A Business Semantic-Based Approach for Underground Cavern Network Modeling and Traffic Congestion Diagnosis
By constructing a business semantic description layer and state machine scheduling in the underground cavern network, the problem of insufficient understanding of vehicle behavior rules in existing technologies is solved, realizing automated management of construction vehicles and generation of diagnostic conclusions to guide construction decisions.
Patent Information
- Application Number
- CN202611251086.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-18
- Publication Date
- 2026-11-17
AI Technical Summary
Existing technologies cannot effectively understand vehicle behavior rules in underground cavern construction scenarios, resulting in simulation results that cannot guide construction decisions, and there is a lack of vehicle state machine cyclic scheduling methods applicable to construction procedure rules.
An independent business semantic description layer is constructed to assign business semantic types to nodes and road segments in the road network. Construction data is converted into simulated traffic flow through a unified data source adaptation protocol, and the arrival of vehicles triggers the execution state transition of road segments in real time, generating congestion diagnosis conclusions expressed in business language.
It has achieved automated management of the entire life cycle of construction vehicles, generated diagnostic conclusions that can guide construction decisions, lowered the threshold for using simulation technology, and improved the degree of automation.
Smart Images

Figure CN123116571A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of underground engineering construction simulation and traffic congestion analysis technology, and in particular to a method for modeling underground cavern network and diagnosing traffic congestion based on business semantics. Background Technology
[0002] Unlike urban roads or highway construction areas, underground cavern complexes have the following significant characteristics: the road network consists of a multi-level tunnel network composed of access tunnels, construction adits, and external roads, with a unique topology; road network nodes have unique business functions such as main plant buildings, slag yards, and parking lots, and their congestion characteristics are completely different from ordinary intersections; construction operations are carried out cyclically according to procedures, resulting in periodic and pulse-like traffic flow; data sources cover three granularities: detailed construction plans, static traffic flow in the OD matrix, and coarse-grained construction plans; demand analysis requires not only knowing whether there is congestion, but also knowing diagnostic conclusions that can guide construction decisions, such as "which work face is causing the congestion" and "which adits are the bottlenecks."
[0003] In recent years, technical solutions have emerged for real-time vehicle behavior control via simulation control interfaces during microscopic traffic simulation operations. However, these methods all employ global optimization paradigms (such as genetic algorithms and Bayesian optimization) to calculate the "optimal scheduling scheme." In the scenario of underground cavern construction, vehicle behavior patterns are determined by construction procedure rules—such as muck trucks must proceed to the slag yard after arriving at the working face and automatically recycle after completing a quota volume—belonging to rule-driven state machine scheduling rather than an optimization problem. Existing technologies do not provide a real-time simulation vehicle scheduling method based on semantically triggered road segments and state machine-driven approaches suitable for construction scenarios. General traffic simulation tools treat road network elements as indiscriminate nodes and road segment numbers, failing to understand the semantics of construction operations such as main plant buildings, slag yards, and access tunnels. This results in simulation results that cannot be directly applied by construction managers. Construction vehicle flow generation relies on manual compilation, especially the static "start-end" travel list for cyclical transportation tasks such as slag removal, which cannot express rule-driven termination logic such as "automatic recycling after volume targets are met." Existing vehicle control methods in simulations use optimization paradigms and lack a state machine cyclic scheduling method for slag removal vehicles based on semantically triggered road segments that is applicable to construction procedure rules. Summary of the Invention
[0004] Purpose of the invention: The purpose of this invention is to provide a method for modeling underground cavern network and diagnosing traffic congestion based on business semantics, thereby solving the problems existing in the background technology.
[0005] Technical Solution: The present invention provides a method for modeling underground cavern network and diagnosing traffic congestion based on business semantics, comprising the following steps: Step 1: Build an independent business semantic description layer on top of the simulated road network topology data, and assign business semantic types to the nodes and road segments in the road network respectively; the business semantic type system includes multiple node types and multiple road segment types, and the semantic data is stored in independent files and decoupled from the road network topology data; Step 2: Through a unified data source adaptation protocol, three different granularity construction data sources, namely detailed construction plan, OD matrix static traffic flow and coarse-grained construction plan, are converted into simulated traffic flow. Among them, detailed construction plan and OD matrix static traffic flow generate static vehicle travel list, and coarse-grained construction plan generates construction vehicle dispatcher. Step 3: Map the business semantic nodes to trigger road segments. During the simulation, detect in real time whether the vehicle has arrived at the trigger road segment. Perform the corresponding state transition according to the type of the arrived trigger road segment, and control the vehicle to circulate between the working face and the slag yard until the task quota is completed. Step 4: Integrate the static vehicle travel list with the construction vehicle dispatcher into the same simulation scenario, start the micro traffic simulation engine to execute the simulation, and output simulation data; Step 5: Calculate congestion indicators based on simulation output data, distribute vehicle delays to each node on the path using a node-level delay allocation algorithm, and attribute congestion indicators to specific business nodes and business road segments by combining the business semantic type system, generating congestion diagnosis conclusions expressed in business language.
[0006] Furthermore, in step 1, the node types include main plant, slag yard, parking lot and intersection, and the road segment types include access tunnel, construction adit and external road; the semantic data is generated by manual annotation, stored independently in JSON format, and is physically separated from the road network topology file but logically related.
[0007] Furthermore, in step 2, the unified data source adaptation protocol includes two methods: generating static traffic flow and generating a construction vehicle scheduler. The detailed construction plan adapter parses precise travel records one by one to generate static traffic flow. The OD matrix adapter uses business semantics to automatically filter start and end nodes and generate batch static traffic flow evenly. The coarse-grained construction plan adapter generates a scheduler for dynamically scheduling construction vehicles. Static traffic flow and construction vehicle scheduler run in a mixed manner in the same simulation scenario.
[0008] Further, step 3 is as follows: Map the working face node to the first triggering road segment and the slag yard node to the second triggering road segment; dynamically create construction vehicles at the departure point and set the initial destination to the first triggering road segment; in each simulation step, detect the road segment where the active vehicle is located, update the route to the second triggering road segment when the vehicle arrives at the first triggering road segment and perform loading and parking, update the route to return to the first triggering road segment when the vehicle arrives at the second triggering road segment and perform unloading and parking; alternately execute the above process to form a cyclical transportation.
[0009] Furthermore, step 3 also includes volume quota termination: the volume quota is calculated based on the total volume and single vehicle load in the coarse-grained construction plan. After each one-way transport from the working face to the slag yard is completed, the cumulative transport time is incremented by one. When the cumulative transport time reaches the volume quota, the vehicle is routed to the departure point for recycling after arriving at the trigger section for the next time, and the cycle is terminated.
[0010] Furthermore, in step 5, the node-level delay allocation algorithm obtains the complete driving path of each vehicle and evenly distributes the total delay time of the vehicles to each node on the path; after summarizing the delay allocation results of all vehicles, it combines business semantic tags to perform attribution analysis according to the business node dimension to locate the construction work surface corresponding to the source of congestion.
[0011] Furthermore, in step 5, the calculated congestion indicators include density, speed, queue length, delay time, saturation, and timeout rate; the start and end times of congestion are identified through time series analysis, and the congested road segment IDs are replaced with semantic tags in combination with the business semantic type system, automatically generating readable diagnostic conclusions containing congestion time periods, congestion levels, and congestion source information.
[0012] The present invention discloses a system for modeling underground cavern network and diagnosing traffic congestion based on business semantics, comprising: The semantic road network modeling module is used to build an independent business semantic description layer on top of the simulated road network topology data, and to assign business semantic types to nodes and road segments; The vehicle type configuration module is used to configure the physical parameters of various construction vehicles; The traffic flow generation module is used to convert three data sources—detailed construction plans, OD matrix static traffic flow, and coarse-grained construction plans—into simulated traffic flow through a unified data source adaptation protocol. The construction vehicle scheduling module is used to map business semantic nodes to trigger road segments, and to detect the status of vehicles arriving at the trigger road segments in real time during the simulation operation and execute the corresponding state machine scheduling. The simulation execution module is used to integrate static traffic flow and construction vehicle dispatcher into the same simulation scenario, drive the micro traffic simulation engine to execute simulation and output data; The congestion diagnosis module is used to calculate congestion indicators and generate congestion diagnosis conclusions expressed in business language by combining a node-level delay amortization algorithm with a business semantic type system.
[0013] The present invention provides a storage medium storing a computer program, wherein the computer program is configured to execute the above-described method at runtime.
[0014] The apparatus of the present invention includes a memory and a processor, wherein the memory stores a computer program and the processor is configured to run the computer program to perform the above-described method.
[0015] Beneficial effects: Compared with the prior art, the present invention has the following significant advantages: (1) Strong domain-specificity: A road network business semantic model was designed specifically for underground cavern groups. The simulation input and diagnostic output both use business language familiar to construction managers, which lowers the threshold for using simulation technology.
[0016] (2) Realistically reflects the characteristics of cyclical operations: The construction semantic nodes are mapped to trigger road sections, and the route update and parking settings are automatically executed when the vehicle arrives. Combined with the volume quota termination condition, the automated management of the entire life cycle of construction vehicles is realized, which truly reflects the characteristics of cyclical operations.
[0017] (3) High degree of automation: The entire chain from road network import, multi-source traffic flow fusion, dynamic simulation execution to semantic congestion diagnosis is automated, which greatly reduces the workload of manual compilation and analysis.
[0018] (4) The diagnostic conclusions are interpretable: Through the node-level delay sharing algorithm and semantic type mapping, the simulation indicators are reversed to the specific work surface and road segment type, generating diagnostic conclusions that can guide construction decisions. Attached Figure Description
[0019] Figure 1 This is a flowchart of the present invention; Figure 2 This is a partial example of the business semantic road network visualization of the present invention, showing the annotation effects of 4 node types and 3 road segment types.
[0020] Figure 3 This is a diagram of the multi-source traffic flow unified fusion protocol architecture of the present invention, showing the layered architecture of three data sources adapted to static traffic flow and construction vehicle dispatcher through the RouteSource protocol.
[0021] Figure 4 The flowchart of the state machine cyclic scheduling algorithm for slag removal and transportation vehicles of the present invention shows the complete process of semantic node mapping → vehicle deployment → road segment arrival detection → state transition → volume quota termination.
[0022] Figure 5 This is a partial schematic diagram of the semantic congestion diagnosis report interface of the present invention, showing the interface layout of the web-based congestion report page, including functional areas such as indicator cards, saturation heatmap, and bottleneck node ranking. Detailed Implementation
[0023] The technical solution of the present invention will be further described below with reference to the accompanying drawings.
[0024] like Figure 1 As shown, this embodiment of the invention provides a method for modeling underground cavern network and diagnosing traffic congestion based on business semantics, including the following steps: This embodiment uses the construction of an underground powerhouse cavern complex at a hydroelectric power station as an application scenario. The underground powerhouse system comprises three main caverns: the main powerhouse, the main transformer room, and the tailrace surge chamber, as well as a multi-level tunnel network consisting of access tunnels, multiple construction adits, and external roads. During construction, multiple work faces simultaneously perform cyclical operations such as drilling, blasting, muck removal, and support, generating a large number of construction vehicles (dump trucks, concrete mixer trucks, etc.) traversing within the tunnel network. This embodiment uses SUMO (Simulation of Urban Mobility) as the microscopic traffic simulation engine, achieving real-time vehicle scheduling during simulation operation through its provided TraCI (Traffic Control Interface) control interface. However, it should be understood that the method of this invention is also applicable to other microscopic traffic simulation platforms that provide runtime vehicle control interfaces.
[0025] Step S1: Construct a business semantic road network model for underground cavern groups The tunnel network of underground cavern complexes differs from urban roads—it is not a simple intersection or street, but a multi-level network composed of access tunnels (main tunnels), construction adits (branch tunnels), and external roads (connecting roads outside the tunnel entrances); the nodes in the network are not ordinary crossroads, but locations with specific engineering functions: such as the main plant (the area where drilling, blasting, and slag removal operations are underway), the slag yard (the waste disposal site), and the parking lot (the vehicle parking area), etc.
[0026] If a road network file is directly imported using a general simulation tool, the simulation engine can only see a bunch of numbered nodes and road segments, and cannot distinguish whether "this node is a working face or a regular intersection" or "this road segment is an access tunnel or a construction adit". As a result, the simulation results can only be presented in the form of node numbers and road segment numbers, which construction managers cannot directly understand and apply.
[0027] To address this issue, this step involves constructing an independent business semantic description layer on top of the simulated road network topology data, specifically including: S101: Road Network File Import and Format Conversion. Receives road network source files (supports OpenDRIVE .xodr files or SUMO .net.xml files). If it is in .xodr format, it calls the simulation engine's built-in netconvert conversion tool to convert it into a .net.xml format road network topology file, which serves as the basis for all subsequent operations.
[0028] S102: Road Network Parsing and Verification. Parses the road network topology file, extracts road segment and node information, and constructs the road network data structure. After parsing, it automatically performs multiple health checks (such as checks for isolated nodes, dead-end roads, and zero-length road segments) to ensure the quality of the road network data.
[0029] S103: Business Semantic Data Acquisition. This involves associating each node and road segment in the road network with its corresponding business semantic type. This embodiment involves four node semantic types: main plant (the work area where construction is underway), slag yard (a waste disposal site), parking lot (a vehicle parking and waiting area), and intersection (a common tunnel intersection); and three road segment semantic types: access tunnel (the main passage tunnel connecting the tunnel entrance to the construction area), construction adit (branch tunnels branching from the access tunnel to various work faces), and external road (the ground road outside the tunnel entrance). The semantic type system can be flexibly expanded as needed to adapt to construction scenarios of different scales and types.
[0030] Semantic information annotation is completed manually (annotation methods include pre-annotation, batch annotation, etc., and are manually reviewed and confirmed): Users directly select nodes or road segments on the road network map in the web interface and choose the corresponding business semantic type (main plant, slag yard, etc.). Semantic data is stored in a separate JSON file, physically separate from the road network topology file but logically related, without affecting the simulation engine's normal parsing of the road network file, and also facilitating independent maintenance and version management of semantic data. (See attached file.) Figure 2 As shown, the semantically annotated road network uses different colors and labels to distinguish between node types and road segment types.
[0031] Step S2: Vehicle type library configuration Construction transport vehicles are diverse, with significant differences in vehicle length, maximum speed, acceleration, and deceleration performance. These physical parameters directly affect their driving behavior and traffic flow characteristics in tunnels. This embodiment uses dump trucks (slag removal trucks) as the primary transport vehicle in the muck removal scenario. These trucks are relatively long and have a large load capacity, serving as the standard configuration for muck removal tasks in the simulation. The system also includes default configurations for various common construction vehicles such as concrete mixer trucks, personnel transport vehicles, flatbed trucks, and water trucks. Each vehicle type is equipped with complete physical parameters (maximum acceleration, maximum deceleration, vehicle length, maximum speed, etc.). Users can add or remove vehicle types or modify parameters according to the actual construction vehicle situation. Step S3: Static traffic flow generation In practical engineering, construction vehicle operation plans typically exist at two granularities: the most precise is the detailed construction plan (which specifies the departure time, origin, destination, and vehicle type for each vehicle), and the medium-granularity is the OD matrix (which presents the distribution of vehicle numbers between different areas in matrix form, such as "50 vehicle trips from the parking lot to work face 3"). This step unifies these two data sources into a static vehicle trip list that the simulation engine can consume. For coarse-grained construction plans and other traffic flows that require dynamic scheduling, step S4 handles them separately.
[0032] S301: Detailed Construction Schedule Adapter. When the data source is a detailed construction schedule accurate to a single trip (JSON format, each record containing fields such as vehicle ID, departure time, origin node, destination node, and vehicle type), the adapter directly parses each record and converts it into a static list of vehicle trips that the simulation engine can consume. This method is suitable for the construction phase where detailed scheduling has already been completed.
[0033] S302: Static Traffic Flow Adapter for OD Matrix. When the data source is an OD matrix, the adapter reads the configuration (e.g., "Parking lot → Main plant: 100 vehicles / hour"), automatically filters start and end nodes using the semantic tags established in S103 (e.g., automatically finding all nodes of type "Parking lot" as start points and all nodes of type "Main plant" as end points), and then evenly generates batch vehicle trips between each OD pair. This method is suitable for representing static traffic flow within a construction area (auxiliary vehicles, personnel commuting, etc., where detailed construction plans are not required).
[0034] Figure 3 A diagram illustrating the unified fusion protocol architecture for multi-source traffic flow shows the layered architecture of three data sources adapted to static traffic flow and construction vehicle dispatchers via the RouteSource protocol. (See attached diagram.) Figure 3 As shown.
[0035] Step S4: Generation of the construction vehicle state machine cyclic scheduler S401: Initialization Phase: Receives coarse-grained construction plan parameters, including the working face (source of slag discharge), slag yard (destination of slag discharge), departure point (initial vehicle deployment location), total slag discharge volume (e.g., 10,000 cubic meters), single vehicle rated load (e.g., 20 cubic meters / vehicle), cycle start time, etc. Automatically calculates the volume quota = rounded up (total volume / single vehicle load) = rounded up (10,000 / 20) = 500 trips. Simultaneously, using the semantic road network established in S1, automatically finds the entrance road segment corresponding to the working face node as the first trigger road segment, and the entrance road segment corresponding to the slag yard node as the second trigger road segment.
[0036] S402: Vehicle Deployment: At the start of the cycle, a specified number of construction vehicles (e.g., 5 vehicles working simultaneously) are dynamically created at the departure point via the simulation control interface. The initial destination for each vehicle is set as the first trigger road segment (the entrance road segment of the work face). These vehicles are as if they are "dispatched" to the work face to begin the first transport.
[0037] S403: State Machine Loop Scheduling: The simulation engine progresses in discrete time steps (e.g., one step per second). In each step, the scheduler detects the current road segment of all active vehicles. When a vehicle enters the first trigger segment (working face entrance), the "loading" state transition is automatically executed: The second trigger segment (slag yard entrance) is set as the vehicle's new route target through the simulation control interface. This triggers the simulation engine's built-in routing mechanism to automatically recalculate the travel path and update the vehicle's route based on the real-time traffic conditions of the current road network. Simultaneously, the vehicle is commanded to stop at the end of the lane to simulate loading waiting time. When a vehicle enters the second trigger segment (slag yard entrance), the "unloading" state transition is automatically executed: The first trigger segment (working face entrance) is reset as the route target through the simulation control interface. The built-in routing mechanism automatically recalculates the return path and updates the route. Simultaneously, the vehicle is commanded to stop to simulate unloading waiting time. The vehicle then automatically begins its next round trip. The state transition process for this step is shown in the attached diagram. Figure 4 As shown.
[0038] During the simulation, the construction vehicles follow a four-state cyclical model: empty vehicle drives to the working face → loading (stops and waits after arriving at the working face entrance) → loaded vehicle drives to the spoil heap → unloading (stops and waits after arriving at the spoil heap entrance) → empty vehicle drives to the working face, and so on. The triggering conditions and behaviors for each state are as follows: (1) Empty vehicle drives to the working face: The vehicle departs from the departure point or slag yard and the destination is the first trigger section (entrance of the working face). It drives normally and does not need to stop.
[0039] (2) Loading: Triggered when the vehicle enters the first triggering section. The scheduler recalculates the path to the second triggering section (slag yard entrance) through the TraCI interface, updates the vehicle route, and sets stops to simulate loading time. After the stops are completed, the vehicle automatically switches to a heavy load state.
[0040] (3) Heavy load driving to the slag yard: The vehicle is fully loaded and drives to the second trigger section. It is driving normally, but the speed is affected by the increased load.
[0041] (4) Unloading: Triggered when the vehicle enters the second triggering section. The dispatcher recalculates the path back to the first triggering section, updates the route, sets stops to simulate unloading time, increments the trip number by 1 after the stop ends, and automatically returns to the empty vehicle state to start the next trip.
[0042] State transition conditions summary: Entering the working face entrance → Loading; Loading completed → Heavy load; Entering the slag yard entrance → Unloading; Unloading completed → Empty vehicle. Anti-repeated triggering mechanism (S405) ensures that each passage through the triggering section triggers only one state transition. Volume quota termination (S404) after the cumulative number of trips reaches the target, the vehicle is recycled, and the cycle ends.
[0043] S404: Volume Quota Termination: Each time a "working face → slag yard" journey is completed, the vehicle's cumulative transport trips increase by 1. When the cumulative transport trips reach the volume quota (500 trips), the vehicle will no longer enter the loop when it arrives at the trigger section next time. Instead, it will be automatically routed to the departure point for recycling, marking the completion of the vehicle's transport task.
[0044] S405: Prevent repeated triggering: Maintain a set of processed road segment markers - when a vehicle enters a triggered road segment, it is marked as processed, and the marker is cleared after leaving, ensuring that each triggered road segment only performs a state transition once during each visit, avoiding repeated triggering in continuous simulation steps.
[0045] S406: Multi-task collaboration: When multiple working faces are simultaneously carrying out muck removal operations, the system creates an independent scheduler instance for each muck removal task, and manages the lifecycle of each scheduler uniformly through the scheduler manager. All schedulers' construction vehicles and static traffic flow run together in the same simulation scenario. The two mechanisms work together and complement each other to ensure the correctness of scheduling in the scenario of parallel muck removal from multiple working faces.
[0046] Step S5: Multi-source traffic flow fusion and simulation execution The static traffic flow generated in step S3 and the construction vehicle dispatcher generated in step S4 are integrated into a single simulation scenario. A micro-traffic simulation engine is then activated to simulate the micro-behaviors of all vehicles in the tunnel network, including driving, following, lane changing, and queuing. The specific sub-steps are as follows: S501: Static route calculation. For the static traffic flow (static traffic flow and detailed construction plan traffic flow) generated by S301 and S302, the routing calculation module of the simulation engine is called to calculate the specific driving route from the origin to the destination for each travel record (which road segments are passed through and which intersections are turned), and generate a routing file with complete route information.
[0047] S502: Simulation Configuration Generation. Integrates all input files to generate a complete simulation configuration file, including: road network topology file (from S1), routing file (static routing from S501), and running parameters such as simulation start and end time, time step (e.g., 1 second), and output options.
[0048] S503: Simulation Run. Start the SUMO simulation engine and load the simulation configuration file. At the start of the simulation, the construction vehicle scheduler created in S4 automatically deploys construction vehicles at each departure point and assigns initial driving routes. The simulation progresses in fixed time steps. At each step, each scheduler checks whether a vehicle has arrived at a trigger segment, triggering the corresponding state transition (loading → transporting → unloading → returning), and automatically updates the vehicle's driving route. Vehicles that have completed their transport tasks automatically leave the road network, and the corresponding scheduler terminates. Construction vehicles and the static traffic flow generated in S3 run together in the same simulation scenario. The SUMO engine uniformly handles all vehicle micro-driving behaviors such as following, lane changing, and queuing.
[0049] S504: Simulation Data Acquisition. After the simulation runs to the preset end time, it automatically collects and outputs four types of structured data files: vehicle trip statistics (the complete driving trajectory and time of each vehicle), floating car trajectory (recording the position and speed of each vehicle by time step), road segment aggregation data (average density, average speed, queue length, etc. of each road segment in each statistical time window), and global statistics (total number of vehicles, total driving distance, total delay time, etc. of the entire road network).
[0050] Step S6: Semantic congestion indicator aggregation and bottleneck diagnosis The raw data output from the simulation is technical—the road segment number is "edge_1234" and the node number is "node_5678"—and construction managers cannot directly derive meaningful conclusions from it. The core task of this step is to "translate" the technical data into business language.
[0051] S601: Parse simulation output data. Read vehicle travel statistics, floating car trajectories, road segment aggregation data, and global statistics from the four types of data files output by S504.
[0052] S602: Calculation of road segment-level indicators. Using road segments as the basic unit and statistical time windows as the granularity, six business congestion indicators are calculated for each road segment and time window: density, speed, queue length, delay time, saturation, and overtime rate.
[0053] Among them, the calculation of road segment-level indicators involves calculating the following six business congestion indicators for each road segment and each time window, using the road segment as the basic unit and the statistical time window as the granularity: (1) Density: (vehicles / km) Where N e The sum of vehicle counts at all sampling times on road segment e within the time window, where T is the simulation step count, and L is the number of simulation steps. e The length of the road segment is in kilometers. This formula is derived from the basic traffic flow density formula k = N / L by taking the mean value over a time window.
[0054] (2) Average speed: (km / h) Where Σv i n is the sum of the instantaneous speeds (m / s) of all vehicles on road segment e within the time window. e 3 represents the number of speed samples, and 3.6 is the conversion factor from m / s to km / h.
[0055] (3) Queue length: (rice) Where q t Max represents the number of vehicles with a speed lower than 1.0 m / s (defined as queuing) on road segment e in the t-th simulation step within the time window (one value per step, forming a time series). t Take the maximum value in each step, L veh This represents the length occupied by a single vehicle (8.0 meters, including vehicle length and safety clearance). The speed threshold of 1.0 m / s is a conservative value within this range, based on the common practice in mainstream microscopic traffic simulation tools such as VISSIM and SUMO, which typically use 1.0–2.0 m / s for queuing determination. This ensures the accuracy of queuing assessment. (4) Average delay time (seconds): Take the timeLoss field of the simulation output, which is the time lost by the vehicle due to traveling at a speed lower than the ideal speed.
[0056] (5) Saturation: Where ρ e D represents the road segment density. max For reference, the saturation density (a system-configurable parameter, defaulting to 50 veh / km·lane; saturation density is the critical density at which traffic flow transitions from free flow to congestion, and according to the basic traffic flow diagram theory, this value is determined by the headway. According to the U.S. Highway Capacity Manual (HCM), density is used as the core indicator for level of service classification. The density threshold for congestion (Level F) is approximately 28 pc / km / lane after conversion. Considering the large size of construction vehicles and the high safety distance requirements, 50 veh / km / lane is a reasonable upper limit).
[0057] (6) Timeout rate: Where N timeout N represents the number of vehicles of a specified model whose timeLoss exceeds a threshold (the threshold is set separately for each model). total This represents the total number of times this vehicle model has been driven.
[0058] S603: Node-level Delay Allocation. The total delay of a vehicle from its origin to its destination is caused by multiple nodes and road segments along its travel path. To allocate the responsibility for delays to specific construction workfaces or tunnel sections, this system employs a node-level delay allocation algorithm: obtaining the complete travel path of each vehicle (the sequence of nodes it passes through), and evenly distributing the vehicle's total delay time to each node along the path. For example, if a vehicle travels from the parking lot through the access tunnel to workface 3 with a total delay of 6 minutes, passing through 3 key nodes, then each node contributes 2 minutes of delay. After summarizing the delay allocation results for all vehicles, and combining them with the semantic road network tags in S1, the system can answer the question "Which workface's slag removal operation causes the most severe congestion?" from the perspective of business nodes.
[0059] S604: Bottleneck Identification and Diagnosis Report. Combining the calculation results of various indicators in S602, time series analysis is used to identify the start and end times of congestion, pinpointing continuous congestion periods and peak saturation sections. Utilizing the semantic type data from S1, congested section IDs are replaced with semantic labels (e.g., "Entrance Traffic Tunnel D-1") to locate the key tunnel section causing congestion and automatically generate readable diagnostic conclusions. For example, "Entrance Traffic Tunnel D-1 experienced severe congestion between 09:00 and 10:30, with a peak saturation of 1.8, a maximum queue length of 350 meters, and an average delay of 4.2 minutes. The congestion source is traced to slag removal vehicles from Main Plant No. 3." Various congestion indicators and diagnostic conclusions are presented in visual charts through a web interface, as shown in the attached diagram. Figure 5 As shown.
Claims
1. A method for modeling road network and diagnosing traffic congestion in a group of underground caverns based on business semantics, characterized in that, Includes the following steps: Step 1: Build an independent business semantic description layer on top of the simulated road network topology data, and assign business semantic types to the nodes and road segments in the road network respectively; the business semantic type system includes multiple node types and multiple road segment types, and the semantic data is stored in independent files and decoupled from the road network topology data; Step 2: Through a unified data source adaptation protocol, three different granularity construction data sources, namely detailed construction plan, OD matrix static traffic flow and coarse-grained construction plan, are converted into simulated traffic flow. Among them, detailed construction plan and OD matrix static traffic flow generate static vehicle travel list, and coarse-grained construction plan generates construction vehicle dispatcher. Step 3: Map the business semantic nodes to trigger road segments. During the simulation, detect in real time whether the vehicle has arrived at the trigger road segment. Perform the corresponding state transition according to the type of the arrived trigger road segment, and control the vehicle to circulate between the working face and the slag yard until the task quota is completed. Step 4: Integrate the static vehicle travel list with the construction vehicle dispatcher into the same simulation scenario, start the micro traffic simulation engine to execute the simulation, and output simulation data; Step 5: Calculate congestion indicators based on simulation output data, distribute vehicle delays to each node on the path using a node-level delay allocation algorithm, and attribute congestion indicators to specific business nodes and business road segments by combining the business semantic type system, generating congestion diagnosis conclusions expressed in business language.
2. The method according to claim 1, wherein, In step 1, the node types include main plant, slag yard, parking lot and intersection, and the road segment types include access tunnel, construction adit and external road. The semantic data is generated by manual annotation and stored independently in JSON format. It is physically separated from the road network topology file but logically related.
3. The method according to claim 1, wherein, In step 2, the unified data source adaptation protocol includes two methods: generating static traffic flow and generating a construction vehicle dispatcher; the detailed construction plan adapter parses each precise travel record to generate static traffic flow. The OD matrix adapter automatically filters start and end nodes using business semantics and generates batch static traffic flows evenly; the coarse-grained construction plan adapter generates a scheduler for dynamically scheduling construction vehicles; static traffic flows and construction vehicle schedulers run in a mixed manner in the same simulation scenario.
4. The method according to claim 1, wherein, Step 3 is as follows: Map the working face node to the first triggering road segment and the slag yard node to the second triggering road segment; dynamically create construction vehicles at the departure point and set the initial destination to the first triggering road segment; in each simulation step, detect the road segment where the active vehicle is located, update the route to the second triggering road segment when the vehicle arrives at the first triggering road segment and perform loading and parking, update the route to return to the first triggering road segment when the vehicle arrives at the second triggering road segment and perform unloading and parking; alternately execute the above process to form a cyclical transportation.
5. The method according to claim 4, wherein, Step 3 also includes volume quota termination: the volume quota is calculated based on the total volume and single vehicle load in the coarse-grained construction plan. After each one-way transport from the working face to the slag yard is completed, the cumulative transport time is increased by one. When the cumulative transport time reaches the volume quota, the vehicle is routed to the departure point for recycling after arriving at the trigger section for the next time, and the cycle is terminated.
6. The method of claim 1, wherein, In step 5, the node-level delay allocation algorithm obtains the complete driving path of each vehicle and evenly distributes the total delay time of the vehicles to each node on the path. After summarizing the delay allocation results of all vehicles, the algorithm combines business semantic tags to perform attribution analysis according to the business node dimension to locate the construction work surface corresponding to the source of congestion.
7. The method according to claim 1, wherein, In step 5, the calculated congestion indicators include density, speed, queue length, delay time, saturation, and timeout rate. The start and end times of congestion are identified through time series analysis, and the congested road segment IDs are replaced with semantic tags in combination with the business semantic type system. Readable diagnostic conclusions containing congestion time period, congestion degree, and congestion source information are automatically generated.
8. A system for modeling road network of underground cavern group and diagnosing traffic congestion based on service semantics, characterized in that, include: The semantic road network modeling module is used to build an independent business semantic description layer on top of the simulated road network topology data, and to assign business semantic types to nodes and road segments; The vehicle type configuration module is used to configure the physical parameters of various construction vehicles; The traffic flow generation module is used to convert three data sources—detailed construction plans, OD matrix static traffic flow, and coarse-grained construction plans—into simulated traffic flow through a unified data source adaptation protocol. The construction vehicle scheduling module is used to map business semantic nodes to trigger road segments, and to detect the status of vehicles arriving at the trigger road segments in real time during the simulation operation and execute the corresponding state machine scheduling. The simulation execution module is used to integrate static traffic flow and construction vehicle dispatcher into the same simulation scenario, drive the micro traffic simulation engine to execute simulation and output data; The congestion diagnosis module is used to calculate congestion indicators and generate congestion diagnosis conclusions expressed in business language by combining a node-level delay amortization algorithm with a business semantic type system.
9. A storage medium having stored therein a computer program, wherein, The computer program is configured to execute the method as described in any one of claims 1-7 at runtime.
10. An apparatus comprising a memory and a processor, wherein, The memory stores a computer program, and the processor is configured to run the computer program to perform the method as described in any one of claims 1-7.
