Quantum-enhanced real-time multi-modal transportation network optimization method and system

Abstract

The application provides a quantum-enhanced real-time multi-mode traffic network optimization method and system, and belongs to the technical field of intelligent traffic control and optimization; the method comprises the following steps: performing edge computing and preprocessing on multi-source motion data in a traffic network to generate traffic state feature data; using a quantum computing model to generate a signal scheme for different traffic control targets; constructing a digital twin model to verify the signal scheme; converting the verified signal scheme into a traffic signal control instruction and delivering the traffic signal control instruction to a traffic signal machine; continuously monitoring traffic conditions and feeding back the monitoring results to the quantum computing model to adjust the signal scheme in real time; and the application can efficiently coordinate multi-mode traffic flow in a large-scale urban network in real time, has rapid response and closed-loop adaptive capacity in emergency events, and can meet the actual needs of future smart cities and vehicle-road collaborative development.

Description

Technical Field

[0001] This invention belongs to the field of intelligent traffic control and optimization technology, and in particular relates to a quantum-enhanced real-time multimodal traffic network optimization method and system. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] With the continuous advancement of urbanization and the diversification of residents' travel needs, urban road networks are characterized by a high density and mixed traffic of various traffic entities, including motor vehicles, bicycles, pedestrians, and public transportation vehicles, posing unprecedented challenges to traffic signal optimization.

[0004] However, existing traffic signal optimization technologies have the following significant drawbacks:

[0005] (1) Existing signal control systems mainly adopt fixed time-period timing, adaptive timing based on local real-time data, or regional coordinated timing. These methods can often only be optimized at a single intersection or within a limited area, making it difficult to take into account the multi-modal traffic demands from the perspective of the overall network. Especially when traffic flow fluctuates drastically or when sudden events (such as road accidents, severe weather, or the later stages of major events) occur, classic adaptive control strategies are prone to getting stuck in local optima or unsolvable states, making it difficult to respond to network-wide congestion in a timely manner.

[0006] (2) Existing network-level optimization methods for large-scale transportation networks mostly rely on metaheuristic algorithms such as genetic algorithms, particle swarm optimization, and simulated annealing. These methods have huge computational loads and slow iterative convergence speed when dealing with combinatorial optimizations of tens of thousands of dimensions, and cannot meet the strict requirements of real-time timing.

[0007] (3) In recent years, quantum computing technology, especially quantum approximation optimization algorithm (QAOA), has shown great potential in solving quadratic unconstrained binary optimization (QUBO) problems. However, its application in real traffic scenarios faces several difficult technical problems, such as: massive data is difficult to accurately map to the model, the feasibility and high quality of the solution are difficult to guarantee under hardware limitations, the feasibility of the scheme is difficult to verify, and the scheme is difficult to adjust in a timely manner after implementation. Summary of the Invention

[0008] To overcome the shortcomings of the prior art, this invention provides a quantum-enhanced real-time multimodal traffic network optimization method and system, which can coordinate multimodal traffic flows in real time and efficiently in large-scale urban networks, and has both rapid response to emergencies and closed-loop adaptive capabilities, which can meet the actual needs of future smart city and vehicle-road cooperative development.

[0009] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions:

[0010] The first aspect of this invention provides a quantum-enhanced real-time multimodal traffic network optimization method.

[0011] A quantum-enhanced real-time multimodal transportation network optimization method includes:

[0012] Real-time acquisition of multi-source motion data within the transportation network;

[0013] Edge computing and preprocessing are performed on the obtained multi-source motion data to generate traffic state feature data;

[0014] Based on the traffic state characteristic data, a quantum computing model is used to generate signal schemes for different traffic control objectives;

[0015] Construct a digital twin model and verify the signal scheme;

[0016] The verified signal scheme is converted into traffic signal control commands and sent to the traffic signal controllers for execution;

[0017] The system continuously monitors traffic conditions after traffic signal control commands are issued and feeds the monitoring results back to the quantum computing model to adjust the signal scheme in real time.

[0018] Furthermore, various sensors used to collect multi-source motion data in the traffic network in real time achieve the acquisition of multi-source motion data through vehicle-to-everything (V2X) communication; at the same time, all sensors adopt a precise clock synchronization protocol to ensure consistent data timestamps.

[0019] Furthermore, the edge computing and preprocessing includes: each edge node fusing multi-source motion data within the same spatial region and performing sliding window filtering and feature smoothing operations; constructing feature vectors and assembling the feature vectors of all edge nodes into a global traffic state matrix by row; and generating traffic state feature data through feature extraction and sorting.

[0020] Furthermore, signal schemes are generated for different traffic control objectives, including: first, setting an initial weight vector for the multi-objective optimization problem and calculating linear and interaction coefficients; then, using a quantum approximation optimization algorithm combined with a classical backoff mechanism to solve the combinatorial optimization problem; and mapping the solution back to specific timing to generate a complete set of timing parameters as a signal scheme.

[0021] Furthermore, the signal scheme is verified, including: importing the required verification data into the digital twin model and calculating the simulation deviation; comparing the obtained simulation deviation with the set key performance indicators, and considering the verification as unsuccessful if any indicator fails to meet the standard.

[0022] Furthermore, traffic signal control commands are encrypted via an edge gateway before being sent to the traffic signal controller.

[0023] Furthermore, traffic conditions are continuously monitored after traffic signal control commands are issued. When an emergency occurs, the relevant edge node detects an accident marker and randomly applies linear penalty reinforcement to the current edge node.

[0024] A second aspect of the present invention provides a quantum-enhanced real-time multimodal traffic network optimization system.

[0025] A quantum-enhanced real-time multimodal transportation network optimization system includes:

[0026] The data acquisition and fusion module is configured to: acquire multi-source motion data within the traffic network in real time;

[0027] The edge computing and preprocessing module is configured to perform edge computing and preprocessing on the acquired multi-source motion data to generate traffic state feature data.

[0028] The quantum optimization module is configured to generate signal schemes for different traffic control objectives based on the traffic state feature data and using a quantum computing model.

[0029] The digital twin simulation module is configured to: construct a digital twin model and verify the signal scheme;

[0030] The classic execution control module is configured to convert the verified signal scheme into traffic signal control commands and send them to the traffic signal controller for execution.

[0031] The adaptive feedback control module is configured to continuously monitor traffic conditions after traffic signal control commands are issued and feed the monitoring results back to the quantum computing model to adjust the signal scheme in real time.

[0032] A third aspect of the present invention provides a computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the steps of a quantum-enhanced real-time multimodal traffic network optimization method as described in the first aspect of the present invention.

[0033] A fourth aspect of the present invention provides an electronic device including a memory, a processor, and a program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of a quantum-enhanced real-time multimodal traffic network optimization method as described in the first aspect of the present invention.

[0034] The above one or more technical solutions have the following beneficial effects:

[0035] (1) This invention generates a low-dimensional latent feature matrix by fusing multi-source data at edge nodes, and constructs a QUBO model covering all network nodes to simultaneously optimize multi-modal traffic flow such as motor vehicles, buses, non-motor vehicles, and pedestrians at the network level. When traffic flow fluctuates drastically or sudden events occur, it can quickly break local optima by strengthening penalties for key nodes and performing parallel incremental optimization in different regions, and respond promptly to the congestion situation of the entire network to achieve global collaborative optimization of multi-modal traffic.

[0036] (2) This invention constructs a compact QUBO model by performing dimensionality reduction and key feature extraction on massive traffic data, and adopts a hybrid solution strategy combining the quantum approximation optimization algorithm (QAOA) with the classical backoff mechanism. The combination of the efficiency of the quantum algorithm in handling high-dimensional combinatorial optimization problems and the fallback guarantee of the classical algorithm significantly reduces the amount of computation and accelerates the iteration convergence speed. It can efficiently handle traffic signal timing combinatorial optimization problems with tens of thousands of dimensions and meet the strict requirements of real-time timing for large-scale urban traffic networks.

[0037] (3) This invention accurately maps massive multi-source data into a compact QUBO model, solving the data mapping problem; by using a hybrid quantum-classical algorithm and a classical backoff mechanism, it ensures the feasibility and high quality of the solution under limited quantum hardware conditions; by using a digital twin simulation module to perform simulation verification and security assessment of the optimization scheme, it solves the problem of scheme feasibility verification; by using an adaptive feedback control module to monitor the execution effect in real time and dynamically adjust the weights and algorithm parameters, it realizes timely adjustment after the scheme is implemented, effectively overcoming the technical obstacles of quantum computing in traffic scenarios.

[0038] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0039] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0040] Figure 1 This is a flowchart of a quantum-enhanced real-time multimodal traffic network optimization method according to Embodiment 1 of the present invention. Detailed Implementation

[0041] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0042] It should be noted that the terminology used herein is for the purpose of describing particular implementations only and is not intended to limit the exemplary implementations of the present invention.

[0043] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0044] The overall approach of this invention is as follows: Based on multi-source sensing and vehicle-to-everything (V2X) technology, and combining the ideas of quantum computing and classical computing for collaborative problem-solving, this invention designs a real-time, efficient quantum-enhanced real-time multi-modal traffic network optimization method for multi-modal traffic flow (including motor vehicles, public transportation, non-motorized vehicles, and pedestrians) in large-scale urban road networks. This method preprocesses sensor and vehicle data streams through edge nodes, then constructs a network-level quadratic unconstrained binary optimization (QUBO) model based on abstract low-dimensional latent features. It utilizes a hybrid solution of the quantum approximation optimization algorithm (QAOA) and classical Lagrange relaxation or heuristic algorithms to achieve real-time optimization of city-level traffic signal timing. Its core lies in mapping the combinatorial optimization problem of traffic signal timing to a quantum-processable framework, and forming a closed-loop adaptive traffic flow management scheme through digital twin simulation and adaptive feedback control mechanisms. This achieves accurate optimization of multi-modal traffic signals in large-scale urban traffic networks and is widely applicable to the field of urban intelligent traffic management.

[0045] To facilitate understanding of the technical solution of this invention, the following terms are explained:

[0046] 1) QUBO (Quadratic Unconstrained Binary Optimization): A quantum computing optimization model used to map traffic optimization problems to a quantum computing framework.

[0047] 2) QAOA (Quantum Approximation Optimization Algorithm): An algorithm for efficiently solving the QUBO model on existing quantum computing devices.

[0048] 3) TCN (Temporal Convolutional Network): A deep learning algorithm suitable for processing time series data.

[0049] 4) Digital twin: A real-time virtual replica of a real transportation system, used for simulation and evaluation.

[0050] 5) V2X communication: Real-time communication technology between vehicles and road infrastructure (V2I) and networks (V2N).

[0051] Example 1

[0052] This embodiment discloses a quantum-enhanced real-time multimodal transportation network optimization method.

[0053] like Figure 1 As shown, a quantum-enhanced real-time multimodal transportation network optimization method includes:

[0054] Step S1: Real-time acquisition of multi-source motion data within the traffic network;

[0055] Step S2: Perform edge computing and preprocessing on the obtained multi-source motion data to generate traffic state feature data;

[0056] Step S3: Based on the traffic state characteristic data, a quantum computing model is used to generate signal schemes for different traffic control objectives;

[0057] Step S4: Construct a digital twin model and verify the signal scheme;

[0058] Step S5: Convert the verified signal scheme into traffic signal control instructions and send them to the traffic signal controller for execution;

[0059] Step S6: Continuously monitor the traffic situation after the traffic signal control command is issued, and feed the monitoring results back to the quantum computing model to adjust the signal scheme in real time.

[0060] Based on the above process, this invention can coordinate multi-modal traffic flows in real time and efficiently in large-scale urban networks, and also possesses rapid response to emergencies and closed-loop adaptive capabilities, which can meet the practical needs of future smart city and vehicle-road cooperative development. To facilitate understanding of the technical solution of this invention, the specific implementation methods of this invention will be further explained and described below.

[0061] In step S1, multi-source motion data within the traffic network are collected in real time.

[0062] Multiple sensors, including lidar, millimeter-wave radar, and high-definition video cameras, are deployed at key intersections and road sections in the city to collect real-time location information, movement trajectories, speed, and density data of motor vehicles, non-motor vehicles, pedestrians, and public transportation vehicles.

[0063] Multiple sensors used to collect multi-source motion data in the traffic network in real time achieve the collection of multi-source motion data through vehicle-to-everything (V2X) communication technology. That is, vehicle-to-infrastructure (V2I) and vehicle-to-network (V2N) communication is established through V2X technology to achieve real-time collection of vehicle status information.

[0064] Time synchronization and data preprocessing are triggered, and all acquisition devices (sensors) uniformly adopt precise clock synchronization (PTP protocol) to ensure consistent data timestamps. After receiving the sensor and V2X raw data, the edge node immediately triggers the preprocessing process, including: 1) Data integrity verification: If data packet loss or abnormal changes are detected, interpolation compensation or anomaly marking is performed; 2) Privacy protection and encryption: If the vehicle positioning information involves personal privacy, anonymization and encrypted transmission are adopted; 3) Data priority allocation: Based on indicators such as road congestion risk and accident probability, the data transmission priority is dynamically adjusted to ensure that critical sensor data is delivered first when the network is congested.

[0065] In step S2, edge computing and preprocessing are performed on the obtained multi-source motion data. Specifically, the original data (multi-source motion data) undergoes noise reduction, filtering, and feature extraction. High-quality traffic state feature data is generated using techniques such as temporal convolutional networks to improve data processing efficiency and reduce subsequent processing delays. This can be achieved through the following methods:

[0066] Step S2-1: Edge nodes fuse multi-sensor data within the same spatial region.

[0067] Extended Kalman Filter (EKF) is used to fuse the detection results of LiDAR and millimeter-wave radar to eliminate false alarms. A pre-trained convolutional neural network (CNN) (such as YOLOv5) is used to detect targets in the image / video stream, and then the cross-frame trajectory of vehicles, pedestrians and bicycles is tracked through a multi-target tracking (MOT) algorithm. Weighted Kalman Filter (WKF) is used on the data from geomagnetic coils and pedestrian counting cameras to generate accurate pedestrian / non-motorized vehicle stationary and movement status.

[0068] The fused information forms a "preliminary list of spatial-temporal objects," laying the foundation for subsequent feature extraction.

[0069] Step S2-2: Sliding window filtering and feature smoothing.

[0070] For vehicle speed and acceleration data collected by V2X, a sliding window median filter is used to remove outliers caused by short-term network jitter; for the statistical data of the edge nodes themselves (such as traffic flow per second, average speed, queue length, etc.), an exponentially weighted moving average (EWMA) is applied to ensure that the features change smoothly over time and with minimal noise.

[0071] Step S2-3: Multi-modal traffic main feature extraction. After edge preprocessing, construct the first... Node number (intersection or road segment) at time The eigenvectors, i.e.:

[0072] ;

[0073] in, Indicates the first Node number at time Traffic density (number of vehicles passing through per unit time). Indicates the first Node number at time The average speed; Indicates the first Node number at time Queue length; Indicates the first Node number at time The number of pedestrians waiting; Indicates the first Node number at time Bus lane occupancy rate; Indicates a safety symbol, if the first Node number at time If an accident or emergency is detected, then ,otherwise ; Indicates the first The remaining duration of the current phase of the traffic light at node number 1; Indicates the first Traffic growth rate of node number within a short future window (e.g., 30 seconds) (output by a pre-trained Temporal Convolutional Network (TCN), in units of relative change rate).

[0074] Step S2-4: Concatenate the feature vectors of all nodes row by row into a global traffic state matrix, i.e.:

[0075] ;

[0076] in, Indicates the total number of nodes to be controlled; Indicates time The feature matrix of all nodes in the network, with each row corresponding to the 8-dimensional features of a node; This indicates that the data is based on eight key indicators (such as traffic flow, speed, and queues). The first in the time-sharing transportation network A real-time status data set consisting of nodes.

[0077] Step S2-5: Perform temporal convolution and attention dimensionality reduction. The central node utilizes a temporal convolutional network (TCN, denoted as...) The data within the historical time window is used to extract temporal features, and then multi-head self-attention (hereinafter referred to as multi-head self-attention) is applied. Extract key node and key moment features in the space-time dimension to generate a low-dimensional latent feature matrix, i.e.:

[0078] ;

[0079] ;

[0080] in, Indicates the timing window length; This indicates that the TCN process will retain the data. One time-series feature channel; Indicates the feature dimension of the TCN output; This represents a temporal convolutional network used to capture nodes in the past. Dynamic changes at any given moment; This represents a multi-head self-attention mechanism, used to... Mapped to the latent feature space; Indicates time The generated low-dimensional latent feature matrix; This represents the potential dimensions of attention output.

[0081] Steps S2-6: Based on the attention weights or principal component analysis (PCA) results, [the following steps are performed]. Sort the "importance score" of each row and keep the top-scoring ones. Let there be nodes, denoted as set. ,in This indicates the retention ratio. For example, This indicates that 70% of the high-impact nodes will be retained. This represents the set of key node indices retained in this round.

[0082] For each reserved node Several candidate timing schemes are predefined, denoted as . and the first The node number is The possible solutions are mapped to binary decision variables, namely:

[0083] , ;

[0084] in, Indicates the first The total number of possible configuration schemes for node number; Indicating the first [unit / item] in this round of optimization Node number 1 was selected. Matching schedule.

[0085] In step S3, advanced quantum computing algorithms are used to solve the combinatorial optimization problem of large-scale transportation networks in real time; for different traffic control objectives (such as reducing congestion, reducing travel delays and reducing emissions), real-time signal optimization schemes are generated to effectively improve the operating efficiency of the entire transportation network.

[0086] In recent years, quantum computing technology, especially quantum approximation optimization algorithms (QAOA), has shown significant potential in solving quadratic unconstrained binary optimization (QUBO) problems. However, its application in real-world traffic scenarios still faces several challenges: First, how to generate compact QUBO models that can be solved using quantum computing after dimensionality reduction and feature extraction from massive amounts of multi-source sensor data and vehicle-to-everything (V2X) information; second, how to ensure the feasibility and high quality of solutions under limited quantum hardware scale and noise conditions through hybrid quantum-classical algorithms; third, how to interface the signal timing scheme derived from quantum optimization with a city-level digital twin platform for simulation verification and security assessment; and fourth, how to quickly adjust algorithm parameters and optimization target weights through adaptive feedback mechanisms when simulation results deviate from actual performance, thereby achieving closed-loop adaptive optimization of the system. Current technologies have not yet formed a comprehensive solution that fully integrates quantum computing with classical traffic simulation, edge computing, and adaptive control.

[0087] This invention overcomes the limitations of existing technologies by employing advanced quantum computing algorithms to solve the combinatorial optimization problem of large-scale transportation networks in real time. Specifically, it can be achieved through the following methods:

[0088] Step S3-1: After all candidate decision variables and their corresponding low-dimensional features are ready, set the initial weight vector for the multi-objective optimization problem. This is used to measure the importance of different objectives. During the morning rush hour, if you want to prioritize reducing average delays, you can set:

[0089] ;

[0090] in, This represents the weight of the objective of "reducing average delay". This indicates the weight of the goal of "reducing emissions". This indicates the weight of the goal of "improving bus punctuality." This indicates the weight of the goal of "ensuring pedestrian safety".

[0091] Step S3-2: Based on low-dimensional latent features And the response curves obtained from training with historical data, for each variable Calculate the linearity and interaction coefficients. Specifically, first predict the interaction coefficients using the response curve. Node number adopts the first The performance change of the proposed scheme compared to the baseline scheme is as follows:

[0092] ;

[0093] in, Indicates the first Intersection No. 1 adopts the first The average delay reduction after implementing this plan; This indicates the corresponding reduction in emissions; This indicates the predicted on-time bus rate under this scheme; This represents the pedestrian safety index predicted under this scheme. Represents binary variables The linear coefficients can be positive or negative, so that the objective function can simultaneously reduce delays and emissions, improve bus punctuality and pedestrian safety.

[0094] For any two binary decision variables and Calculate its second-order interaction coefficients If the solutions corresponding to these two variables are at the same intersection... There is a phase conflict, at which point:

[0095] ;

[0096] Its meaning is: when There is a need to impose strong penalties on conflicts. For a large positive value; if ,intersection and Distance is And since there is a timing synergy effect between the two, let:

[0097] ;

[0098] in, A constant representing the degree of decay in timing coordination benefits; Indicates intersection and The Euclidean distance between them; It is a decay parameter that determines the rate at which the benefits of timing coordination decay as the distance increases. It is based on low-dimensional latent features The calculated coordination strength coefficient is obtained when the traffic correlation between nodes is stronger. The value is greater.

[0099] If there is a conflict between bus priority and normal phase in the schemes corresponding to these two variables, the interaction coefficient is dynamically adjusted based on the bus arrival time (ETA). However, the overall form is still to represent the coordination or conflict between different schemes using a matrix.

[0100] All of these and After the merger, it constitutes a The QUBO matrix, where, Let represent the total number of binary variables representing all possible solutions for all nodes currently participating in the optimization. At this point, the objective function for Quadratic Unconstrained Binary Optimization (QUBO) is written as:

[0101] ;

[0102] in, This represents the vectorized form of all binary decision variables; the first term is a linear term, reflecting the contribution or penalty of each individual variable to the objective; the second term is a quadratic interaction term, reflecting the synergistic or conflicting effects between different variables.

[0103] Step S3-3: After the QUBO model is constructed, the Quantum Approximation Optimization Algorithm (QAOA) combined with the classical backoff mechanism is used to solve this large-scale high-dimensional combinatorial optimization problem. First, the number of iterations in QAOA is set to... And on the quantum module (which can be real quantum hardware or a high-precision quantum simulator), parameterized quantum circuits are constructed, namely:

[0104] ;

[0105] ;

[0106] in, Indicates all The bits are initially in a uniform superposition state; It is the cost Hamiltonian, which, after diagonalization, is related to the QUBO objective function. equivalence; It is a mixed Hamiltonian. Indicates the action in the Pauli- on 1 qubit X matrix; and It is the parameter vector to be optimized.

[0107] In the first layer of QAOA, cost evolution is applied first. Then apply hybrid evolution The second layer applies first. Reapplication And so on until the 1st Layer. After the entire quantum circuit has been executed, the expected energy value can be obtained through multiple samplings:

[0108] ;

[0109] This expected value is then fed back to the classic optimizer to update the parameter vector. and until the convergence condition is met. Or reach the maximum number of iterations .in, Represents the expected energy value with respect to the parameter The gradient; It is a pre-set convergence threshold.

[0110] However, considering the current scale and noise limitations of quantum hardware, when QAOA fails to find a feasible solution within the specified time window, the system will initiate a classical backoff mechanism. At this point, the QUBO problem is handed over to a classical CPU / GPU cluster for solving using Lagrange relaxation algorithms or heuristic algorithms such as simulated annealing or genetic algorithms, yielding a suboptimal solution. In subsequent iterations of QAOA, a suboptimal solution is used as an initial guess to accelerate convergence to a higher-quality solution. After several rounds of alternating quantum and classical iterations, the solution optimal in terms of traffic network performance metrics (such as average delay) is finally selected, denoted as... ,and:

[0111] .

[0112] Step S3-4: When the optimal or near-optimal binary decision vector is obtained... Then, it is mapped back to the specific timing. For all Make This determined the first Intersection No. 1 should adopt the first The green light duration of traffic lights and pedestrian / bus priority strategies are used. This yields a complete set of timing parameters, namely:

[0113] ;

[0114] For the first Intersection No. 1: Indicates the duration of the first green light phase, in seconds; Indicates the duration of the yellow light transition, in seconds; Indicates the duration of the red light, in seconds.

[0115] This timing scheme will be packaged into two parts: one part will be sent to the digital twin module for simulation verification, and the other part will be converted into actual signal control commands later.

[0116] In step S4, a high-fidelity real-time digital twin model of urban traffic is established to simulate and verify the traffic signal control scheme (i.e., the signal scheme) in real time. This allows for rapid evaluation of the optimization effect before the actual implementation of the scheme, ensuring its effectiveness and safety. Specifically, this can be achieved through the following methods:

[0117] Step S4-1: In the digital twin simulation, the topology of the entire urban road network (road length, number of lanes, traffic signal locations, etc.), the historical OD (Origin-Destination) traffic matrix, and the latest traffic flow data collected by real-time edge nodes are first imported into the simulation platform (such as SUMO). The simulation platform adopts a "residual-driven calibration" strategy, that is, after the initial loading is completed, the observation values ​​of the edge nodes are compared with the same indicators output in the simulation to calculate the simulation error, i.e.:

[0118] ;

[0119] in, This represents the average delay obtained in the simulation, in seconds. This represents the actual observed average delay, in seconds. For relative error, if If the simulation model matches reality well, the system will automatically adjust the simulation parameters or flow distribution until it is calibrated to an acceptable range.

[0120] Next, the timing parameters output in step S3 are loaded into the signal controller interface of the simulation platform, and the simulation is run in accelerated mode for a period of time. The simulation duration is usually set to... Instant arrival Within seconds. During this process, the simulation platform will continuously collect the following key performance indicators ( ),Right now:

[0121] The average speed of all vehicles in the simulation, expressed in meters per second;

[0122] : The average delay time of all vehicles in the simulation, in seconds;

[0123] Vehicles in simulation Total emissions, based on the velocity-emission model, are expressed in grams.

[0124] The on-time rate of buses in the simulation ranges from [value range missing]. ;

[0125] The average delay time of buses in the simulation, in seconds;

[0126] The average waiting time for pedestrians in the simulation, in seconds;

[0127] The pedestrian safety index in the simulation is calculated based on crossing waiting time and conflict probability, and its value ranges from [value range missing]. ;

[0128] If a sudden event is injected, the response time of the event in the simulation is in seconds.

[0129] If the above indicators meet the preset thresholds, the scheme is deemed to have passed verification; if any indicator fails to meet the thresholds, the difference information from this round of simulation needs to be reported. Return to the adaptive feedback control stage, continue to adjust the weights and parameters, and then iterate through step S3 again.

[0130] Step S4-2: After validating in normal scenarios, extreme scenarios must be tested. In the same digital twin environment, possible sudden scenarios are simulated sequentially: when a large-scale event ends and a large influx of people and vehicles suddenly occurs, the simulation will artificially increase the flow to a preset peak at a certain moment; if a main road is closed due to an accident or construction, the simulation will also forcibly reduce the road's capacity to zero; under severe weather conditions such as heavy rain or dense fog, the simulation will automatically reduce vehicle speeds and increase pedestrian crossing waiting times according to the corresponding speed model. If any of the above extreme simulations occur... Exceeding the threshold or If the value falls below the safety threshold, the current constraints are considered insufficient, and the relevant constraints need to be upgraded to hard constraints, with the penalty term applied. Increase these new constraints and return to step S3 to rebuild QUBO for optimization.

[0131] In step S5, the optimized scheme verified through digital twin simulation is converted into specific traffic signal control commands and issued to the traffic signal controller for execution, thus implementing it in real time within the traffic signal control system to ensure the accurate and effective execution of the optimized strategy. This can be achieved through the following methods:

[0132] Step S5-1: After all simulation verifications pass, the timing scheme is officially deployed to the actual traffic signal system. First, the timing parameters need to be... Replace with one that conforms to the local signal control protocol (such as...) or specific manufacturers The control command package is in a specific format. Each command package contains the signal number, phase sequence, and the corresponding light color duration for each phase.

[0133] Step S5-2: These command packets are encrypted and sent to each traffic signal controller via the edge gateway's private network or VPN link. For new-generation smart traffic signals already connected to V2I, they can communicate directly with the edge gateway via TCP / IP; while for older traffic signals that have not yet been upgraded, they are relayed to the cloud platform via a 4G / 5G public network VPN link. Upon receiving the command, the traffic signal controller needs to complete phase switching preparation within 2 seconds and execute it uniformly at the agreed time to ensure that the delay from command issuance to actual execution does not exceed the set maximum delay. .

[0134] Step S5-3: During the execution of the new scheme by the signal controller, the edge node will deploy a monitoring agent to report the signal controller's operating status in real time, including bulb status, controller operation logs, and abnormal alarms. If the edge node detects a delay or loss of instructions in the signal controller's command execution (i.e., ... This immediately triggers the "local backup plan," temporarily loading the timing scheme or classic heuristic scheme that was most recently verified by the digital twin onto the intersection until the fault is resolved. Simultaneously, this fault information is pushed to the operations and maintenance center.

[0135] Step S5-4: As the traffic signal controllers begin operating according to the new scheme, roadside sensors continuously collect data on actual traffic flow, vehicle speed, queue length, etc., and calculate the actual data in real time at the edge nodes. ,Right now:

[0136] : The average delay of all vehicles, in seconds;

[0137] Average queue length on actual road sections, in meters;

[0138] : The average delay time of actual public transport vehicles, in seconds;

[0139] : The average waiting time for actual pedestrians, in seconds;

[0140] Actual vehicle measurements Emissions, based on the velocity-emission model, are expressed in grams;

[0141] These actual With simulation To make a comparison, the simulation error is calculated, namely:

[0142] ;

[0143] If error If the threshold is exceeded, it indicates that there is a significant deviation between the digital twin model and reality. In this case, the next step is to enter adaptive feedback control to recalibrate the digital twin model (returning to the calibration stage in step S4); otherwise, the current scheme is maintained and the next round of data collection and optimization is carried out.

[0144] In step S6, the signal optimization effect is continuously monitored, and the optimization implementation status is fed back to the quantum computing model in real time to dynamically adjust the optimization parameters and strategies, achieving continuous optimization and adaptive control. This can be achieved through the following methods:

[0145] Step S6-1: The key to closed-loop control lies in adaptive feedback control. After each optimization cycle, the actual data generated in step S5 is collected. sequence:

[0146] ;

[0147] Also record the number of iteration layers used in this round of QAOA. Current quantum hardware error rate Information such as the number of times the classic rollback mechanism was triggered.

[0148] Step S6-2: After obtaining this real-time feedback data, recalculate the adaptive weight function. Specifically, the actual weight of the current period... of , , , Substituting into the soft maximization function, we get:

[0149] ;

[0150] in, Represents a certain point in the current cycle ; It corresponds The sensitivity parameter; the exponential function ensures that when a certain KPI value deviates significantly from expectations (e.g., the average delay is too high), the system will automatically increase the corresponding weight to prioritize reducing that KPI in the next optimization.

[0151] Step S6-3: Obtain the weight vector again Next, it is necessary to analyze the simulation and actual errors. To dynamically adjust the QAOA parameter configuration. If the actual error is small and the current quantum hardware error rate... If the situation is under control, then we choose to increase the number of QAOA iteration layers in the next round, that is:

[0152] ;

[0153] This enables quantum circuits to express more complex quantum states; otherwise, if the actual error is large or If the increase is significant, reduce the number of QAOA layers or increase the classic fallback trigger threshold, i.e.:

[0154] ;

[0155] in, This represents the maximum time window during which the quantum algorithm is allowed to run in this round. If this time window is exceeded, the algorithm will automatically revert to classical solutions. It is a decay parameter used to adjust how quickly classical backoff is triggered when the simulation error is large.

[0156] Step S6-4: Summarize the historical optimization results and execution feedback collected by local edge nodes during multiple rounds of optimization to the central server, and train two models at the central server using a federated learning mechanism: the first is a traffic state prediction model. This allows for a more precise understanding of historical characteristics. The first is to predict the feature distribution of future short-term windows; the second is the QUBO parameter prediction model. Used to directly extract from low-dimensional latent features Predicting linear coefficients With interaction coefficient This allows for a more efficient construction of the QUBO matrix during the next round of optimization.

[0157] When the urban road network structure is adjusted (e.g., adding new roads, upgrading traffic signals, or launching new bus routes), an offline model retraining is automatically triggered, and the updated model is distributed to all edge nodes so that they can obtain the latest training parameters, thereby obtaining more accurate features and QUBO construction results in subsequent iterations.

[0158] Step S6-5: When certain emergencies occur (such as intersection accidents, sudden traffic congestion due to heavy rain, etc.), the edge nodes can detect them immediately. The accident is flagged, and then the emergency optimization subprocess begins. First, the linear penalty at the intersection is strengthened, namely:

[0159] ;

[0160] Therefore, the linear coefficients related to the accident node are amplified. acc This forces the optimizer to prioritize traffic management around this node; then, when building the QUBO, the relevant conflict phases are penalized. The threshold for short-term parallel processing is dynamically lowered, allowing conflicting phases to open in parallel within a very short time to quickly release backlogged vehicles. Since this is an emergency optimization, a full QAOA is not required across the entire network. Instead, a partitioned parallel strategy is used, performing incremental optimization only on the affected area to quickly generate an emergency timing plan, which is then shared with... Submit the solution to the digital twin for accelerated simulation within seconds: If the simulation results show that the solution is better than the baseline delay value and no high-risk conflict occurs, immediately send it to the actual signal controller in step S5 and execute it; after execution, the incident handling process will be continuously monitored. If the delay exceeds expectations, we will revert to a more conservative local backup solution.

[0161] Through the implementation of the above steps, this invention, based on multi-source sensing and quantum enhancement optimization, extracts and reduces noise of the fused traffic features through temporal convolutional networks and self-attention mechanisms, effectively reducing the interference of sensor data and model errors on the optimization results. It can achieve high-precision coordinated timing of vehicles, bicycles, pedestrians and public transportation in the road network, providing strong support for the global optimization of city-level intelligent transportation and the popularization and development of vehicle-road cooperation.

[0162] Example 2

[0163] This embodiment discloses a quantum-enhanced real-time multimodal transportation network optimization system.

[0164] A quantum-enhanced real-time multimodal transportation network optimization system includes:

[0165] The data acquisition and fusion module is configured to: acquire multi-source motion data within the traffic network in real time;

[0166] The edge computing and preprocessing module is configured to perform edge computing and preprocessing on the acquired multi-source motion data to generate traffic state feature data.

[0167] The quantum optimization module is configured to generate signal schemes for different traffic control objectives based on the traffic state feature data and using a quantum computing model.

[0168] The digital twin simulation module is configured to: construct a digital twin model and verify the signal scheme;

[0169] The classic execution control module is configured to convert the verified signal scheme into traffic signal control commands and send them to the traffic signal controller for execution.

[0170] The adaptive feedback control module is configured to continuously monitor traffic conditions after traffic signal control commands are issued and feed the monitoring results back to the quantum computing model to adjust the signal scheme in real time.

[0171] Example 3

[0172] The purpose of this embodiment is to provide a computer-readable storage medium.

[0173] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of a quantum-enhanced real-time multimodal traffic network optimization method as described in Embodiment 1 of this disclosure.

[0174] Example 4

[0175] The purpose of this embodiment is to provide an electronic device.

[0176] An electronic device includes a memory, a processor, and a program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of a quantum-enhanced real-time multimodal transportation network optimization method as described in Embodiment 1 of this disclosure.

[0177] The steps and methods involved in the apparatuses of Embodiments 2, 3, and 4 above correspond to those in Embodiment 1. For specific implementation details, please refer to the relevant description section of Embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood as including any medium capable of storing, encoding, or carrying an instruction set for execution by a processor and enabling the processor to perform any of the methods in this invention.

[0178] Those skilled in the art will understand that the modules or steps of the present invention described above can be implemented using general-purpose computer devices. Optionally, they can be implemented using computer-executable program code, thereby allowing them to be stored in a storage device for execution by a computer device, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. The present invention is not limited to any particular combination of hardware and software.

[0179] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A quantum-enhanced real-time multimodal traffic network optimization method, characterized in that, include: Real-time acquisition of multi-source motion data within the transportation network; Edge computing and preprocessing are performed on the obtained multi-source motion data to generate traffic state feature data; Based on the traffic state characteristic data, a quantum computing model is used to generate signal schemes for different traffic control objectives, which consists of four steps: The first step, after all candidate decision variables and their corresponding low-dimensional features are ready, is to set an initial weight vector for the multi-objective optimization problem. This is used to measure the importance of different objectives; during the morning rush hour, if the priority is to reduce average delay, then set: ； in, This indicates the weight given to the objective of reducing average delay. This indicates the weight of the goal of reducing emissions. This indicates the weight of the goal of improving bus punctuality. This indicates the weight given to the goal of ensuring pedestrian safety; The second step is based on low-dimensional latent features. And the response curves obtained from training with historical data, for each variable Calculate the linear and interaction coefficients; first predict the response curve at the 1st... Node number adopts the first The performance change of the proposed scheme compared to the baseline scheme is as follows: ； in, Indicates the first Intersection No. 1 adopts the first The average delay reduction after implementing this plan; This indicates the corresponding reduction in emissions; This indicates the predicted on-time bus rate under this scheme; This indicates the pedestrian safety indicators predicted under this scheme; Represents binary variables The linear coefficients are determined so that the objective function can simultaneously reduce delays and emissions, improve bus punctuality and pedestrian safety. For any two binary decision variables and Calculate its second-order interaction coefficients If the solutions corresponding to these two variables are at the same intersection There is a phase conflict, at which point: ； Its meaning is: when There is a need to impose strong penalties on conflicts. For a large positive value; if ,intersection and Distance is And since there is a timing synergy effect between the two, let: ； in, A constant representing the degree of decay in timing coordination benefits; Indicates intersection and The Euclidean distance between them; It is a decay parameter that determines the rate at which the benefits of timing coordination decay as the distance increases. It is based on low-dimensional latent features The calculated coordination strength coefficient is obtained when the traffic correlation between nodes is stronger. The value is greater; The third step is to use a quantum approximation optimization algorithm combined with a classical backoff mechanism to solve this large-scale high-dimensional combinatorial optimization problem after the QUBO model is built. The fourth step is to obtain the optimal or near-optimal binary decision vector. Then, map it back to the specific timing; for all Make This determined the first Intersection No. 1 should adopt the first The green light duration of traffic lights is correlated with pedestrian and public transport priority strategies; this yields a complete set of timing parameters, namely: ； Among them, for the first Intersection No. 1: Indicates the duration of the first green light phase, in seconds; Indicates the duration of the yellow light transition, in seconds; Indicates the duration of the red light, in seconds; Construct a digital twin model and verify the signal scheme; The verified signal scheme is converted into traffic signal control commands and sent to the traffic signal controllers for execution; The system continuously monitors traffic conditions after traffic signal control commands are issued and feeds the monitoring results back to the quantum computing model to adjust the signal scheme in real time.

2. The quantum-enhanced real-time multimodal traffic network optimization method as described in claim 1, characterized in that, Multiple sensors used to collect multi-source motion data in the traffic network in real time achieve the acquisition of multi-source motion data through vehicle-to-everything (V2X) communication; at the same time, all sensors adopt a precise clock synchronization protocol to ensure consistent data timestamps.

3. The quantum-enhanced real-time multimodal traffic network optimization method as described in claim 1, characterized in that, The edge computing and preprocessing include: each edge node fuses multi-source motion data within the same spatial region and performs sliding window filtering and feature smoothing operations; by constructing feature vectors, the feature vectors of all edge nodes are concatenated into a global traffic state matrix by row; and by feature extraction and sorting, traffic state feature data is generated.

4. The quantum-enhanced real-time multimodal traffic network optimization method as described in claim 1, characterized in that, The signal scheme is verified, including: importing the required verification data into the digital twin model and calculating the simulation deviation; comparing the obtained simulation deviation with the set key performance indicators, and considering the verification as unsuccessful if any indicator fails to meet the standard.

5. The quantum-enhanced real-time multimodal traffic network optimization method as described in claim 1, characterized in that, Traffic signal control commands are encrypted via an edge gateway before being sent to the traffic signal controller.

6. The quantum-enhanced real-time multimodal transportation network optimization method as described in claim 1, characterized in that, The system continuously monitors traffic conditions after traffic signal control commands are issued. When an emergency occurs, the relevant edge node detects an accident marker and randomly applies linear penalty reinforcement to the current edge node.

7. A quantum-enhanced real-time multimodal traffic network optimization system, employing the traffic network optimization method as described in any one of claims 1-6, characterized in that, include: The data acquisition and fusion module is configured to: acquire multi-source motion data within the traffic network in real time; The edge computing and preprocessing module is configured to perform edge computing and preprocessing on the acquired multi-source motion data to generate traffic state feature data. The quantum optimization module is configured to generate signal schemes for different traffic control objectives based on the traffic state feature data and using a quantum computing model. The digital twin simulation module is configured to: construct a digital twin model and verify the signal scheme; The classic execution control module is configured to convert the verified signal scheme into traffic signal control commands and send them to the traffic signal controller for execution. The adaptive feedback control module is configured to continuously monitor traffic conditions after traffic signal control commands are issued and feed the monitoring results back to the quantum computing model to adjust the signal scheme in real time.

8. A computer-readable storage medium having a program stored thereon, characterized in that, When executed by a processor, the program implements the steps of a quantum-enhanced real-time multimodal transportation network optimization method as described in any one of claims 1-6.

9. An electronic device, comprising a memory, a processor, and a program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps in the quantum-enhanced real-time multimodal transportation network optimization method as described in any one of claims 1-6.

Images