy matching engine.

[0017] As preferred, the unique binding relationship of the person, the vehicle and the goods in the transportation order is established based on the RFID binding algorithm and is synchronized to the database, which further comprises:

[0018] The goods are configured with a physical label for storing the identification information of the goods, the road vehicle is configured with an identification device for reading the physical label, and the driver is equipped with an identity tag carrying the identification information of the driver;

[0019] The goods information of the physical label, the vehicle information of the identification device and the driver information of the identity tag are collected through the station automatic verification system, and the consistency of the three is verified based on a binding verification algorithm, and the binding verification algorithm is:

[0020]

[0021] Among them, RFID1, RFID2 and RFID3 are the RFID of the person, the vehicle and the goods respectively k identification information, db k is the pre-stored associated information in the database, and δ(·) is a matching function. When Mbind=1, it is determined as an effective binding.

[0022] A cross-system data interaction platform is constructed, based on a multimodal transport data exchange standard, data formats of railway freight plan data and highway vehicle dynamic data are unified through a data conversion interface, distributed data processing tools are used to clean the data in real time after unification of the data formats, to ensure consistency of key fields, the data conversion interface realizes automatic correspondence of XML and JSON fields through pre-defined mapping rules, and the key fields include cargo ID, arrival time, cargo location number and vehicle ID;

[0023] Based on a cargo RFID state change event triggered information pushing mechanism, when it is detected that the unloading of the cargo at the railway end is completed, the unloading time and cargo location information are automatically pushed to the vehicle terminal and the driver terminal of the associated highway vehicle.

[0024] Preferably, the standardization processing of the railway transport state data, the highway transport state data and the freight station operation data by the multi-source data fusion center comprises:

[0025] Real-time public and private state data is collected, including fusion of railway CTC system, highway GPS positioning, weather warning information and related multi-source real-time data streams;

[0026] ApacheFlink distributed processing framework is used to access the multi-source data stream in real time, and the data stream is processed through

[0027] IsolationForest anomaly detection algorithm is used to identify abnormal points in the state data, and the abnormal score calculation formula is:

[0028]

[0029] Wherein, h(x) is the path length of sample x in the isolated tree, E(h(x) is the expected value of the path length in the multiple isolated trees, c(n) is the path length normalization factor, which is used to eliminate the influence of sample size on the result, and n is the sample number;

[0030] Random forest missing value completion algorithm is used to fill the missing values in the public and private state data after removing the abnormal points, and the completion formula is:

[0031]

[0032] Wherein, x^ j is the predicted value of the feature column X j to be completed, N is the number of decision trees in the random forest, f i is the prediction function of the i-th decision tree, and X -j is the other feature column except X j .

[0033] Preferably, the analyzing the intermodal status information based on the LSTM collaborative scheduling model further comprises:

[0034] obtaining standardized intermodal status information, extracting emergency condition features, the emergency condition features including a railway train delay time, a highway congestion section length, and a freight station operation abnormal type;

[0035] inputting the emergency condition features into the LSTM collaborative scheduling model, and outputting a scheduling adjustment intensity A t , the calculation formula being:

[0036] A t =ω rail ·ΔT late +ω road ·L congest +ω bind ·S bind ,

[0037] wherein ΔT late is the railway train delay time, L congest is the highway congestion section length, S bind is the driver and vehicle binding stability, and w rail , w road , and w bind are weight coefficients.

[0038] Preferably, the dynamically matching the highway trucks and the railway carriages based on the mixed integer programming model is a dynamic optimal matching.

[0039] obtaining a railway carriage set, a highway truck set, and intermodal status information, the railway carriage set including a carriage ID, a load, a volume, and an available time, the highway truck set including a truck ID, a load, a volume, and a real-time position, and the intermodal status information including a railway freight arrival time and a highway vehicle connection time;

[0040] constructing a mixed integer programming model, a target function being a minimum total cost, the formula being:

[0041]

[0042] wherein C represents the truck set, C represents the carriage set, d ij represents a transportation distance of the truck i and the carriage j, t ij represents a time window matching degree, α, β, and γ represent weight coefficients, and s ij represents a freight-carriage specification adaptation degree; the freight-carriage specification adaptation degree s ij is obtained by collecting a total weight W good and a total volume V good of the freight through a freight RFID tag and a truck load Wtruck , volume V truck The calculation is as follows

[0043]

[0044] Solving the mixed integer programming model, obtaining the optimal matching scheme, and storing the scheme to the cloud database, synchronizing to the highway truck through the vehicle terminal and the driver APP.

[0045] As preferred, the construction of the freight station three-dimensional visualization navigation system further comprises:

[0046] Based on the point cloud scanning equipment, the freight station is modeled for the whole scene, and a centimeter-level digital twin model containing coordinate system mapping relationship is generated;

[0047] Deploy UWB positioning base station and RFID landmark at the entrance of the freight station and related key nodes of loading and unloading sites to form a spatial grid positioning reference;

[0048] Collect the vehicle position in real time through the vehicle-mounted UWB module, correct the positioning error combined with the RFID landmark data, and generate the vehicle driving track;

[0049] Dynamically plan the path through the freight station internal path planning cost function, assign a dedicated channel according to the vehicle type, and display dynamic guidance instructions on the LED screen.

[0050] As preferred, the steps of the freight station internal path planning cost function are:

[0051] Collect the distance data and real-time congestion coefficient of all road segments in the freight station;

[0052] Get the number of turns of each road segment;

[0053] Calculate the cost C(n) of each path inside the freight station, and the formula is:

[0054] C(n) = d(n)·(1+k cong ·ρ(n))+k turn ·t(n),

[0055] Wherein, d(n) is the path distance, ρ(n) is the road congestion coefficient (0-1), k ong is the congestion influence coefficient, t(n) is the number of turns, and k trun is the turn penalty coefficient;

[0056] Solve the minimum cost path through the path planning algorithm, and send the optimal path to the driver APP and the vehicle terminal.

[0057] As preferred, the construction of the cross-regional policy compliance evaluation model comprises:

[0058] Collecting policy texts of vehicle access permissions, cargo security standards and RFID technology application specifications in various provinces and cities across the country to generate a multi-region policy adaptation database;

[0059] Extracting policy core elements through NLP technology, including vehicle access permissions, cargo security standards and RFID technology application specifications, and establishing a mapping relationship between policies and technical requirements;

[0060] Calling a dynamic policy matching engine, when the vehicle dispatch plan involves cross-regional transportation, automatically querying the policy requirements of the passing regions according to the transportation route, and checking whether it meets the local regulations combined with the RFID information of the vehicle and the cargo;

[0061] Cross-regional policy compliance P comply Evaluation calculation:

[0062]

[0063] Where, p m is the mth regional policy requirement, s m is the actual satisfaction state of the vehicle fleet, δ(·) is the compliance function, M is the total number of policies involved, and P comply is greater than or equal to 0.9, which is determined as passable.

[0064] The application also provides a public and private transport combined dispatching system based on digital twinning and multi-agent collaboration, comprising:

[0065] A digital twinning model is used to receive transportation orders submitted by consignors, establish a unique binding relationship between people, vehicles and goods in the transportation order based on an RFID binding algorithm and synchronize it to a database, and construct an initial digital twinning model including a transportation network, cargo station facilities and vehicle equipment.

[0066] A data acquisition module is used to acquire real-time railway transportation state data, highway transportation state data and cargo station operation data, wherein the railway transportation state data includes railway cargo arrival time and railway compartment position, the highway transportation state data includes real-time highway vehicle position and loading state, and the cargo station operation data includes cargo station vehicle parking position and loading and unloading equipment state.

[0067] A multi-source data fusion center is used to standardize the railway transportation state data, highway transportation state data and cargo station operation data through the multi-source data fusion center, break down the information barriers between railways and highways, and generate real-time shared combined transportation state information.

[0068] The dynamic scheduling module is configured to analyze the intermodal transportation state information based on the LSTM collaborative scheduling model, dynamically adjust the connection time, route and stop position of the road vehicle when detecting railway delays, road congestion, abnormal cargo station operation and related sudden conditions, and output the road vehicle rerouting and waiting time estimation scheme under the sudden condition.

[0069] The resource matching module is configured to predict the railway carriage demand through the LSTM, dynamically match the optimal matching of the road freight vehicle and the railway carriage based on the mixed integer programming model, take the shortest transportation distance, the highest time matching degree and the maximum loading rate as the objective function, and generate the optimal matching scheme of the road freight vehicle and the railway carriage.

[0070] For the prior art, the present application has the following beneficial effects:

[0071] 1. Improve the reliability of human, vehicle and cargo matching and the efficiency of information sharing: by constructing a full-domain digital twin of intermodal transportation, the digital twin maps the static attributes and dynamic states of personnel, vehicles, goods, railway networks, highway networks and cargo station facilities in the physical world in real time through Internet of Things sensing terminals and data interfaces, realizes the unique binding of personnel, vehicles and goods by using an RFID intelligent identification system, eliminates information errors caused by manual checking by combining a cross-system data interaction platform and an RFID trigger type information pushing mechanism, breaks down the information barriers between railways and highways, greatly reduces the mismatch rate of goods and vehicles and the risk of misdelivery of goods, and significantly shortens the empty waiting time of road transporters caused by information lag.

[0072] 2. Enhance dynamic scheduling response and emergency collaborative ability: the multi-source real-time data fusion center and the collaborative scheduling model based on LSTM realize the rapid perception and processing of sudden conditions such as railway delays and road congestion, significantly shorten the average delay repair time, reduce the scheduling confusion caused by unclear binding relationship between drivers and vehicles, and greatly improve the cross-department emergency collaborative efficiency.

[0073] 3. Improve resource matching efficiency and reduce waste of transport capacity: the transport capacity demand prediction model and the intelligent matching algorithm based on mixed integer programming effectively solve the supply-demand mismatch problem of road freight vehicles and railway carriages, improve the utilization rate of railway carriages in peak seasons, reduce the empty load rate of road freight vehicles in off-seasons, and significantly reduce the operating costs of enterprises.

[0074] 4. Improve the operation efficiency and orderliness of the last mile: the three-dimensional visual navigation system and the dynamic traffic control of the cargo station provide accurate parking guidance for road vehicles, shorten the time spent by drivers in searching for goods, reduce the congestion rate inside the cargo station, make the connection between loading and unloading more orderly, and significantly improve the operation efficiency of the last mile.

[0075] 5. Strengthen the positioning accuracy and safety warning ability: high-precision positioning combined with safety distance real-time monitoring algorithm of safety monitoring network, realize the accurate tracking of personnel, vehicle, cargo position and effective control of safety distance, improve the positioning accuracy, shorten the warning response time of the driver off-duty and the cargo out of the safety range, greatly reduce the loss of goods and operation risk.

[0076] 6. Promote standardization and cross-regional adaptation: multi-region policy adaptation database and dynamic policy matching engine, solve the adaptation problem caused by policy differences in different regions, shorten the adaptation cycle of the system to new regional policies, improve the policy compliance rate, lay a solid foundation for the standardization promotion and large-scale application of technical solutions, and promote the development of intelligent and efficient intermodal transportation. BRIEF DESCRIPTION OF DRAWINGS

[0077] Figure 1 A functional architecture diagram of an intermodal transportation scheduling system based on digital twinning and multi-agent collaboration in an embodiment of the present application;

[0078] Figure 2 A technical architecture diagram of an intermodal transportation scheduling system based on digital twinning and multi-agent collaboration in an embodiment of the present application;

[0079] Figure 3 A business process example diagram of an intermodal transportation scheduling system based on digital twinning and multi-agent collaboration in an embodiment of the present application. DETAILED DESCRIPTION

[0080] To make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of the present application.

[0081] The term "comprising" and its variants as used herein are open-ended, that is "including but not limited to". The term "based on" is "at least partially based on". The term "an embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". Related definitions of other terms will be given in the following description.

[0082] Embodiment one

[0083] The application provides a public and railway combined transport scheduling method based on digital twinning and multi-agent cooperation, which is suitable for a public and railway combined transport scheduling system based on digital twinning and multi-agent cooperation, and a general function architecture diagram is shown in Figure 1. The method mainly comprises a public and railway combined transport fleet intelligent scheduling management and control system service portal, a public and railway combined transport fleet intelligent scheduling management and control system platform end, a mobile end and a third party docking, and is combined with business Figure 3The system comprises the following function layers: (1) a service portal layer, which provides: a shipper / forwarder registration interface, receives enterprise information input and qualification upload data; an order creation and tracking module, generates a combined transport order and triggers payment of a prepayment; (2) a dispatch platform layer connected with the service portal layer through an API gateway, which is used to: respond to an order creation instruction, perform RFID tag association order / waybill operation, and monitor RFID nodes throughout the process; based on a public road transfer operation linkage signal, trigger a yard RFID automatic verification and an abnormal order alarm; the dispatch platform layer comprises: an RFID full life cycle management module, which performs: responds to an order creation instruction of the service portal layer, generates a passive RFID tag code; through a vehicle-mounted read-write device state monitoring interface, real-time verifies the binding relationship between the tag and the waybill; a dynamic scheduling engine module, which performs: receives a goods RFID transfer signal reported by the mobile application layer, updates the waybill status; when the waybill status is abnormal, pushes a reprocessing instruction to the mobile application layer. The dispatch platform layer performs cross-system strategy cooperation: in the generation of a receiving plan stage, synchronizes road scheduling demand to a warehouse company system; when LSTM predicts that the station time deviation is greater than a threshold value, automatically adjusts the road capacity matching parameter and notifies the shipper and forwarder. (3) a mobile application layer, which communicates with the dispatch platform layer through a message middleware, and provides: a driver identity information authentication and RFID device binding interface; a task receiving and goods RFID transfer state reporting channel; the mobile application layer comprises an abnormality processing cooperation mechanism: when the goods RFID transfer verification fails, the mobile application layer triggers the following operations: sends an abnormal code and on-site image data to the dispatch platform layer; receives a public road transfer operation adjustment instruction issued by the dispatch platform layer; based on the new instruction, performs APP end connection navigation to a standby loading and unloading position. (4) a third-party interface layer, which interacts with the dispatch platform layer through a data exchange center, and realizes: reports road transport RFID track data to the Ministry of Transport; transmits payment vouchers to financial institutions and synchronizes electronic receipt to warehouse companies; the third-party interface layer realizes the following interactions: (a) interaction with financial institutions: the dispatch platform layer generates an electronic receipt hash value after RFID automatic verification; the third-party interface layer matches the electronic receipt with the payment prepayment voucher, and triggers financial institution settlement; (b) interaction with the Ministry of Transport: the dispatch platform layer extracts road transport RFID track key nodes (goods loading, breakpoint, and arrival verification); the third-party interface layer converts the track data according to the Ministry of Transport data specification and reports through a special VPN channel. The interaction between the service portal layer and the dispatch platform layer comprises: (a) the service portal layer responds to the qualification upload operation, and calls the qualification audit interface of the dispatch platform layer; (b) the dispatch platform layer synchronizes the following data to the service portal layer in real time: RFID automatic verification result and yard passage state; sub-waybill tracking information after order splitting; (c) the service portal layer unlocks task allocation permissions based on the payment prepayment state and pushes them to the dispatch platform layer.The interaction between the scheduling platform layer and the mobile application layer includes: when the scheduling platform layer distributes tasks to the mobile application layer, the following encrypted data packets are synchronized: digital twin navigation path coordinate sequence; expected handover time window of goods RFID tag; mobile application layer real-time execution: collecting goods RFID handover evidence data through handheld terminal; packaging and returning transportation track and RFID state tracking record to scheduling platform layer.

[0084] Those skilled in the art can understand that the system comprises: 1. Intelligent scheduling and control system service portal of combined road and rail truck fleet: freight owner / forwarder registration, enterprise information maintenance, qualification upload, audit tracking, order creation, order tracking, information publishing, online payment. 2. Intelligent scheduling and control system platform of combined road and rail truck fleet: associated RFID device, RFID tag life cycle management, RFID terminal operation permission configuration, vehicle-mounted read-write device state monitoring, waybill state information update, order splitting, RFID tag associated order / waybill, abnormal order alarm, full-process RFID node record, station RFID automatic verification, highway transportation RFID track monitoring, combined road and rail handover operation linkage, interactive RFID data, data exchange center connection, waybill state abnormality triggering. 3. Mobile terminal: driver registration and login, identity information authentication, handheld terminal binding, RFID device binding, task receiving, goods RFID handover, RFID data uploading, transportation track viewing. 4. Third-party connection: connection with the Ministry of Transport, truck fleet, freight owner and forwarder, financial institutions, warehousing companies. See Figure 2 As shown, the project planning supports the access of multiple terminals, is designed according to the actual user's role and use scenario, and supports PC terminal, mobile phone terminal, handheld device, WeChat applet, etc. The UI layer mainly displays the core technology system for the user terminal using HTML5, VUE related framework and VUE related components, etc. The access layer is mainly responsible for the data docking between internal systems of the platform and the gateway access of the related subsystems of the group and the third party, mainly including API gateway and load balancing. The business service layer includes business end and application end, the business end is responsible for the scheduled task processing, API interface, message pushing, short message task, organization structure and other public services of the platform; the application end is mainly for each business subsystem. The data persistence layer mainly includes data cache processing, transaction, data synchronization, file service and NOSQL non-relational database. The data storage layer supports Oracle enterprise database, MYSQL database, domestic database and other databases. The running environment supports domestic operating system, Windows operating system, Linux operating system, etc., and includes JVM JAVA running environment, Asp.netCore running environment, Docker container and related Web application server. The infrastructure layer mainly includes related network services, security services, storage, servers and related hardware devices.

[0085] Referring to Figure 3 As shown, the implementation includes the following:

[0086] S1: Receive the transport order submitted by the consignor terminal, establish the unique binding relationship of people, vehicles and goods in the transport order based on the RFID binding algorithm and synchronize to the database, build an initial digital twin model containing the transport network, freight station facilities and vehicle equipment, realize the full-factorial and high-faithful mapping from the physical world to the digital world; build a multimodal transport global digital twin, which maps the static attributes and dynamic states of personnel, vehicles, goods, railway network, highway network and freight station facilities in the physical world in real time through Internet of Things sensing terminals and data interfaces. The core principle of this step is “virtual-real mapping” and “data-driven”. The construction of the digital twin model is not to create a simple three-dimensional visualization scene, but to establish a virtual entity that corresponds one-to-one with the physical world, is bidirectionally connected, and can be simulated and calculated. Its construction process follows the following logic:

[0087] Static geometric model: Through technologies such as laser point cloud scanning, CAD drawing import, photogrammetry, etc., a high-precision three-dimensional model of the transport network (railway, highway), freight station facilities (platform, track, warehouse, passageway), and vehicle equipment is pre-constructed. Dynamic business data: Through RFID binding, order analysis, GPS positioning, etc., the real-time business status and attributes of people, vehicles and goods are obtained. The dynamic business data is “mounted” or “injected” as attributes or states to the corresponding static geometric model, so that the virtual model “comes to life” and becomes a digital twin that truly reflects the current state of the physical world. RFID binding event triggers the key instruction for instantiation of the digital twin. When the order is received and binding is completed, the digital twin model contains the pre-set static environment model, and according to the current order and binding relationship, a virtual entity representing a specific person, vehicle or goods in this transport task is created or activated in the virtual space, and is given an initial state and attribute, thereby starting the whole life cycle mirroring and monitoring of the entire transport task. In one way, a high-precision freight station digital twin sub-model is constructed, based on laser point cloud scanning and computer vision fusion modeling, a three-dimensional virtual environment with centimeter-level precision is generated, which is used to simulate and guide the precise parking of vehicles; a safety monitoring digital twin sub-model is constructed, which monitors the positions and behaviors of personnel, vehicles and goods in real time in the virtual space through the fusion of UWB positioning and AI vision data, and executes safety distance violation judgment; a policy and regulation digital twin sub-model is constructed, which extracts policy elements in various places through NLP technology to form a structured knowledge base, providing compliance decision basis for the multi-agent system. In the digital twin, a corresponding virtual twin is created for each physical entity (person, vehicle, goods); the binding verification algorithm is used to verify the consistency of the mapping relationship between the physical world data and the virtual twin, ensuring the data authenticity of the digital twin.

[0088] To solve the problem of insufficient reliability of matching between people, vehicles and goods, and the problem of information sharing lag, the specific steps are as follows:

[0089] The physical label storing the goods identification information is configured for the goods, the identification device for reading the physical label is configured for the road vehicle, and the identity card carrying the driver identification information is equipped for the driver; by introducing the RFID intelligent identification system, the passive RFID label storing the goods ID and specification information is pasted for the goods, the RFID reader is installed on the vehicle, and the identity card integrated with RFID is equipped for the driver, the unique binding of people, vehicles and goods is realized through the station automatic verification system, and the binding information is uploaded to the cloud database in real time.

[0090] The goods information of the physical label, the vehicle information of the identification device and the driver information of the identity card are collected by the station automatic verification system, and the consistency of the three is verified based on the binding verification algorithm, and the binding verification algorithm is:

[0091]

[0092] Among them, RFID1, RFID2 and RFID3 are the RFID of people, vehicles and goods respectively k identification information, db k is the pre-stored associated information in the database, δ(·) is a matching function, which returns 1 if consistent, otherwise 0, and Mbind=1 is determined as valid binding;

[0093] A cross-system data interaction platform is constructed, based on the national multimodal transport data exchange standard, the data format of railway freight plan data and road vehicle dynamic data is unified through a data conversion interface, and distributed data processing tools (ApacheNiFi) are used to clean the data in real time after the data is unified, to ensure the consistency of the key fields, the data conversion interface realizes the automatic correspondence of XML and JSON fields through pre-defined mapping rules, and the key fields include goods ID, arrival time, position number and vehicle ID;

[0094] Based on the information push mechanism triggered by the change of the goods RFID state, when the unloading completion state of the goods at the railway end is detected, that is, when the RFID reader of the railway freight station detects the unloading completion of the goods (identifies that the goods RFID tag leaves the carriage), the unloading time and the cargo position information are automatically pushed to the vehicle-mounted terminal and the driver terminal of the associated highway vehicle through the Kafka message queue and the MQTT protocol, thereby effectively avoiding the empty waiting of the highway vehicle. Through the RFID intelligent identification system, the unique binding of people, vehicles and goods is realized. Combined with the cross-system data interaction platform and the RFID triggered information push mechanism, the information error caused by manual checking is eliminated, the information barrier between the railway and the highway is broken, the mismatch rate of goods and vehicles and the risk of misoperation of goods are greatly reduced, and the empty waiting time of the highway carrier caused by information lag is significantly shortened.

[0095] S2: Real-time collection of railway transportation state data, highway transportation state data and freight station operation data, the railway transportation state data including railway goods arrival time, railway carriage position, the highway transportation state data including highway vehicle real-time position, loading state, and the freight station operation data including vehicle parking position in the freight station, and unloading equipment state;

[0096] S3: Standardization processing of the railway transportation state data, the highway transportation state data and the freight station operation data by the multi-source data fusion center, breaking the information barrier between the railway and the highway, generating real-time shared intermodal state information, in order to solve the problems of dynamic scheduling response lag and weak emergency coordination ability, specifically including:

[0097] Collecting real-time public and private state data, the real-time public and private state data including fusion of railway CTC system, highway GPS positioning, weather warning information and related multi-source real-time data flow; integrating railway CTC system, highway GPS positioning data, weather warning information and other multi-source real-time data flow, using ApacheFlink for distributed real-time processing, ensuring data quality through IsolationForest anomaly detection algorithm and random forest missing value completion algorithm in the data cleaning stage, constructing a real-time dynamic scheduling system, effectively solving the problem of dynamic scheduling response lag in public and private intermodal transportation.

[0098] Using ApacheFlink distributed processing framework to access multi-source data flow in real time, through

[0099] IsolationForest anomaly detection algorithm identifies abnormal points in state data, and the abnormal score calculation formula is:

[0100]

[0101] wherein h(x) is the path length of sample x in the isolated tree, E(h(x) is the expected value of the path length in multiple isolated trees, c(n) is a path length normalization factor for eliminating the influence of sample size on the result, and n is the sample quantity; the normalization factor calculation formula is: wherein H(k) represents a harmonic series, and the approximate calculation formula is H(k)≈ln(k)+0.5772, and n represents the number of sub-samples used when the isolated tree is constructed.

[0102] The missing value filling formula of the random forest missing value filling algorithm is:

[0103]

[0104] wherein x^ j is the predicted value of the feature column X j to be filled, N is the number of decision trees in the random forest, f i is the prediction function of the i-th decision tree, and X -j is the other feature column except X j .

[0105] S4: analyzing the intermodal status information based on the LSTM collaborative scheduling model, dynamically adjusting the connection time, route and stopping position of the highway vehicle when detecting railway delay, highway congestion, abnormal cargo station operation and related sudden conditions, and outputting the highway vehicle rerouting and waiting time estimation scheme under the sudden condition, specifically including the following:

[0106] Obtaining the standardized intermodal status information, extracting the sudden condition features, and the sudden condition features including the railway train delay duration, highway congestion section length and cargo station operation abnormal type;

[0107] Inputting the sudden condition features into the LSTM collaborative scheduling model, and outputting the scheduling adjustment intensity A t , the calculation formula is:

[0108] A t =ω rail ·ΔT late +ω road ·L congest +ω bind ·S bind ,

[0109] wherein ΔT late is the railway train delay duration, L congest is the highway congestion section length, S bind is the driver and vehicle binding stability, and w rail , w road , w bind are weight coefficients.

[0110] At the same time, a burst condition-based collaborative scheduling model is developed based on LSTM neural network, which inputs train delay duration, congestion section length, vehicle real-time position and other features, introduces attention mechanism to focus on key influencing factors, and outputs highway vehicle diversion, waiting time estimation and other adjustment schemes. The scheme is pushed to railway dispatchers and highway fleet managers through a cross-department collaborative platform to achieve rapid response. The multi-source real-time data fusion center and the LSTM-based collaborative scheduling model realize rapid sensing and processing of railway delay, highway congestion and other burst conditions, significantly shorten the average delay repair time, reduce the scheduling confusion caused by unclear driver and vehicle binding relationship, and greatly improve the cross-department emergency collaboration efficiency.

[0111] S5: Predict the demand of railway carriages through LSTM, and dynamically match the dynamic optimal matching of highway trucks and railway carriages based on a mixed integer programming model. The objective function is to minimize transportation distance, maximize time matching degree and loading rate to generate the optimal matching scheme of highway trucks and railway carriages. In order to solve the problem of low resource matching efficiency and serious waste of transport capacity, the specific steps are as follows:

[0112] Real-time cleaning and standardization processing of multi-source heterogeneous data is carried out by using distributed ETL tool, LSTM neural network is used to predict future carriage demand, and dynamic optimal matching is realized based on mixed integer programming model, which effectively solves the problem of resource waste caused by mismatch between highway trucks and railway carriages. At the same time, an intelligent matching algorithm based on RFID data is provided, which uses the information of goods weight and volume collected by goods RFID tags, combines the load and volume parameters of highway trucks, and based on the mixed integer programming model, takes the shortest transportation distance, the highest time matching degree and the maximum loading rate as the target to automatically match highway trucks and railway arrival goods. In the peak season, high-priority goods such as fresh food and emergency supplies are given priority in transport capacity, and in the off-season, idle trucks are guided to undertake backlog goods. The prediction formula is:

[0113]

[0114] wherein, is the final prediction value, W y is the weight matrix, h t is the hidden state, C t is the cell state, and σ is the sigmoid activation function, b y is the bias term. The model training uses Attention mechanism to enhance the weight of key features.

[0115] Obtain a railway carriage set, a highway truck set and a combined transport state information, the railway carriage set includes carriage ID, load, volume and available time, the highway truck set includes truck ID, load, volume and real-time position, and the combined transport state information includes railway freight arrival time and highway vehicle connection time;

[0116] Construct a mixed integer programming model, and a target function is to minimize total cost, and a formula is as follows:

[0117]

[0118] Wherein, C represents the truck set, C represents the carriage set, d ij represents the transportation distance of the truck i and the carriage j, t ij represents the time window matching degree, and alpha, beta and gamma represent weight coefficients, s ij represents the freight-carriage specification adaptation degree;

[0119] The freight-carriage specification adaptation degree s ij The total weight W of the freight collected through the freight RFID tag good , and the total volume V good The truck load W truck , and the volume V truck are calculated, and a formula is as follows

[0120] F fit When greater than or equal to 0.8, it is determined that the adaptation is good, a mixed integer programming model is solved, an optimal matching scheme is obtained, and the scheme is stored in a cloud database and synchronized to the highway truck through a vehicle terminal and a driver APP. The above scheme capacity demand prediction model and the intelligent matching algorithm based on the mixed integer programming effectively solve the supply-demand mismatch problem of the highway truck and the railway carriage, improve the utilization rate of the railway carriage in the transportation peak season, reduce the empty load rate of the highway truck in the off-season, and significantly reduce the enterprise operating cost.

[0121] At present, the scheme has low efficiency and disordered connection: the short-distance transfer from the railway freight station to the highway distribution center lacks accurate guidance, the electronic port platform or the dispatching system only displays the approximate area (such as "freight station A area") where the freight is located, and does not mark the specific parking point coordinates, and the driver needs to inquire and find on site after arriving on site, and the average time consumption is more than 15 minutes. At the same time, the internal of the freight station lacks planning of the vehicle driving route, which further aggravates the congestion at the loading and unloading link, resulting in low efficiency of the "last kilometer" transportation. In order to solve the above defects, S6: a three-dimensional visual navigation system of the freight station is constructed, a centimeter-level digital twin model is established based on point cloud scanning to generate centimeter-level accurate parking guidance, a partition dynamic traffic control strategy is combined, and the vehicle is guided to enter the specified parking position according to the preset route, and specifically includes:

[0122] Based on point cloud scanning equipment, the whole scene of the freight station is modeled, and a centimeter-level digital twin model containing coordinate system mapping relationship is generated.

[0123] At the entrance of the freight station and the relevant key nodes such as loading and unloading sites, UWB positioning base stations and RFID landmarks are deployed to form a spatial grid positioning reference.

[0124] The vehicle position is collected in real time through the vehicle-mounted UWB module, and the positioning error is corrected combined with the RFID landmark data to generate the vehicle driving trajectory. Combined with the optimal path generated by the intelligent scheduling model, the centimeter-level parking guidance (such as "go straight along A channel for 30 meters and turn left to No. 5 loading and unloading site") is displayed on the driver's APP.

[0125] The path is dynamically planned through the path planning cost function inside the freight station, and the dedicated channel is allocated according to the vehicle type, and the dynamic guidance instructions are displayed on the LED screen to implement the partitioned dynamic traffic control. According to the real-time vehicle flow (detected by camera and ground coil) and the cargo loading and unloading progress in the freight station, the intelligent gate and LED guidance screen are used to allocate dedicated channels for different types of vehicles (such as large trucks and small distribution vehicles) to avoid parking chaos.

[0126] The path planning cost function inside the freight station: collect the distance data and real-time congestion coefficient of all road segments in the freight station; get the number of turns of each road segment; calculate the cost C(n) of each path inside the freight station, the formula is: C(n) = d(n)·(1+k cong ·ρ(n))+k turn ·t(n),

[0127] Where d(n) is the path distance, ρ(n) is the road congestion coefficient (0-1), k ong is the congestion influence coefficient, t(n) is the number of turns, and k trun is the turn penalty coefficient. The smaller the C(n) value, the better the path. The minimum cost path is calculated by the path planning algorithm and sent to the driver's APP and vehicle-mounted terminal. Through the combination of high-precision positioning and safety monitoring network, the real-time monitoring algorithm of safety distance is realized to accurately track the position of personnel, vehicles and goods and effectively control the safety distance, improve the positioning accuracy, shorten the response time of the driver's off-duty and the cargo's out-of-safety-range warning, and greatly reduce the risk of cargo loss and operation.

[0128] The current scheme relies on GPS positioning technology, and the positioning error in complex environments such as freight stations and warehouses can reach 5-10 meters, which cannot meet the monitoring needs of close cooperation between personnel, vehicles and goods. It can only judge the route anomaly through trajectory deviation, but cannot monitor the physical distance between personnel and goods, and vehicles and goods in real time. When the driver is off-duty and the cargo is out of the safety range, the system cannot timely alarm, and there is a risk of cargo loss or operation. In order to solve the problems of limited positioning accuracy and delayed safety warning:

[0129] S7: Real-time monitoring of personnel, vehicle, and cargo position information through high-precision positioning and safety monitoring network, combined with safety distance real-time monitoring algorithm, triggering safety warning when detecting personnel off-duty, cargo out of safety range, or vehicle and cargo distance exceeding threshold;

[0130] Establish a high-precision positioning and safety monitoring network, using UWB ultra-wideband positioning technology inside the freight station and warehouse to obtain real-time three-dimensional coordinates of personnel, vehicles, and goods; use GPS + Beidou dual-mode positioning during transportation, combined with the movement trajectory of the goods RFID tag, to achieve full-process location tracking. Develop safety distance real-time monitoring and early warning algorithm, set safety distance threshold in different scenarios (such as personnel and cargo distance ≥ 2 meters during loading and unloading, goods and vehicle edge distance ≥ 0.5 meters during transportation), calculate real-time distance through sliding time window, when less than threshold, immediately issue sound and light warning through vehicle terminal and driver APP, and push to monitoring center. Real-time safety distance calculation and early warning determination:

[0131]

[0132] When D real is less than D threshold , trigger warning, where (x1, y1, z1) and (x2, y2, z2) are the real-time coordinates of the two monitored objects, and D threshold is the safety distance threshold.

[0133] In the current scheme, railways and highways belong to different management departments, and there are differences in cross-department and cross-regional policies (such as vehicle access rights, cargo security standards, RFID technology application specifications, etc.). The system technical solution needs to be repeatedly adapted to the management requirements of different regions, making it difficult to achieve standardized promotion and limited scale application. In order to solve the problems of standardization and policy implementation, and cross-regional adaptability:

[0134] S8: Build a cross-regional policy compliance evaluation model, extract policy core elements through NLP technology, and verify cross-regional transportation compliance through a dynamic policy matching engine, including:

[0135] Collect policy texts on vehicle access rights, cargo security standards, and RFID technology application specifications from provinces and cities across the country to generate a multi-regional policy adaptation database;

[0136] Extract policy core elements through NLP technology, including vehicle access rights, cargo security standards, and RFID technology application specifications, such as a province requiring dangerous goods transport vehicles to install specific RFID tags, and establish a mapping relationship between policies and technical requirements;

[0137] Call the development dynamic policy matching engine, when the vehicle team scheduling plan involves cross-regional transportation, automatically query the policy requirements of the passing area according to the transportation route, and combine the RFID information of the vehicle and the goods to check whether it meets the local regulations, such as whether the vehicle has access permission and whether the goods security meets the standard, and give adjustment suggestions (such as replacing the vehicle that meets the requirements) for the items that do not meet the requirements.

[0138] Cross-regional policy compliance P comply Evaluation calculation:

[0139]

[0140] Where, p m is the mth regional policy requirement, s m is the actual satisfaction state of the vehicle team, delta (·) is the compliance function, M is the total number of policies involved, P comply is greater than or equal to 0.9, it is determined to be passable. Through the multi-region policy adaptation database and the dynamic policy matching engine, the adaptation problem caused by the difference between different regional policies is solved, the adaptation cycle of the system to new regional policies is shortened, the policy compliance rate is improved, a solid foundation is laid for the standardization promotion and large-scale application of the technical scheme, and the development of intelligent and efficient intermodal transportation is promoted.

[0141] Embodiment two

[0142] Based on the same idea, the present application provides an intermodal transportation scheduling system based on digital twinning and multi-agent collaboration, comprising

[0143] A digital twinning model is used to receive the transportation order submitted by the consignor, establish a unique binding relationship between people, vehicles and goods in the transportation order based on an RFID binding algorithm, and synchronize it to the database, and build an initial digital twinning model including the transportation network, the cargo station facility and the vehicle equipment.

[0144] A data acquisition module is used to acquire real-time railway transportation state data, highway transportation state data and cargo station operation data, wherein the railway transportation state data includes railway cargo arrival time and railway compartment position, the highway transportation state data includes real-time highway vehicle position and loading state, and the cargo station operation data includes cargo station vehicle parking position and loading and unloading equipment state.

[0145] A multi-source data fusion center is used to standardize the railway transportation state data, highway transportation state data and cargo station operation data through the multi-source data fusion center, break through the information barrier between railways and highways, and generate real-time shared intermodal transportation state information.

[0146] The dynamic scheduling module is configured to analyze the intermodal transportation state information based on the LSTM collaborative scheduling model, and dynamically adjust the connection time, route and stopping position of the road vehicle when detecting railway delays, road congestion, abnormal cargo station operation and related emergencies, and output the road vehicle rerouting and waiting time estimation scheme under the emergency condition.

[0147] The resource matching module is configured to predict the railway carriage demand through the LSTM, dynamically match the dynamic optimal matching of the road freight vehicle and the railway carriage based on the mixed integer programming model, take the shortest transportation distance, the highest time matching degree and the maximum loading rate as the objective function, and generate the optimal matching scheme of the road freight vehicle and the railway carriage.

[0148] The precise navigation module is configured to construct a three-dimensional visual navigation system of the cargo station, generate a centimeter-level precise parking guide based on the centimeter-level digital twin model established by the point cloud scanning, and guide the vehicle to enter the designated stopping position according to the preset route in combination with the dynamic traffic control strategy.

[0149] The safety warning module is configured to monitor the position information of personnel, vehicles and goods in real time through high-precision positioning and safety monitoring network, and trigger a safety warning when detecting that the personnel are off duty, the goods are out of the safety range or the distance between the vehicle and the goods exceeds the threshold in combination with the real-time safety distance monitoring algorithm.

[0150] The policy adaptation module is configured to construct a cross-regional policy compliance evaluation model, extract the core elements of the policy through the NLP technology, and verify the cross-regional transportation compliance in combination with the dynamic policy matching engine.

[0151] The implementation principles of the above modules have been described in the foregoing embodiments, and thus will not be repeated here.

[0152] Embodiment three

[0153] Based on the same concept, in some embodiments of the present application, an electronic device is also provided. The electronic device includes a memory and a processor. The memory is configured to store a processing program, and the processor is configured to execute the processing program according to the instructions. When the processor executes the processing program, the digital twin and multi-agent collaborative intermodal transportation scheduling method in the foregoing embodiments is implemented.

[0154] In some embodiments of the present application, a readable storage medium is also provided. The readable storage medium can be a non-volatile readable storage medium or a volatile readable storage medium. The readable storage medium stores instructions. When the instructions are run on a computer, the electronic device containing the readable storage medium executes the digital twin and multi-agent collaborative intermodal transportation scheduling method in the foregoing embodiments.

[0155] It can be understood that, for the aforementioned one based on digital twin and multi-agent collaborative highway-railway intermodal transportation scheduling method, if all are realized in the form of software function modules and sold or used as independent products, it can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application essentially or the part that contributes to the prior art or the whole or part of the technical solutions can be embodied in the form of a software product, which is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method of each embodiment of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (Read-only memory, ROM), a random access memory (Random access memory, RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0156] The computer readable storage medium can include a data signal carried in the baseband or as part of a carrier wave propagating through the program code readable by the computer. Such a propagating data signal can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. The readable storage medium can also be any readable medium other than the readable storage medium, which can send, propagate or transmit programs for use by or in conjunction with instruction execution systems, devices or apparatuses. The program code contained on the readable storage medium can be transmitted by any suitable medium, including but not limited to wireless, wired, optical cable, RF, etc., or any suitable combination of the above.

[0157] The program code for executing the technical solutions disclosed in the present application can be written in any combination of one or more programming languages, including object-oriented programming languages such as Java, C++, etc., and conventional procedural programming languages such as C language or similar programming languages. The program code can be executed entirely on the user computing device, partially on the user device, as an independent software package, partially on the user computing device and partially on the remote computing device, or entirely on the remote computing device or server. In the case of remote computing device, the remote computing device can be connected to the user computing device through any kind of network, including local area network (LAN) or wide area network (WAN), or can be connected to external computing device (for example, connected through Internet by using Internet service provider).

[0158] It should be pointed out finally that the above embodiments are only used to illustrate the technical solutions of the present application, but not to limit the same; and although the present application has been described in detail with reference to the foregoing embodiments, it should be appreciated by those skilled in the art that the technical solutions recorded in the foregoing embodiments can be modified, or some technical features thereof can be replaced equivalently; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A road-rail intermodal transport scheduling method based on digital twins and multi-agent collaboration, characterized in that, include: The system receives transportation orders submitted by cargo owners, establishes a unique binding relationship between people, vehicles, and goods in the transportation order based on RFID binding algorithms, and synchronizes it to the database to build an initial digital twin model that includes the transportation network, freight station facilities, and vehicle equipment. Real-time collection of railway transportation status data, highway transportation status data, and freight station operation data. The railway transportation status data includes railway freight arrival time and railway carriage location. The highway transportation status data includes real-time location and cargo loading status of highway vehicles. The freight station operation data includes vehicle parking positions and loading / unloading equipment status within the freight station. The railway transportation status data, highway transportation status data, and freight station operation data are standardized through a multi-source data fusion center, breaking down information barriers between railways and highways and generating real-time shared intermodal transport status information. The intermodal transport status information is analyzed based on the LSTM collaborative scheduling model. When railway delays, highway congestion, freight station operation abnormalities and related emergencies are detected, the connection time, route and stop position of highway vehicles are dynamically adjusted, and the road detour and waiting time prediction scheme for highway vehicles under emergencies is output. The analysis of the intermodal transport status information based on the LSTM collaborative scheduling model further includes: acquiring standardized intermodal transport status information, extracting emergency situation features, including railway train delay duration, highway congestion length, and freight station operation anomaly type; inputting the emergency situation features into the LSTM collaborative scheduling model, and outputting the scheduling adjustment intensity A. t The calculation formula is: ,in, For railway train delay duration, L congest S represents the length of the congested road section. bind To ensure driver-vehicle stability, w rail w road w bind These are the weighting coefficients; The system predicts railway carriage demand using LSTM and dynamically matches the optimal combination of road freight cars and railway carriages based on a mixed-integer programming model. The objective function is to minimize transportation distance, maximize time matching, and maximize loading rate, generating the optimal matching scheme for road freight cars and railway carriages. The dynamic matching of road freight cars and railway carriages based on the mixed-integer programming model includes: acquiring a set of railway carriages, a set of road freight cars, and intermodal transport status information. The railway carriage set includes carriage ID, load capacity, volume, and available time; the road freight car set includes freight car ID, load capacity, volume, and real-time location; and the intermodal transport status information includes railway freight arrival time and road vehicle connection time. A mixed-integer programming model is constructed, with the objective function being to minimize the total cost, as shown in the formula: ,in, C represents a set of freight cars, and D represents a set of carriages. ij The distance t represents the transport distance between truck i and carriage j. ij Indicates time window matching degree, Indicates the weighting coefficient, s ij Indicates the cargo-carriage size fit; cargo-carriage size fit s ij The total weight W of the goods collected by the cargo RFID tag good Total volume V good With truck load capacity W truck Volume V truck The calculation formula is as follows: Solve the mixed integer programming model to obtain the optimal matching scheme, and store the scheme in the cloud database. Then, synchronize it to the truck on the highway through the vehicle terminal and the driver's APP.

2. The road-rail intermodal transport scheduling method based on digital twin and multi-agent collaboration as described in claim 1, characterized in that, Also includes: A 3D visualization navigation system for freight stations is constructed. Based on point cloud scanning, a centimeter-level digital twin model is established to generate centimeter-level precise parking guidance. Combined with a zoned dynamic traffic control strategy, vehicles are guided to enter designated parking spaces according to preset routes. The high-precision positioning and security monitoring network monitors the location information of personnel, vehicles and goods in real time. Combined with the real-time security distance monitoring algorithm, a security warning is triggered when personnel leave their posts, goods leave the safe range or the distance between vehicles and goods exceeds the threshold. A cross-regional policy compliance assessment model is constructed, which extracts the core elements of policies through NLP technology and combines them with a dynamic policy matching engine to verify the compliance of cross-regional transportation.

3. The road-rail intermodal transport scheduling method based on digital twin and multi-agent collaboration as described in claim 1, characterized in that, The step of establishing a unique binding relationship between people, vehicles, and goods in a transportation order based on an RFID binding algorithm and synchronizing it to the database further includes: Physical tags storing cargo identification information are configured on goods, identification devices for reading the physical tags are configured on highway vehicles, and drivers are equipped with identification badges carrying driver identification information. The station's automated verification system collects cargo information from the physical tags, vehicle information from the identification devices, and driver information from the identification badges, and verifies the consistency of the three based on a binding verification algorithm. The binding verification algorithm is as follows: ， Among them, RFID1, RFID2, and RFID3 are RFID tags for people, vehicles, and goods, respectively. k Identification information, db k For the pre-stored association information in the database, For the matching function, Mbind=1 indicates a valid binding; A cross-system data interaction platform is constructed based on the multimodal transport data exchange standard. The data format of railway freight plan data and road vehicle dynamic data is unified through a data conversion interface. Distributed data processing tools are used to clean the unified data in real time to ensure the consistency of key fields. The data conversion interface realizes the automatic correspondence between XML and JSON fields through predefined mapping rules. The key fields include cargo ID, arrival time, cargo location number and vehicle ID. Based on the cargo RFID status change event triggering information push mechanism, when the unloading status of cargo at the railway end is detected to be completed, the unloading time and cargo location information are automatically pushed to the vehicle-mounted terminal and driver terminal of the associated highway vehicle.

4. The road-rail intermodal transport scheduling method based on digital twin and multi-agent collaboration as described in claim 1, characterized in that, The standardization process of the railway transportation status data, highway transportation status data, and freight station operation data through the multi-source data fusion center includes: Real-time road and rail status data is collected, which includes data from the railway CTC system, highway GPS positioning, meteorological early warning information, and related multi-source real-time data streams. The Apache Flink distributed processing framework is used to access multi-source data streams in real time. The IsolationForest anomaly detection algorithm is used to identify anomalies in the state data. The anomaly score is calculated using the following formula: , Where h(x) is the path length of sample x in the isolated tree, E(h(x)) is the expected value of the path length in multiple isolated trees, c(n) is the path length standardization factor used to eliminate the influence of sample size on the results, and n is the number of samples. The missing value imputation algorithm of random forest is used to fill in the missing values ​​of the public railway status data after removing outliers. The imputation formula is as follows: , in, For feature column X to be completed j The predicted value, where N is the number of decision trees in the random forest, f i Let X be the prediction function of the i-th decision tree. -j To divide X j Other feature columns besides.

5. The road-rail intermodal transport scheduling method based on digital twin and multi-agent collaboration according to claim 2, characterized in that, The construction of the cargo station 3D visualization navigation system further includes: The freight station is modeled in its entirety using point cloud scanning equipment, generating a centimeter-level digital twin model that includes coordinate system mapping relationships. UWB positioning base stations and RFID landmarks are deployed at key nodes such as cargo station entrances and loading / unloading points to form a spatial grid-based positioning benchmark. The vehicle's location is collected in real time by the on-board UWB module, and the positioning error is corrected by combining RFID landmark data to generate the vehicle's driving trajectory. The system dynamically plans routes using the internal route planning cost function of the freight station, allocates dedicated lanes according to vehicle type, and displays dynamic guidance instructions on the LED screen.

6. The road-rail intermodal transport scheduling method based on digital twin and multi-agent collaboration as described in claim 5, characterized in that, The steps of the internal route planning cost function of the freight station are as follows: Collect distance data and real-time congestion coefficients for all road sections within the freight station; Get the number of turns for each road segment; The cost C(n) of each path within the freight station is calculated using the following formula: , Where d(n) is the path distance, ρ(n) is the road segment congestion coefficient (0-1), and k ong Let t(n) be the congestion impact coefficient, t(n) be the number of turns, and k be the number of turns. trun This is the turning penalty coefficient; The minimum cost path is determined by a path planning algorithm, and the optimal path is then distributed to the driver's app and the vehicle terminal.

7. The road-rail intermodal transport scheduling method based on digital twin and multi-agent collaboration according to claim 2, characterized in that, The construction of the cross-regional policy compliance assessment model includes: Collect policy documents from all provinces and cities across the country regarding vehicle access rights, cargo security inspection standards, and RFID technology application specifications, and generate a multi-regional policy adaptation database; The core elements of the policy are extracted using NLP technology. These core elements include vehicle access rights, cargo security inspection standards, and RFID technology application specifications, and a mapping relationship between the policy and technical requirements is established. The system calls upon a dynamic policy matching engine to automatically query the policy requirements of the regions along the transportation route when the fleet scheduling plan involves cross-regional transportation, and combines the RFID information of the vehicles and goods to verify whether they comply with local regulations. Cross-regional policy compliance P comply Evaluation calculation: ， Where, p m For the m-th regional policy requirement, s m This represents the actual condition of the fleet. For the conforming function, M is the total number of policies involved, P comply A value greater than or equal to 0.9 is considered passable.

8. A road-rail intermodal transport scheduling system based on digital twin and multi-agent collaboration, characterized in that, Implementing the road-rail intermodal transport scheduling method based on digital twins and multi-agent collaboration as described in any one of claims 1 to 7, comprising: The digital twin model is used to receive transportation orders submitted by cargo owners, establish a unique binding relationship between people, vehicles and goods in the transportation order based on RFID binding algorithm and synchronize it to the database, and build an initial digital twin model including transportation network, freight station facilities and vehicle equipment; The data acquisition module is used to collect real-time railway transportation status data, highway transportation status data, and freight station operation data. The railway transportation status data includes the arrival time of railway goods and the location of railway carriages. The highway transportation status data includes the real-time location of highway vehicles and their loading status. The freight station operation data includes the parking positions of vehicles within the freight station and the status of loading and unloading equipment. The multi-source data fusion center is used to standardize the railway transportation status data, highway transportation status data, and freight station operation data, break down information barriers between railways and highways, and generate real-time shared intermodal transport status information. The dynamic scheduling module is used to analyze the intermodal transport status information based on the LSTM collaborative scheduling model. When railway delays, highway congestion, freight station operation abnormalities and related emergencies are detected, the module dynamically adjusts the connection time, route and stop position of highway vehicles and outputs the highway vehicle detour and waiting time prediction scheme under emergencies. The resource matching module is used to predict railway car demand using LSTM and dynamically match the optimal combination of road freight cars and railway cars based on a mixed integer programming model. The objective function is to generate the optimal matching scheme between road freight cars and railway cars with the shortest transportation distance, the highest time matching degree, and the highest loading rate.

Images