A traffic flow electrical engineering monitoring system

Abstract

This invention discloses a traffic flow electrical engineering monitoring system, belonging to the field of intelligent transportation. The system includes: a perception layer with a multimodal electrical sensor array; an edge computing layer, whose nodes contain neuromorphic computing modules for encoding sensor signals into pulses and performing ultra-low power, millisecond-level event detection; and a digital twin layer, which constructs a virtual traffic model. This invention innovatively applies neuromorphic computing to front-end perception, solving the bottlenecks of edge real-time performance and energy efficiency. Simultaneously, it deeply integrates digital twins and federated learning, utilizing a global model in virtual space for policy simulation, optimization, and security verification, forming a closed loop of "perception-decision-simulation-control." This invention achieves high real-time performance, low power consumption, and privacy-preserving traffic monitoring and dynamic optimization, significantly improving the safety and efficiency of traffic management decisions.

Description

Technical Field

[0001] This invention relates to the fields of intelligent transportation and electrical engineering technology, specifically to a traffic flow monitoring system, and more particularly to an intelligent monitoring and optimization system and method that integrates neuromorphic computing, digital twins, and federated learning edge computing. Background Technology

[0002] With the development of smart cities, traffic flow monitoring technology is constantly evolving. Existing technologies mainly rely on single or combined sensors such as video surveillance, inductive loop detectors, and radar, and process data through centralized servers or cloud computing. However, these solutions have several industry pain points: 1) Insufficient real-time performance: video analysis has high latency and performance drops sharply in severe weather; 2) Low energy efficiency: edge computing nodes consume a lot of power and rely on continuous power supply; 3) Data silos and privacy risks: centralized data processing is prone to leaking sensitive information, and cross-system sharing is difficult; 4) High decision-making risks: control strategy optimization relies on historical data and simple models, making it difficult to conduct security simulations and rapid iterations in complex dynamic environments.

[0003] Neuromorphic computing, simulating the impulse information processing mechanism of the human brain, features event-driven operation, ultra-low power consumption, and extremely high parallelism, making it suitable for processing continuous spatiotemporal signal streams generated by sensors. Digital twins, by constructing high-fidelity virtual models of physical entities, can be used for simulation, prediction, and optimization. Federated learning allows multiple clients to train models locally and share only parameters, protecting data privacy. Currently, there are no reports on deeply integrating these three technologies to construct an integrated transportation system encompassing "millisecond-level perception," "virtual space safety decision-making," and "closed-loop control in the physical world." Summary of the Invention

[0004] This invention aims to overcome the shortcomings of existing technologies and provide a traffic flow monitoring and optimization system and method based on neuromorphic computing and digital twins. The system applies neuromorphic computing to edge perception to achieve ultra-low power consumption and ultra-high real-time event detection; simultaneously, it constructs a digital twin of the traffic system and combines it with a federated learning framework to perform secure and efficient decision-making and optimization in virtual space, ultimately forming a complete closed loop of perception-decision-simulation.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a traffic flow electrical engineering monitoring system, characterized in that it comprises:

[0006] The perception layer includes a multimodal electrical sensor array deployed on the roadside for acquiring raw traffic signals;

[0007] The edge computing layer includes multiple edge computing nodes, each node being connected to at least one sensor array of the perception layer. The edge computing nodes include neuromorphic computing modules for encoding the raw traffic signals into pulse sequences and processing them to generate structured traffic event data.

[0008] A digital twin layer, deployed in the cloud or on a regional server, constructs a virtual traffic model that maps to the physical traffic environment. The virtual traffic model is configured to: receive traffic event data from the edge computing layer to synchronize its state, and load a global prediction model trained by a federated learning framework; and in the virtual traffic model, run a simulation optimization algorithm based on the global prediction model and the current state to generate a traffic control strategy.

[0009] The system forms a closed loop, and the optimized traffic control strategy is verified and then sent to the physical traffic control equipment for execution.

[0010] Preferably, the neuromorphic computing module includes a neuromorphic chip configured to run a spiking neural network model, which employs pulse time-dependent plasticity rules for unsupervised learning to identify pulse events corresponding to traffic flow patterns.

[0011] Preferably, the multimodal electrical sensor array includes a piezoelectric sensor, a giant magnetoresistive sensor, and a geomagnetic sensor; the neuromorphic computing module is configured to perform differential pulse coding on signals from different sensors and input them into the spiking neural network model for fusion and event detection.

[0012] Preferably, a federated learning framework is deployed between the edge computing layer and the digital twin layer;

[0013] The edge computing node is configured to: train a local prediction model using local traffic event data and upload encrypted model parameters to the digital twin layer;

[0014] The digital twin layer is configured to aggregate model parameters from multiple edge computing nodes to update the global prediction model, and then distribute the updated global prediction model parameters.

[0015] Preferably, the digital twin layer is configured to perform multi-scenario simulation verification in the virtual traffic model before applying the updated global prediction model to the physical world.

[0016] Preferably, the simulation optimization algorithm is a deep reinforcement learning algorithm, whose reward function is constructed based on the traffic efficiency, delay time or congestion index in the virtual traffic model, and is used to optimize and generate traffic light timing schemes in the simulation.

[0017] Preferably, the edge computing node further includes a traditional microprocessor, the neuromorphic computing module acts as a coprocessor, specifically responsible for feature extraction and event detection of the original signal, and the traditional microprocessor is responsible for protocol communication, task scheduling, and data interaction with the digital twin layer.

[0018] A method for monitoring traffic flow electrical engineering includes the following steps:

[0019] S1: Acquire the raw electrical signals of the road surface through a multimodal electrical sensor array;

[0020] S2: At the edge computing node, the original electrical signal is encoded into a pulse sequence using a neuromorphic computing module and processed by a spiking neural network model to generate traffic event data in real time;

[0021] S3: In the digital twin layer, traffic event data from various edge computing nodes is used to drive the state synchronization of the virtual traffic model;

[0022] S4: In the virtual traffic model, based on the global prediction model generated by federated learning and the current state, a simulation optimization algorithm is run to obtain an optimized traffic control strategy;

[0023] S5: Simulate and verify the optimized traffic control strategy in the virtual traffic model;

[0024] S6: The verified traffic control strategy is sent to the traffic control equipment in the physical world for execution, completing the closed-loop control.

[0025] Compared with the prior art, the beneficial effects of the present invention are:

[0026] Extreme real-time performance and ultra-low power consumption: Neuromorphic computing enables millisecond-level event detection with power consumption only one-thousandth that of traditional solutions, making it suitable for solar-powered edge scenarios.

[0027] Decision-making safety and efficiency: All control strategies are fully simulated and tested in a digital twin virtual environment, eliminating the risk of "bad strategies" directly affecting real traffic. The simulation speedup ratio is high, and the optimization cycle is significantly shortened.

[0028] Privacy protection and collaborative intelligence: Through federated learning, global model collaborative evolution is achieved without data leaving the local environment, breaking down data silos and protecting data privacy.

[0029] System closed loop and self-optimization: A complete closed loop of "physical perception - edge processing - cloud twin simulation - policy distribution" is formed, enabling the system to continuously learn and self-optimize. Attached Figure Description

[0030] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0031] Figure 1 This is a block diagram of the traffic flow electrical engineering monitoring system of the present invention;

[0032] Figure 2 This is a flowchart of the traffic flow electrical engineering monitoring method of the present invention. Detailed Implementation

[0033] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0034] Please see Figures 1 to 2 The present invention provides a technical solution:

[0035] A traffic flow electrical engineering monitoring system, comprising:

[0036] The perception layer includes a multimodal electrical sensor array deployed on the roadside for acquiring raw traffic signals;

[0037] The edge computing layer includes multiple edge computing nodes, each node being connected to at least one sensor array of the perception layer. The edge computing nodes include neuromorphic computing modules for encoding the raw traffic signals into pulse sequences and processing them to generate structured traffic event data.

[0038] A digital twin layer, deployed in the cloud or on a regional server, constructs a virtual traffic model that maps to the physical traffic environment. The virtual traffic model is configured to: receive traffic event data from the edge computing layer to synchronize its state, and load a global prediction model trained by a federated learning framework; and in the virtual traffic model, run a simulation optimization algorithm based on the global prediction model and the current state to generate a traffic control strategy.

[0039] The system forms a closed loop, and the optimized traffic control strategy is verified and then sent to the physical traffic control equipment for execution.

[0040] Furthermore, the core of the neuromorphic computing module is a neuromorphic chip (such as Intel Loihi), which runs a spiking neural network model and uses pulse time-dependent plasticity rules for unsupervised learning to adaptively identify vehicle pulse patterns at specific intersections.

[0041] Furthermore, the system integrates a federated learning framework. Edge nodes train lightweight prediction models locally, and only the model parameters are encrypted and uploaded to the digital twin layer for secure aggregation, forming a globally optimized model. Before being applied to the physical world, this global model undergoes multi-scenario, high-concurrency simulation verification within the digital twin to ensure the policy's security and effectiveness.

[0042] Furthermore, the simulation optimization algorithm used in the digital twin layer is a deep reinforcement learning algorithm. Its reward function R can be designed as the negative of the total delay time, i.e. .

[0043] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0044] like Figure 1 As shown, the system is divided into three layers. The sensing layer consists of an array of piezoelectric, giant magnetoresistive, and geomagnetic sensors arranged in a certain topology, responsible for collecting raw analog signals such as changes in charge and magnetic fields caused by the passing of vehicles. These signals are transmitted to the edge computing layer.

[0045] Edge computing node hardware includes a main control MCU and a neuromorphic chip that acts as a coprocessor. For example... Figure 2 As shown, the continuous analog signal from the sensor is first preprocessed and then encoded into a pulse sequence. For example, a pulse is generated when the piezoelectric sensor signal exceeds a threshold θ:

[0046]

[0047] The pulse sequence is input into a pre-trained pulse convolutional neural network. The network follows the STDP rule.

[0048]

[0049] The system learns local traffic patterns and ultimately outputs pulse clusters representing events such as "vehicles passing in a type A manner" and "congestion begins," which are then converted into structured traffic event data by the main MCU.

[0050] This data is used to train a lightweight federated learning client model (such as a small LSTM prediction model) locally, and is also uploaded to the digital twin layer via an encrypted channel to drive the vehicle entities in the virtual traffic model and maintain virtual-real synchronization.

[0051] In the digital twin layer, a high-fidelity virtual road environment and real-time traffic flow are constructed. Parameters from federated learning models at various edge nodes are securely aggregated here to form a more accurate global traffic prediction model. A decision engine (e.g., a deep reinforcement learning agent) takes the current virtual traffic state and prediction results as input, tries different signal control actions in the simulation environment, and learns the optimal policy π* based on rewards (e.g., negative total delay). The learned new policy is then stress-tested in the digital twin environment (e.g., simulating traffic accidents, heavy rain, etc.) to verify its robustness. Once verified, the policy is compiled into control commands and sent to the physical intersection signal controllers for execution, thus completing one optimization loop.

[0052] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A traffic flow electrical engineering monitoring system, characterized by, include: The perception layer includes a multimodal electrical sensor array deployed on the roadside for acquiring raw traffic signals; The edge computing layer includes multiple edge computing nodes, each node being connected to at least one sensor array of the perception layer. The edge computing nodes include a neuromorphic computing module for encoding the raw traffic signals into pulse sequences and processing them to generate structured traffic event data. A digital twin layer, deployed in the cloud or on a regional server, constructs a virtual traffic model that maps to the physical traffic environment. The virtual traffic model is configured to: receive traffic event data from the edge computing layer to synchronize its state, and load a global prediction model trained by a federated learning framework; and in the virtual traffic model, run a simulation optimization algorithm based on the global prediction model and the current state to generate a traffic control strategy. The system forms a closed loop, and the optimized traffic control strategy is verified and then sent to the physical traffic control equipment for execution.

2. The traffic flow electrical engineering monitoring system according to claim 1, characterized in that: The neuromorphic computing module includes a neuromorphic chip configured to run a spiking neural network model, which uses pulse time-dependent plasticity rules for unsupervised learning to identify pulse events corresponding to traffic flow patterns.

3. The traffic flow electrical engineering monitoring system according to claim 1, characterized in that: The multimodal electrical sensor array includes piezoelectric sensors, giant magnetoresistive sensors, and geomagnetic sensors; the neuromorphic computing module is configured to perform differential pulse encoding on signals from different sensors and input them into the spiking neural network model for fusion and event detection.

4. The traffic flow electrical engineering monitoring system according to claim 1, characterized in that: A federated learning framework is deployed between the edge computing layer and the digital twin layer; The edge computing node is configured to: train a local prediction model using local traffic event data and upload encrypted model parameters to the digital twin layer; The digital twin layer is configured to aggregate model parameters from multiple edge computing nodes to update the global prediction model, and then distribute the updated global prediction model parameters.

5. The traffic flow electrical engineering monitoring system according to claim 4, characterized in that: The digital twin layer is configured to perform multi-scenario simulation verification in the virtual traffic model before applying the updated global prediction model to the physical world.

6. The traffic flow electrical engineering monitoring system according to claim 1, characterized in that: The simulation optimization algorithm is a deep reinforcement learning algorithm. Its reward function is constructed based on the traffic efficiency, delay time or congestion index in the virtual traffic model, and is used to optimize and generate traffic light timing schemes in the simulation.

7. The traffic flow electrical engineering monitoring system according to claim 1, characterized in that: The edge computing node also includes a traditional microprocessor. The neuromorphic computing module acts as a coprocessor, specifically responsible for feature extraction and event detection of the original signal. The traditional microprocessor is responsible for protocol communication, task scheduling, and data interaction with the digital twin layer.

8. A traffic flow electrical engineering monitoring method applied to any one of claims 1-7, characterized in that, Includes the following steps: S1: Acquire the raw electrical signals of the road surface through a multimodal electrical sensor array; S2: At the edge computing node, the original electrical signal is encoded into a pulse sequence using a neuromorphic computing module and processed by a spiking neural network model to generate traffic event data in real time; S3: In the digital twin layer, traffic event data from various edge computing nodes is used to drive the state synchronization of the virtual traffic model; S4: In the virtual traffic model, based on the global prediction model generated by federated learning and the current state, a simulation optimization algorithm is run to obtain an optimized traffic control strategy; S5: Simulate and verify the optimized traffic control strategy in the virtual traffic model; S6: The verified traffic control strategy is sent to the traffic control equipment in the physical world for execution, completing the closed-loop control.

Images