Urban rail transit intelligent customer service center system based on AI and big data deep fusion

Abstract

The invention discloses an urban rail transit intelligent customer service center system based on AI and big data deep fusion, and relates to the technical field of intelligent customer service. Comprising the following steps: constructing an intelligent system covering the whole life cycle of metro operation by taking demand-decision-execution as a core; based on the construction of an intelligent system covering the whole life cycle of subway operation, an urban rail transit intelligent customer service center system is established through multi-source data fusion, multi-task collaborative learning, multi-modal interaction and digital twin simulation. The inquiry and travel requirements of most passengers are met, great convenience is provided for in-station activities of the passengers, the passengers are assisted in making decisions, related travel suggestions are provided, the future upgrading direction of the system can be automatically sorted out, the service mode is optimized, the quality is guaranteed, the number of personnel is reduced, the efficiency is improved, and the practical value is very high.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent customer service, and particularly relates to a city rail transit intelligent customer service center system based on AI and big data deep fusion. BACKGROUND

[0002] The subway traditional customer service mode has the following core pain points: first, the artificial service efficiency is insufficient, the surge of consultation volume during holidays and peak hours leads to response delay, which easily causes station congestion; second, the system data is fragmented, the information of ticketing, security, dispatching and other subsystems has not realized interconnection and intercommunication, which restricts the supply of accurate service; third, the self-service terminal function is solidified, the ticketing machine, inquiry machine and other equipment cannot dynamically adapt to the differentiated needs of passengers; fourth, the service channel coordination is poor, passengers need to repeatedly describe the problem at different service touch points, the service record is difficult to trace across channels, and the complaint closed-loop processing rate is low; fifth, the emergency response mechanism is inefficient, the sudden conditions (such as train delay, equipment failure) rely on manual dispatching, and the disposal efficiency needs to be improved. In addition, the traditional customer service mode highly depends on manual operation, and the repetitive and high-intensity work easily leads to operation fatigue, increases the risk of human error, and at the same time, the service scene is solidified, the tool is single, and the place is limited, which is difficult to meet the personalized and proactive service needs of passengers.

[0003] At present, the city rail transit intelligent customer service system has made a stage breakthrough in the field of virtual customer service and voice interaction, realizes the real scene interaction between passengers and virtual customer service based on intelligent voice and computer vision technology, but the language support ability is limited, and it is difficult to cover the multi-language and multi-dialect service scene. Although the intelligent robot as an unattended customer service supplement has basic functions such as ticketing processing and remote inquiry, it has defects such as single query information dimension, lack of data integration and analysis ability, and limited mobile service range, and cannot realize the whole process guiding service of passengers.

[0004] Some smart customer service solutions exist both domestically and internationally. For example, the Chengdu Zhiyuanhui Intelligent Customer Service Center system utilizes a rule-based intelligent question-and-answer system, integrated with BI tools to visualize passenger flow data and provide standardized answers through keyword matching. It also supports voice-to-text input and integrates GIS maps for visual monitoring of station equipment status. However, it only supports fixed sentence matching, resulting in low accuracy in complex semantic scenarios. Passenger flow warnings require manual confirmation to trigger dispatch instructions, resulting in response delays of 5-10 minutes. It is unable to provide personalized services based on passenger history, such as recommending optimal transfer options. Another example is the CloudX platform, an IoT platform based on edge computing and digital twin technologies, which enables real-time monitoring of equipment status. It uses AI video analysis to provide platform congestion warnings, supports remote station switching, and can construct 3D station models for virtual inspections of equipment failures. However, it only responds to existing events and cannot predict potential risks. It lacks correlation analysis of deeper data, such as passenger profiles and operational efficiency. True data sharing is lacking, and interfaces with ticketing, emergency command, and other systems are not integrated, creating new data silos.

[0005] In order to break through the existing technical bottlenecks, it is urgent to build a full-link intelligent service architecture. By breaking through system data barriers, strengthening multimodal interaction capabilities, and expanding smart terminal service scenarios, we can achieve full-cycle intelligent management of passengers from entry to exit, improve operational efficiency and service quality, and build a proactive, personalized, and integrated smart customer service system.

[0006] Currently, the functions of urban rail transit smart customer service centers are primarily limited to ticket management and information inquiries. They lack comprehensive transportation information announcements, cultural and creative products, and other value-added services tailored to passenger needs. Furthermore, their support for text and language support is limited, creating significant inconvenience for passengers from around the world. Intelligent robots are also limited in their range of operation, making them ineffective in guiding passengers, and they lack statistical analysis and problem collection capabilities.

[0007] Therefore, an urban rail transit intelligent customer service center system based on the deep integration of AI and big data is proposed, aiming to solve the shortcomings of existing customer service centers in service efficiency, intelligence level and data analysis capabilities. Summary of the Invention

[0008] In view of this, the present invention provides an urban rail transit intelligent customer service center system based on the deep integration of AI and big data, which meets the inquiry and travel needs of most passengers, provides great convenience for passengers' activities in the station, assists passengers in making decisions, provides relevant travel suggestions, and can automatically sort out the future upgrade direction of the system, optimize the service mode, ensure quality, reduce staff and increase efficiency, and has strong practical value.

[0009] In order to achieve the above object, the present application adopts the following technical solutions: An urban rail transit intelligent customer service center system based on AI and big data deep fusion comprises the following steps: Demand-decision-execution is taken as the core to build an intelligent system covering the whole life cycle of subway operation; Based on the intelligent system covering the whole life cycle of subway operation, the urban rail transit intelligent customer service center system is established through multi-source data fusion, multi-task collaborative learning, multi-modal interaction and digital twin simulation.

[0010] Optionally, the intelligent system covering the whole life cycle of subway operation specifically comprises an urban rail transit cloud platform, an urban rail transit service platform and a device platform. The urban rail transit cloud platform is used to provide real-time data and algorithm support for the urban rail transit service platform and the device platform. The urban rail transit service platform is used to collect demand data and feed back to the urban rail transit cloud platform to form a demand-decision-execution closed loop, assist passenger decision-making and provide personalized suggestions for passengers. The device platform is used for the execution and environmental perception of intelligent terminals, and converts the decision of the urban rail transit service platform into services.

[0011] Optionally, the urban rail transit cloud platform comprises a data perception layer, an intelligent decision layer, a service implementation layer and an effect evaluation layer. The data perception layer is used to fuse the subway AFC system, vehicle-mounted sensors and passenger mobile phone signaling to realize data acquisition. The intelligent decision layer is used to build a multi-task learning model based on the Transformer architecture to simultaneously process passenger flow prediction, device fault diagnosis and service resource scheduling tasks. The service implementation layer is used to deploy three execution modules of digital human customer service, AR navigation and intelligent scheduling instruction issuing to support automatic handling in complex scenarios. The effect evaluation layer is used to establish an evaluation system, dynamically optimize service strategies based on A / B testing, and form an execution-evaluation-iteration closed loop optimization mechanism.

[0012] Optionally, the urban rail transit service platform comprises ticket management, information inquiry, information announcement, cultural and creative services and other services. The ticket management includes ticket selling, recharging, ticket refunding and ticket replacing. The information inquiry includes ticket information and route information, the ticket information includes historical records, balance and validity period, and the route information includes subway lines, optimal routes, surrounding scenic spots and travel suggestions. Information bulletin includes weather information, train information, scenic spot introduction, emergency announcement, city introduction, medical travel; Text and creativity services include providing customized photos, special scene customization services, and 3D printing personalized services. Other services include rescue, fire fighting, AED, alarm emergency service.

[0013] Optionally, the device platform includes intelligent robots, multi-modal interaction recognition modules, dynamic resource scheduling modules, digital twin emergency deduction modules; The intelligent robot is used to guide passengers to the platform, the toilet or to find the station staff, and to make voice broadcast; The multi-modal interaction recognition module is used to integrate BERT and GPT pre-training models, combine with the track transportation field knowledge graph, realize context understanding, sentiment analysis and cross-domain knowledge reasoning, cover all-scene passenger demand through multi-language voice recognition and OCR technology, and individualize recommendation system user portrait and dialect recognition; The dynamic resource scheduling module is used to realize the transition from experience-driven to data-driven; The digital twin emergency deduction module is used to build a station-level digital twin, integrate mechanical and electrical equipment positions and evacuation channel information, realize scene dynamic simulation, simulate personnel evacuation path through Pathfinder software, and realize cross-city emergency plan sharing through federated learning technology.

[0014] Optionally, the dynamic resource scheduling module includes a transport capacity scheduling unit, a personnel scheduling unit and a big data platform; The transport capacity scheduling unit is based on a deep reinforcement learning train marshalling, stop time and gate opening number dynamic optimization model, takes transport capacity matching degree, energy consumption and punctuality rate as three target optimization functions, and realizes transport capacity utilization rate improvement; The personnel scheduling unit constructs a service personnel skill label system, and realizes optimal matching of work order-personnel through the Hungarian algorithm; The big data platform is used to establish a unified data lake, integrate passenger consultation, ticketing and equipment log data, predict equipment failure through an XGBoost model, and mine customer complaint hotspots through an LDA topic model.

[0015] According to the technical scheme, compared with the prior art, the present application provides a city rail transit intelligent customer service center system based on AI and big data deep fusion, which has the following advantages: (1) The present application realizes the transition of rail transit from "passive response" to "active prediction" and from "experience-driven" to "data-driven" through full-link intelligent transformation, and helps the subway operator to create a safe, efficient and humanized next-generation intelligent transportation system; (2) The application is based on the construction of a multi-task learning model based on the Transformer architecture, and realizes dynamic strategy optimization by combining reinforcement learning, realizes passenger flow prediction and equipment early warning technology, simulates personnel evacuation paths through Pathfinder software, and realizes cross-city emergency plan sharing technology through federated learning technology.

[0016] (3) The application meets the query and travel needs of most passengers, provides great convenience for passenger activities in the station, assists passengers in decision-making, provides relevant travel suggestions, can automatically arrange the future upgrading direction of the system, optimizes the service mode, and thus has strong practical value.

[0017] (4) The application integrates ticket management, information inquiry, information announcement, cultural and creative services and other services into the city rail transit service platform of the intelligent customer service center, uses cloud platform big data, AI technology, video, GIS, Internet of Things, converged communication and other technologies to assist passengers in decision-making, provides suggestions and reduces the travel burden of passengers. BRIEF DESCRIPTION OF DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiment or prior art description. Obviously, the drawings in the following description are only a part of the embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor on the basis of the provided drawings.

[0019] Figure 1 A system block diagram of the intelligent customer service center of urban rail transit based on AI and deep fusion of big data is provided for the present application. Figure 2 A schematic diagram of the intelligent customer service center of urban rail transit is provided for the present application. Figure 3 A schematic diagram of the device platform is provided for the present application. Figure 4 A schematic diagram of the intelligent robot is provided for the present application. DETAILED DESCRIPTION

[0020] The technical solutions in the embodiments of the present application will be described in detail below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0021] REFERENCE Figure 1As shown, the application discloses a kind of city rail transit wisdom customer service center system based on AI and big data deep fusion, including following steps: With demand-decision-execution as core, build intelligent system covering the whole life cycle of subway operation; Based on the intelligent system covering the whole life cycle of subway operation, through multi-source data fusion, multi-task collaborative learning, multi-modal interaction and digital twin simulation, a city rail transit wisdom customer service center system is established.

[0022] Further, the intelligent system covering the whole life cycle of subway operation specifically includes: city rail transit cloud platform, city rail transit service platform and equipment platform; The city rail transit cloud platform is used to provide real-time data and algorithm support for the city rail transit service platform and equipment platform; The city rail transit service platform is used to collect demand data and feedback to the city rail transit cloud platform, forming a demand-decision-execution closed loop, assisting passenger decision-making and providing personalized recommendations for passengers; The equipment platform is used for intelligent terminal execution and environment perception, and converts the decision of the city rail transit service platform into service.

[0023] Specifically, the city rail transit cloud platform, as a technical support and data hub, is used to provide real-time data and algorithm support for the service platform and equipment platform, support intelligent operation of the whole link, and ensure service response speed and decision accuracy; The city rail transit service platform, as the interaction interface between the city rail transit cloud platform and passengers, is used to collect demand data and feedback to the city rail transit cloud platform, forming a "demand-decision-execution" closed loop, assisting passenger decision-making and providing personalized recommendations; The equipment platform is responsible for intelligent terminal execution and environment perception, converts the decision of the city rail transit cloud platform into physical world service, and performs data back transmission to optimize the model.

[0024] Further, the city rail transit cloud platform includes data perception layer, intelligent decision layer, service implementation layer and effect evaluation layer; The data perception layer is used to fuse subway AFC system, vehicle-mounted sensor and passenger mobile phone signaling to realize data acquisition; The intelligent decision layer is used to build a multi-task learning model based on Transformer architecture, which simultaneously processes passenger flow prediction, equipment fault diagnosis and service resource scheduling tasks; The service implementation layer is used to deploy digital human customer service, AR navigation and intelligent scheduling instruction issuing three execution modules, supporting automatic handling of complex scenarios; The effect evaluation layer is used to establish an evaluation system, dynamically optimize service strategies based on A / B testing, and form a closed-loop optimization mechanism of execution-evaluation-iteration.

[0025] Specifically, such as Figure 2 As shown, the data perception layer integrates 12 data sources, including the subway's AFC system, onboard sensors, and passenger mobile phone signals, enabling millisecond-level data collection and standardized processing through edge computing nodes. Key technologies include multi-protocol data gateways, streaming computing frameworks (such as Kafka), and millisecond-level data synchronization mechanisms to ensure real-time data across the entire region.

[0026] Intelligent Decision-Making Layer: A multi-task learning model based on the Transformer architecture simultaneously handles three tasks: passenger flow forecasting, equipment fault diagnosis, and service resource scheduling. This model utilizes knowledge distillation technology to compress and deploy on the edge, combined with reinforcement learning for dynamic policy optimization. This model boasts inference speeds three times faster than traditional solutions, supporting 15-minute passenger flow heatmap predictions and 72-hour warnings for equipment failures.

[0027] Service implementation layer: Deploy digital human customer service (3D virtual image based on NeRF), AR navigation (combined with LiDAR point cloud to achieve centimeter-level positioning) and intelligent scheduling instruction issuance module. Through the linkage of the rule engine and the emergency plan library, 90% of routine problems can be automatically handled and complex scenarios can be responded to in seconds.

[0028] Effect evaluation layer: Establish an evaluation system that includes 20 indicators such as response speed, resolution rate, and passenger satisfaction, dynamically optimize service strategies based on A / B testing, and form a closed-loop optimization mechanism of "execution-evaluation-iteration".

[0029] Furthermore, the urban rail transit service platform includes ticket management, information inquiry, information announcement, cultural and creative services, and other services; among them, Ticket management includes ticket sales, recharge, refund and replacement; Information query includes ticket information and route information. Ticket information includes history, balance, and validity period. Route information includes subway lines, optimal routes, surrounding attractions, and travel suggestions. Information announcements include weather information, train information, scenic spot introductions, emergency announcements, city introductions, and medical travel; Cultural and creative services include customized photos, special scene customization services, and 3D printing personalized services; Other services include rescue, firefighting, AED, and police emergency services.

[0030] Furthermore, the equipment platform includes intelligent robots, multimodal interaction recognition modules, dynamic resource scheduling modules, and digital twin emergency simulation modules; Intelligent robots are used to guide passengers to platforms, toilets or station staff, and make voice broadcasts; A multi-modal interaction recognition module is used to integrate BERT and GPT pre-training models, combine a rail transit field knowledge graph, realize context understanding, sentiment analysis and cross-field knowledge reasoning, cover all-scenario passenger needs through multi-language voice recognition and OCR technology, and individualize a user portrait of a recommendation system to improve a service click rate by 40%. A dynamic resource scheduling module is used to realize a transition from experience-driven to data-driven. A digital twin emergency deduction module is used to construct a station-level digital twin, integrate electromechanical equipment positions and evacuation channel information, realize scene dynamic simulation, simulate personnel evacuation paths through Pathfinder software, and realize cross-city emergency plan sharing through federated learning technology.

[0031] Specifically, as shown in Figure 3 , Figure 4 The intelligent robots of the equipment platform guide passengers unfamiliar with the station to platforms, toilets or station staff. For passengers who need special care, such as the old, young, sick, disabled and pregnant, the intelligent robots will make timely voice broadcasts during the guidance process, such as "someone is passing by, please let one go" or "a guide dog is working, please do not touch, please let one go". The action speed of the intelligent robots can also be freely adjusted to meet the needs of different passengers. In addition, the intelligent robots are also equipped with voice guidance functions and can provide emotional relief services when passengers have negative emotions.

[0032] Specifically, the multi-modal interaction recognition module integrates pre-training models such as BERT and GPT, combines a rail transit field knowledge graph (including 2 million entities and 10 million relationships), realizes context understanding, sentiment analysis and cross-field knowledge reasoning. Through multi-language voice recognition (supporting 12 dialects) and OCR technology, all-scenario passenger needs are covered, and a user portrait of an individualized recommendation system (such as frequently visited stations and historical inquiries) is improved by 40% in service click rate. The dialect recognition accuracy is ≥92% (based on Wav2Vec 2.0 fine-tuning); the knowledge graph reasoning delay is <200 ms (Neo4j graph database); and the multi-modal fusion decision (vision + voice + text) supports complex demand analysis.

[0033] Specifically, the digital twin emergency simulation module builds a station-level digital twin (based on BIM technology, with an accuracy of ±2cm), integrating information on electromechanical equipment locations and evacuation routes. It supports dynamic simulation of 10 scenarios, including fires and high passenger flow. Evacuation routes are simulated using Pathfinder software, and federated learning technology enables cross-city emergency plan sharing (with Paillier homomorphic encryption ensuring data security). A single simulation takes less than 10 minutes (using a 40-core CPU with 512GB of RAM), and the emergency plan matching accuracy is ≥85% (based on a historical case library).

[0034] Furthermore, the dynamic resource scheduling module includes a capacity scheduling unit, a personnel scheduling unit, and a big data platform; The capacity dispatch unit uses a dynamic optimization model for train formation, stop time, and gate opening quantity based on deep reinforcement learning. It optimizes the three objectives of capacity matching, energy consumption, and punctuality to improve capacity utilization. The personnel scheduling unit builds a service personnel skill labeling system and uses the Hungarian algorithm to achieve the optimal matching between work orders and personnel; The big data platform is used to build a unified data lake, integrate passenger consultation, ticketing, and equipment log data, predict equipment failures through the XGBoost model, and mine customer complaint hotspots through the LDA topic model.

[0035] Specifically, the capacity dispatching unit uses a dynamic optimization model for train formation, stop time, and number of gate openings based on deep reinforcement learning (PPO algorithm), with capacity matching, energy consumption, and punctuality as the three-objective optimization function, to achieve a 25% increase in capacity utilization.

[0036] The personnel scheduling unit builds a service personnel skill labeling system and uses the Hungarian algorithm to achieve optimal matching between work orders and personnel, shortening the average response time from 15 minutes to 8 minutes.

[0037] The big data platform establishes a unified data lake (Hadoop cluster with a storage capacity of 1PB), integrates data such as passenger inquiries, ticketing, and equipment logs, uses the XGBoost model to predict equipment failures (providing 72 hours of advance warning), and uses the LDA topic model to identify hot spots in customer complaints (such as "recharge failure" accounting for 30%).

[0038] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0039] The foregoing description of the disclosed embodiments enables a person skilled in the art to make or use the application. Modifications of these embodiments will occur to persons of skill in the art, and that the appended claims are intended to cover all such modifications that do not depart from the true spirit and scope of the application. Therefore, the application is not limited to the embodiments shown but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An urban rail transit smart customer service center system based on the deep integration of AI and big data, characterized by: include: Focusing on demand-decision-execution, we will build an intelligent system covering the entire life cycle of subway operations. Based on building an intelligent system covering the entire life cycle of subway operations, an urban rail transit smart customer service center system is established through multi-source data fusion, multi-task collaborative learning, multi-modal interaction and digital twin simulation.

2. The urban rail transit intelligent customer service center system based on deep integration of AI and big data according to claim 1 is characterized in that: The intelligent system covering the entire life cycle of subway operations includes: urban rail transit cloud platform, urban rail transit service platform and equipment platform; Urban rail transit cloud platform, used to provide real-time data and algorithm support for urban rail transit service platforms and equipment platforms; The urban rail transit service platform is used to collect demand data and feed it back to the urban rail transit cloud platform, forming a demand-decision-execution closed loop, assisting passengers in decision-making, and providing personalized recommendations. The device platform is used for the execution and environmental perception of intelligent terminals, converting the decisions of the urban rail transit service platform into services.

3. The urban rail transit intelligent customer service center system based on deep integration of AI and big data according to claim 2 is characterized in that: The urban rail transit cloud platform includes data perception layer, intelligent decision-making layer, service promotion layer and effect evaluation layer; among them, The data perception layer is used to integrate the subway AFC system, on-board sensors, and passenger mobile phone signaling to achieve data collection; The intelligent decision-making layer is used to build a multi-task learning model based on the Transformer architecture, and simultaneously handle passenger flow prediction, equipment fault diagnosis, and service resource scheduling tasks; The service implementation layer is used to deploy three execution modules: digital human customer service, AR navigation, and intelligent dispatch instruction issuance, supporting automated processing of complex scenarios; The effect evaluation layer is used to establish an evaluation system, dynamically optimize service strategies based on A / B testing, and form a closed-loop optimization mechanism of execution-evaluation-iteration.

4. The urban rail transit intelligent customer service center system based on deep integration of AI and big data according to claim 2 is characterized in that: The urban rail transit service platform includes ticket management, information inquiry, information announcement, cultural and creative services, and other services; among them, Ticket management includes ticket sales, recharge, refund and replacement; Information query includes ticket information and route information. Ticket information includes history, balance, and validity period. Route information includes subway lines, optimal routes, surrounding attractions, and travel suggestions. Information announcements include weather information, train information, scenic spot introductions, emergency announcements, city introductions, and medical travel; Cultural and creative services include customized photos, special scene customization services, and 3D printing personalized services; Other services include rescue, firefighting, AED, and police emergency services.

5. The urban rail transit intelligent customer service center system based on deep integration of AI and big data according to claim 2 is characterized in that: The equipment platform includes intelligent robots, multimodal interaction recognition modules, dynamic resource scheduling modules, and digital twin emergency simulation modules; Intelligent robots are used to guide passengers to platforms, restrooms, or station staff, and provide voice announcements; The multimodal interactive recognition module integrates BERT and GPT pre-trained models with rail transit knowledge graphs to achieve contextual understanding, sentiment analysis, and cross-domain knowledge reasoning. It uses multilingual speech recognition and OCR technology to cover all passenger needs in all scenarios, including user profiling and dialect recognition for personalized recommendation systems. Dynamic resource scheduling module, used to achieve the transition from experience-driven to data-driven; The digital twin emergency simulation module is used to build station-level digital twins, integrate the location of electromechanical equipment and evacuation channel information, realize dynamic scene simulation, simulate personnel evacuation routes through Pathfinder software, and use federated learning technology to realize cross-city emergency plan sharing.

6. The urban rail transit intelligent customer service center system based on deep integration of AI and big data according to claim 5 is characterized in that: The dynamic resource scheduling module includes the transport scheduling unit, the personnel scheduling unit and the big data platform; The capacity dispatch unit uses a dynamic optimization model for train formation, stop time, and gate opening quantity based on deep reinforcement learning. It optimizes the three objectives of capacity matching, energy consumption, and punctuality to improve capacity utilization. The personnel scheduling unit builds a service personnel skill labeling system and uses the Hungarian algorithm to achieve the optimal matching between work orders and personnel; The big data platform is used to build a unified data lake, integrate passenger consultation, ticketing, and equipment log data, predict equipment failures through the XGBoost model, and mine customer complaint hotspots through the LDA topic model.

Images