A traffic signal adaptive control system and method

Abstract

The application discloses a traffic signal adaptive control system and method, comprising: a perception and transmission module, which collects multi-modal traffic flow data of an intersection in real time and compresses and transmits; a space-time graph module, which abstracts the intersection into a dynamic space-time graph according to the traffic flow data; a node represents a physical lane, an edge connects adjacent intersections or upstream and downstream road sections, and the weight of the edge is dynamically calculated and updated in real time according to historical turning probabilities, phase constraint influence factors and real-time queue length influence factors; a decision module, which generates an action space containing phase switching timing according to a state space based on the space-time graph, so that the number of vehicles passing through per unit time is maximized and the maximum waiting time difference of each direction is minimized; and a phase dynamic optimization module, which receives the action space, responds to a regular optimization trigger condition, dynamically adjusts the phase green light duration and switching sequence, and can solve the problems of long-term waiting of vehicles on the next arterial road and the decrease of intersection traffic efficiency in the tidal flow or emergency scene.

Description

Technical Field

[0001] This invention relates to the field of traffic control system technology, specifically to a traffic signal adaptive control system and method. Background Technology

[0002] In urban intelligent transportation management systems, the dynamic optimization and configuration of traffic signals has become a core element in improving road network operational efficiency. Existing adaptive signal control systems mainly rely on a single data source or static optimization, lacking real-time data-driven capabilities and struggling to cope with the spatiotemporal dynamics of traffic flow.

[0003] For example, there is a Chinese patent with publication number CN104040605B, which relates to a traffic signal control method and a traffic signal controller, realizing the control of traffic signals. However, this patent ignores the upstream and downstream road network topology relationships, such as turning probability and queuing overflow effect, and is prone to falling into the trap of local optima. It only optimizes single-point congestion, exacerbates secondary queuing at related intersections, and its phase decision does not incorporate historical traffic patterns, making it unable to adapt to scenarios such as tidal flow. Taking "maximizing network throughput" as the single objective and ignoring the lane-level waiting time balance will lead to a significant increase in vehicle delays on secondary arterial roads when traffic flow is uneven. Summary of the Invention

[0004] To address the issues of long waiting times on secondary arterial roads and reduced intersection efficiency caused by a lack of vehicle dynamic intent perception and insufficient spatiotemporal correlation between upstream and downstream traffic in tidal flow or emergency scenarios, this invention proposes a traffic signal adaptive control system and method. By quantifying the coupling effect of historical turning probabilities and real-time phase conflicts through spatiotemporal graphs and combining it with dual-objective optimization, the system achieves adaptive adaptation to changes in traffic demand.

[0005] A further objective of this invention is to prevent wasted or insufficient green light time through dynamic phase splitting and priority rearrangement.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a traffic signal adaptive control system, comprising:

[0007] The sensing and transmission module collects multimodal traffic flow data at intersections in real time and compresses and transmits it.

[0008] The spatiotemporal graph module abstracts intersections into dynamic spatiotemporal graphs based on traffic flow data: nodes represent physical lanes, edges connect adjacent intersections or upstream and downstream road segments, and the weights of the edges are dynamically calculated and updated in real time based on historical turning probabilities, phase constraint influence factors, and real-time queue length influence factors.

[0009] The decision module, based on the spatiotemporal diagram, generates an action space containing phase switching timing according to the state space, so as to maximize the number of vehicles passing through per unit time and minimize the difference in maximum waiting time in each direction.

[0010] The phase dynamic optimization module receives the action space, responds to the conventional optimization trigger conditions, and dynamically adjusts the phase green light duration and switching sequence.

[0011] In this technical solution, real-time data acquisition and ultra-low latency, high-compression transmission are achieved through the perception and transmission modules; dynamic topology modeling is realized through the spatiotemporal graph module, which introduces dynamic spatiotemporal graphs for the first time to represent intersections, dynamically adjusting edge weights, and taking into account the impact of phase conflicts and real-time queue lengths to improve traffic efficiency and control flexibility; the decision-making module adopts dual-objective optimization to simultaneously satisfy the maximum number of vehicles passing through per unit time and the minimum difference in maximum waiting time in each direction, balancing efficiency and waiting time balance; and millisecond-level green light adjustment is achieved through the phase dynamic optimization module.

[0012] Preferably, the phase dynamic optimization module includes: responding to emergency event requests, inserting a priority phase and locking the red light in the conflicting direction; when a V2X emergency event is detected, interrupting the current service program, locking the green light in the priority passage direction and the red light in the conflicting direction within the response time (less than 1 second), introducing an emergency response mechanism, eliminating the need for manual processing, and improving emergency passage efficiency and safety; executing at least one dynamic adjustment strategy among phase splitting, merging, skipping, or priority reordering; enabling the system to have "anti-mutation immunity" capabilities, quickly completing signal reconfiguration in emergency scenarios such as accidents and rescues, avoiding congestion spread, and shortening emergency response time.

[0013] Preferably, the weight of the edge is obtained by summing the historical turning probability, phase constraint influence factor, and real-time queue length influence factor after applying their respective weight coefficients: the weight coefficient of the historical turning probability is dynamically adjusted according to the sudden event; the weight coefficient of the phase constraint influence factor is dynamically adjusted according to the phase conflict matrix; the weight coefficient of the real-time queue length influence factor is dynamically adjusted according to the current traffic flow; the weight adaptive mechanism enables the spatiotemporal map to quickly complete topology reconstruction in sudden scenarios such as accidents and lane closures, thereby improving the real-time performance and accuracy of the control strategy.

[0014] Preferably, the conventional optimization triggering condition is: when the lane saturation is greater than the first threshold or the queue length difference is greater than the second threshold, the green light duration is adjusted; the daily congestion self-recovery is achieved through a dual-threshold triggering strategy, for example, when the saturation is >80% or the queue length difference is >40%, the bottleneck phase green light is automatically extended to improve peak-hour traffic capacity.

[0015] Preferably, the dynamic adjustment strategy includes: splitting a single-direction traffic surge in a composite phase into an independent phase; inserting a co-release phase into a symmetrical release phase; skipping a phase when there are no vehicles passing for multiple consecutive cycles; merging multiple independent phases into a composite phase when there is severe idle traffic, thereby maximizing the utilization rate of green lights through flexible phase control, reducing idle traffic rate, reducing signal cycle loss, and reducing the average vehicle delay during off-peak hours.

[0016] Preferably, the dynamic adjustment strategy includes: when the proportion of flow in the tidal flow direction is greater than the third threshold, increasing the priority of the phase in that direction and extending the green light duration; increasing the priority of high-demand phases to improve the flexibility and rationality of resource flow.

[0017] Preferably, in the spatiotemporal graph module, the attributes of the nodes include current flow, average speed, and saturation, and the spatiotemporal graph module reconstructs the graph structure at regular intervals to update the real-time status.

[0018] Preferably, the sensing and transmission module uses a Kalman filter algorithm to correct the drift of the high-precision positioning data collected by the vehicle terminal and matches it with the high-precision map to lane-level accuracy, ensuring the real-time performance and accuracy of vehicle dynamic data.

[0019] Preferably, in the decision-making module, the state space includes the current phase combination, the number of vehicles in each direction, the saturation in each direction, and the remaining time of the phase, and the action space includes the green light duration adjustment amount, the red light duration adjustment amount, and the phase priority adjustment coefficient.

[0020] The present invention also adopts the following technical solution: a traffic signal adaptive control method, based on the above-mentioned traffic signal adaptive control system, comprising the following steps:

[0021] S1 collects vehicle dynamic data, lane-level traffic parameters and historical traffic flow data in real time to build a perception network;

[0022] S2, construct a dynamic spatiotemporal graph based on the sensing network and the current phase state;

[0023] S3 generates a signal control strategy based on the state data of the spatiotemporal diagram;

[0024] S4, execute the strategy and adjust the traffic lights, and feed back the execution results to the perception network for re-collection.

[0025] The beneficial effects of this invention are:

[0026] 1) Construct an integrated vehicle-road-cloud architecture to break through the limitations of perception dimensions and achieve accurate prediction of sudden behaviors;

[0027] 2) Dynamic spatiotemporal graph modeling is adopted, nodes are dynamically reconstructed, and edge weights are adaptively adjusted according to historical turning probabilities, phase conflict influence factors and real-time queuing influence factors to eliminate local optimization traps and improve traffic efficiency.

[0028] 3) A dynamic phase adjustment strategy is adopted, which automatically skips low-flow phases and extends the green light for high-flow phases, reducing signal cycle loss and improving emergency response speed;

[0029] 4) Adopt a dual-objective optimization of traffic efficiency and waiting time balance to maintain lane-level waiting time balance and prevent increased vehicle delays on secondary arterial roads when traffic flow is uneven;

[0030] 5) Achieve high-precision regional-level adaptive control of traffic signals while balancing the contradiction between global collaborative optimization and local real-time response. Attached Figure Description

[0031] Figure 1 This is a flowchart of the method in Embodiment 2 of the present invention.

[0032] Figure 2 This is a phase pattern diagram of the intersection signal scheme in Embodiment 2 of the present invention.

[0033] Attached label: North-South entrance straight ahead 1, North-South entrance left turn 2, East-West entrance straight ahead 3, East-West entrance left turn 4, North entrance straight ahead and North entrance left turn 5, South entrance straight ahead and North entrance left turn 6, East entrance straight ahead and North entrance left turn 7, West entrance straight ahead and North entrance left turn 8. Detailed Implementation

[0034] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only one preferred embodiment of this invention and are only used to explain this invention. They do not limit the scope of protection of this invention. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0035] Example 1

[0036] This embodiment provides a traffic signal adaptive control system for achieving millisecond-level traffic light regulation in complex scenarios such as congestion, imbalance, tidal flow, and emergencies. It includes a sensing and transmission module, a spatiotemporal map module, a decision-making module, and a phase dynamic optimization module. All modules are interconnected through a 5G NR V2X network with an end-to-end latency of no more than 10ms and a peak rate of up to 100Mbps.

[0037] The sensing and transmission module includes a sensing unit and a data acquisition unit.

[0038] The data collection unit consists of three layers working together: vehicle-mounted, roadside, and cloud-based.

[0039] The vehicle-mounted unit (OBU) collects high-precision positioning, instantaneous speed, turn signal status, and autonomous driving decision-making intent in real time at a frequency of no less than 10Hz. The data is first filtered by Kalman filtering to eliminate positioning drift, and then matched with a high-precision map to the specific lane.

[0040] RSU radar-visual fusion equipment is deployed on traffic light poles or gantries along the roadside to continuously output the number of vehicles passing through each lane, arrival rate, queue length, and headway. As a supplement, geomagnetic coils provide occupancy data to compensate for the lack of video detection in adverse weather conditions.

[0041] The system accesses a historical traffic database in the cloud, summarizes long-term traffic flow characteristics at 1-minute or 5-minute granularity, and pulls weather information such as visibility and precipitation intensity in real time. All time-series features are mapped to a unified vector space through the Time2Vec encoder, which facilitates deep model processing.

[0042] To reduce channel load, the transmission unit uses a wavelet transform compression algorithm at the transmitting end to compress the original data by at least 80%, and then broadcasts it to the entire network via 5G NR V2X to achieve millisecond-level data sharing.

[0043] The spatiotemporal diagram module abstracts intersections into dynamic spatiotemporal diagrams.

[0044] Nodes correspond to physical lanes, and their attributes include current traffic flow, average speed, and saturation. Edges are divided into spatial edges and temporal edges. Spatial edges connect upstream and downstream lanes and are used to describe the physical connectivity of "where the cars come from and where they go." Temporal edges connect the same lane at adjacent time steps, depicting the evolution of traffic flow over time.

[0045] The weights of edges are no longer simply marked by whether there is a conflict, but are dynamically calculated by combining historical turning probabilities, phase constraint influence factors and real-time queue length influence factors, as detailed below.

[0046] The weighting coefficient for historical turning probability is 0.7 when there are no emergencies, and drops below 0.3 when there are emergencies; the phase constraint influence factor is 1 when the light is green and 0 when the light is red, and can be temporarily increased to 1.5 when the vehicle has priority; the real-time queue length influence factor is 0.3 during off-peak hours and 0.6 during peak hours.

[0047] The entire graph structure is reconstructed at fixed time intervals (e.g., 15 seconds), and node attributes and edge weights are updated in real time to ensure a second-level response to traffic conditions.

[0048] The decision-making module generates signal control strategies based on the spatiotemporal diagram.

[0049] The state space includes the current phase combination, the number of vehicles in each direction, the saturation in each direction, and the remaining time of the phase; the action space includes the green light duration adjustment, the red light duration adjustment, and the phase priority adjustment coefficient.

[0050] The reward function balances traffic efficiency and waiting time balance: traffic efficiency is measured by the ratio of the number of vehicles passing through a single cycle to the cycle length; waiting time balance is measured by the normalized difference between the maximum waiting time and the average waiting time in each direction, and the goal is to minimize this difference.

[0051] Deep reinforcement learning algorithms (such as PPO combined with spatiotemporal Transformer) can output the optimal action within milliseconds, enabling real-time optimization of traffic lights.

[0052] The phase dynamic optimization module is responsible for converting the strategy into an executable signal.

[0053] The conventional triggering mechanism evaluates the situation after each signal cycle. If the saturation of any lane exceeds 80% or the difference in queue length exceeds 40%, the green light duration is adjusted according to the optimization results.

[0054] Upon receiving an emergency request (such as an accident, rescue, or bus priority) from the vehicle-mounted V2X, the emergency response mechanism immediately interrupts the current procedure, inserts the priority phase, and keeps the red light locked in the conflicting direction.

[0055] The dynamic strategy library supports phase splitting, merging, skipping, and priority reordering: when traffic surges in a certain direction within a composite phase, it can be split into an independent phase; when no vehicles pass through for several consecutive cycles, the phase can be skipped to reduce idle green lights; when multiple independent phases are idle, they can be merged into a composite phase; high-demand phases output by reinforcement learning (such as tidal flow directions) can be automatically released in advance.

[0056] The entire control process forms a closed loop of "real-time perception - intelligent decision-making - precise control". The execution effect is fed back in real time by the perception and transmission modules for the next cycle diagram reconstruction and strategy update.

[0057] The system's operation process is described in detail below.

[0058] The full-domain perception network collects and compresses data in real time, and transmits it to the cloud via 5G NR V2X; the spatiotemporal map module reconstructs the traffic map in seconds; the decision-making module generates strategies in milliseconds; and the phase dynamic optimization module performs signal adjustments and feeds the results back to the perception network, realizing adaptive management of traffic signals throughout the entire process.

[0059] The traffic signal adaptive control system of this embodiment has dynamic topology reconstruction capability and millisecond-level response speed, which can effectively cope with complex scenarios such as congestion, imbalance, tidal flow, and sudden events. It provides a scalable decision framework for intelligent transportation systems and promotes the paradigm shift of urban road signal control from "fixed timing" to "intelligent regulation".

[0060] Example 2

[0061] This embodiment provides a traffic signal adaptive control method, based on the aforementioned traffic signal adaptive control system, such as... Figure 1 As shown, it includes the following steps.

[0062] Step S1: Construction of the vehicle-road-cloud collaborative perception network.

[0063] Deploy onboard units (OBU), roadside sensing units (RSU), and geomagnetic coils to collect multi-source data, including real-time vehicle dynamic data, lane-level traffic parameters, and historical traffic flow data.

[0064] The collected vehicle dynamic data includes location, speed, and acceleration; lane-level traffic parameters include occupancy, queue length, and headway; and historical traffic flow data includes data from the same historical period.

[0065] The system constructs a three-dimensional perception network through three types of data sources, as detailed below.

[0066] Vehicle-side dynamic data: Real-time high-precision vehicle positioning (GPS / BeiDou), instantaneous speed, turn signal status, and autonomous driving decision intentions, such as lane changing and braking, are obtained through V2X communication. The sampling frequency is ≥10Hz. After eliminating positioning drift through the Kalman filter algorithm, the data is matched to the specific lane with the high-precision map.

[0067] Roadside perception data: By deploying radar-visual fusion equipment, video streams are used to analyze the number of vehicles passing through each lane, arrival rate, queue length, and headway.

[0068] Cloud-based collaborative data: Integrating historical traffic databases, weather information, road network structure, etc., and mapping temporal features into vector space through the Time2Vec encoder.

[0069] The historical traffic database is integrated at a certain time granularity, such as 1 minute, 5 minutes, etc.; weather information includes, but is not limited to, visibility and precipitation intensity.

[0070] In this embodiment, the data transmission mechanism can use 5G NR V2X communication technology to achieve data transmission, supporting a peak rate of 100Mbps, an end-to-end latency of ≤10ms, and a data compression algorithm based on wavelet transform to compress the original data volume by more than 80%.

[0071] Step S2: Abstract the intersection into a dynamic spatiotemporal diagram.

[0072] The nodes in the dynamic spatiotemporal graph represent "physical lanes," breaking down intersections into specific lanes, such as the left-turn lane at the east entrance and the straight-through lane at the west exit.

[0073] The edge includes spatial edge and temporal edge. Spatial edge connects the upstream and downstream lanes, reflecting the physical connectivity of "where the car comes from and where it goes". Temporal edge connects the same lane at two time steps t and t+1, representing the evolution of traffic flow over time.

[0074] In dynamic spatiotemporal graphs, edge weights are no longer simply a matter of conflict or non-conflict, but rather a combination of historical turning probabilities, current signal phases, and dynamic values ​​of real-time traffic flow, speed, and saturation. This allows them to be used to predict or optimize traffic conditions at the next moment.

[0075] To characterize traffic flow turning relationships and phase conflict rules, the edge weights are determined by the historical turning probabilities of vehicle trajectories and signal control constraints, and are dynamically updated in real time based on the current phase state.

[0076] A turning relationship refers to the travel path of a vehicle from one entrance lane (upstream node) to another exit lane (downstream node). For example, the journey from the east entrance left-turn lane to the north exit straight lane is a turning relationship from "east" to "north left-turn". In the dynamic spatiotemporal diagram of an intersection, the turning relationship is represented by a spatial edge, pointing from the upstream node to the downstream node.

[0077] Phase conflict refers to the conflict of traffic flow trajectories between different signal phases at the same time or in adjacent time periods in traffic signal control. Different signal phases are combinations of green light releases in different directions, which are specifically manifested as physical conflicts and signal conflicts.

[0078] Physical conflict refers to the overlapping paths of traffic flows from different phases in space, such as intersections or merging, which may lead to collisions. For example, when the eastbound left-turn phase (green light) and the northbound straight-ahead phase (green light) are both open, left-turning vehicles must cross the straight-ahead traffic, creating a point of intersection conflict. When northbound and southbound straight-ahead traffic is open at the same time, if the phase is not cleared during the full red time, vehicles may be stuck at the intersection.

[0079] Signal conflict refers to a time conflict that occurs when traffic flow in the previous phase is not cleared and traffic flow in the next phase is allowed to proceed during phase switching. For example, after the green light for east-west straight traffic ends, the left-turn phase starts immediately, but straight-going vehicles may conflict with left-turning vehicles because the queue has not completely cleared the intersection.

[0080] Phase conflicts can be characterized using a conflict matrix. The conflict matrix is ​​a binary table that marks which phases cannot be allowed to proceed simultaneously; a value of 1 indicates a conflict, and a value of 0 indicates compatibility.

[0081] Historical turning probability refers to the probability of a certain turning relationship occurring according to past statistics. For example, the probability of turning from east left to the north exit according to past statistics is 65%. Historical turning probability can determine the probability of potential conflict, and signal control constraints can be reflected by the phase conflict matrix.

[0082] The current phase state refers to the green light phase that the signal controller is currently allowing. The traffic signal adaptive control method of the present invention also dynamically updates the edge weights in real time according to the current phase state. For example, when the current phase is green, the edge weight of the corresponding flow direction is increased, while according to the conflict matrix, the weight of the conflict phase is reduced to zero, so that the weight of the associated edge is suppressed, and the algorithm avoids selecting a signal strategy that may cause conflict.

[0083] For example, when the green light is on for left turns from north to south, the weight of the corresponding left-turn side is increased, while the weight of the side with conflict phase clearance flow is reduced to zero.

[0084] Specifically, in this embodiment, the comprehensive weight of the edge is obtained by summing the historical turning probability, the phase constraint influence factor, and the real-time queue length influence factor after applying their respective weight coefficients.

[0085] The three weighting coefficients generally take values ​​between 0 and 1.

[0086] The weighting coefficient of historical turning probability can be adaptively adjusted according to the scenario. For example, when no event occurs, it relies more on historical turning patterns, and the weighting coefficient of historical turning probability can be 0.7. In the event of a sudden event, the weighting coefficient of historical turning probability is reduced to below 0.3 to weaken the influence of historical data and prioritize responding to real-time changes.

[0087] The weighting coefficient of the phase constraint influence factor is 1 during the green light phase, which opens the corresponding turn signal; it is 0 during the red light phase, which completely blocks the turn signal; and it is 1.5 when the vehicle has priority, which temporarily increases the weight to speed up the clearing process.

[0088] The weighting coefficient of the real-time queue length influencing factor is adaptively adjusted according to the scenario. Specifically, the queue length has a small impact on the intersection during off-peak hours, and the weighting coefficient of the real-time queue length influencing factor can be taken as 0.3; the queue length during peak hours reflects the congestion of the intersection, and the weighting coefficient of the real-time queue length influencing factor can be taken as 0.6.

[0089] Step S3: Generate signal control strategies based on the spatiotemporal graph using a deep reinforcement learning framework.

[0090] The state space should contain key information that can affect the efficiency of traffic signal control in each direction.

[0091] Considering the key characteristics of intersection signal control and the road network structure, this embodiment defines the state space of intelligent traffic signal control at intersections as a set including the current phase combination, the number of vehicles in each direction, the saturation of each direction, and the remaining time of the phase.

[0092] In order to better manage traffic flow at intersections, traffic signal control systems need to select appropriate phases based on the current intersection conditions.

[0093] When designing the action space, select an action space for intelligent traffic signal control at intersections based on a phase pattern.

[0094] Based on the current state, the agent can choose to continue with the current phase or switch to the next phase with higher priority.

[0095] Here, the action space is defined as a set including the green light duration adjustment, the red light duration adjustment, and the phase priority adjustment coefficient.

[0096] Taking a crossroads as an example, without considering right turns, there are eight phase modes for the intersection signal scheme: 1. North-South entrance straight ahead; 2. North-South entrance left turn; 3. East-West entrance straight ahead; 4. East-West entrance left turn; 5. North entrance straight ahead and North entrance left turn; 6. South entrance straight ahead and North entrance left turn; 7. East entrance straight ahead and North entrance left turn; 8. West entrance straight ahead and North entrance left turn. Figure 2 As shown.

[0097] Then, a bi-objective optimization function is designed to balance traffic efficiency and waiting time balance. Traffic efficiency refers to the number of vehicles passing through per unit time, while waiting time balance refers to the maximum difference in waiting time in each direction.

[0098] In this embodiment, two objective functions are adopted: traffic efficiency and waiting time balance, as detailed below.

[0099] Traffic efficiency is obtained by the ratio of the sum of the number of vehicles passing through N phases in a single cycle to the total duration of the signal cycle. The number of vehicles in each lane during the green light period is counted in real time by roadside sensing equipment, and the green light duration is dynamically adjusted to reduce the idle green light time.

[0100] Waiting time balance is obtained by normalizing the difference between the maximum waiting time of each lane and the average waiting time of all lanes by dividing it by the average waiting time of all lanes. The cumulative waiting time of each vehicle is obtained through vehicle-to-vehicle V2X to balance the waiting time of vehicles in all directions and balance the vehicle delays in each direction.

[0101] Step S4, phase dynamic optimization logic.

[0102] The real-time optimization trigger mechanism includes regular optimization and emergency response.

[0103] Regular optimization refers to initiating a strategy evaluation after each signal cycle. When lane saturation is detected to be >80% or queue length difference is detected to be >40%, the green light duration of the signal scheme is adjusted.

[0104] Emergency response refers to immediately initiating the interruption service procedure after receiving an emergency event (such as an accident, rescue, or other priority request) from the vehicle-mounted V2X, inserting the priority release phase or maintaining the priority phase green light and locking the conflicting direction red light.

[0105] The dynamic phase sequence adjustment strategy includes phase splitting, phase merging and skipping, phase time optimization, and priority reordering. The dynamic phase sequence adjustment strategy is described in detail below.

[0106] When traffic flow surges in a certain direction during a composite phase (such as north-south straight + left turn), it is split into an independent phase for separate control, or a simultaneous release phase (simultaneous release at the south entrance) is inserted into a symmetrical release phase (such as north-south straight and north-south left turn) to increase traffic efficiency.

[0107] When a phase has no passing vehicles or queued vehicles for several consecutive cycles, skip that phase to reduce idle green light periods. Alternatively, when multiple independent phases have few passing vehicles and significant idle periods, the independent release phases can be combined into a composite phase to reduce cycle loss.

[0108] Based on traffic flow, the green light duration for each phase is adaptively adjusted to reduce queuing time in high-flow phases and reduce idle time in low-flow phases.

[0109] Based on reinforcement learning output sorting, high-demand phases (such as tidal flow direction) are automatically adjusted.

[0110] The phase dynamic optimization logic achieves full-process adaptive management of traffic signal control through a closed-loop mechanism of real-time perception, intelligent decision-making, and precise control.

[0111] The benefits of phase dynamic optimization include the following three aspects.

[0112] First, a dual triggering mechanism: integrating regular traffic threshold early warning with emergency event response, a seamless system is built from daily optimization to emergency handling.

[0113] Second, flexible phase control: through a combination of strategies such as phase splitting and merging, duration optimization, and priority reordering, traffic efficiency and resource utilization are dynamically balanced, reducing the green light time allocation error by 25%-40%.

[0114] Third, multi-objective collaborative optimization: by combining reinforcement learning and real-time data fusion technology, while reducing queuing difference (<15%) and reducing green light idling (idling rate <5%), the peak-hour traffic capacity is improved by more than 30%.

[0115] This embodiment of the traffic signal adaptive control method collects multimodal data in real time through a global perception network, transmits it to a cloud-based large model platform via a communication network, models the spatiotemporal features of traffic flow based on a deep prediction model of spatiotemporal Transformer, and generates dynamic signal control strategies by combining reinforcement learning algorithms, which can cope with the dynamic changes of complex traffic scenarios.

[0116] Example 3

[0117] To verify the effectiveness of the technical solution of this invention, a typical urban intersection was selected as an implementation case. This intersection is located in the core area of ​​the city, surrounded by commercial areas, residential areas and schools. The traffic flow is large and complex, with obvious tidal flow phenomena during morning and evening rush hours, and congestion can occur during off-peak hours due to unexpected events.

[0118] Step S1: Construct a vehicle-road-cloud collaborative perception network.

[0119] At this intersection, 100 on-board units (OBUs), 5 roadside sensing units (RSUs), and 8 geomagnetic coils have been deployed. During the morning rush hour, the on-board units use a sampling frequency of 10Hz to acquire real-time data such as vehicle high-precision positioning, instantaneous speed, and turn signal status via V2X communication.

[0120] For example, when a vehicle approaches an intersection, its location data is processed by a Kalman filter algorithm and accurately matched to the left-turn lane at the east entrance, providing precise vehicle dynamic information for subsequent traffic flow analysis.

[0121] The roadside radar-visual fusion equipment analyzes the video stream in real time and calculates parameters such as the number of vehicles passing through each lane and the arrival rate.

[0122] For example, in a certain 5-minute period, the number of vehicles passing through the straight lane at the west entrance was 200, the arrival rate was 40 vehicles per minute, and the queue length reached 50 meters.

[0123] Meanwhile, cloud-based collaborative data collected historical traffic data for the same time period over the past week, as well as real-time weather information (good visibility and no precipitation on that day). These temporal features were mapped into a vector space using the Time2Vec encoder, providing comprehensive data support for subsequent modeling and decision-making.

[0124] By utilizing 5G NR V2X communication technology and combining wavelet transform data compression algorithms, a large amount of raw data collected can be efficiently transmitted to the cloud with a transmission latency of less than 10ms and a data volume compression of more than 80%, ensuring the real-time performance and efficient transmission of data.

[0125] Step S2: Spatiotemporal graph network modeling.

[0126] The intersection is abstracted as a dynamic spatiotemporal graph, with each physical lane as a node.

[0127] Taking the left-turn lane at the east entrance during the morning rush hour as an example, the attributes of this node include information such as current traffic flow (15 vehicles per minute), average speed (15 km / h), and saturation (0.7).

[0128] The edges connect adjacent intersections and upstream and downstream sections of the same road, reflecting the patterns of time evolution.

[0129] The weights of the edges are calculated by summing the weights of the historical turning probability, the phase constraint influence factor, and the real-time queue length influence factor after applying their respective weight coefficients.

[0130] During normal morning rush hour, assuming no unforeseen events occur, the weighting coefficient for the historical turning probability is 0.7, and the historical turning probability shows a 60% probability of turning left from the east to the north exit; currently, the green light for left turns to the north and south is in effect, the weighting coefficient for the phase constraint influence factor is 1, and the phase constraint influence factor is 1; the real-time queue length influence factor is calculated based on the queue length, and the weighting coefficient for the real-time queue length influence factor is 0.6.

[0131] The weight of this edge is calculated and used for subsequent signal control decisions.

[0132] The graph structure is reconstructed every 15 seconds based on real-time traffic condition changes.

[0133] For example, when the queue length of the straight lane at the west entrance suddenly increases, the attributes of the corresponding nodes and edges, as well as the edge weights, will be updated in a timely manner to accurately reflect changes in traffic conditions.

[0134] Step S3: Multi-objective optimization decision.

[0135] In the traffic signal control at this intersection, the state space includes the current phase combination, the number of vehicles in each direction, the saturation of each direction, and the remaining time of the phase. The current phase combination is a green light for north-south straight traffic. The number of vehicles in each direction is counted by roadside sensing devices, such as 30 vehicles waiting at the east entrance and 25 vehicles waiting at the west entrance. The saturation of each direction is calculated based on traffic flow and lane capacity. The remaining time of the phase is updated in real time.

[0136] The action space is a set including the green light duration adjustment, the red light duration adjustment, and the phase priority adjustment coefficient. At a certain moment, the agent finds that there are many vehicles turning left at the east entrance and the saturation is high based on the current state. It decides to increase the green light duration of the left-turn phase at the east entrance by 5 seconds and correspondingly reduce the red light duration by 5 seconds. At the same time, it adjusts the phase priority coefficient to increase the priority of the left-turn phase at the east entrance.

[0137] The reward function optimizes the signal control strategy by balancing traffic efficiency and waiting time uniformity. Within a signal cycle, the number of vehicles passing through each lane during the green light period is counted by roadside sensing devices to calculate traffic efficiency.

[0138] For example, if 150 vehicles pass through the north-south straight-through phase and 80 vehicles pass through the east-west left-turn phase in a certain cycle, and the total signal cycle duration is 120 seconds, then the traffic efficiency is (150+80)÷120≈1.92 vehicles / second.

[0139] The cumulative waiting time of each vehicle is obtained through vehicle-to-vehicle (V2X) communication, and a waiting time balance index is calculated. For example, if the average waiting time across all lanes is 30 seconds, and the maximum waiting time for a particular lane is 45 seconds, then the waiting time balance index is (45-30)÷30=0.5. By continuously adjusting the action space and optimizing the reward function, intelligent control of traffic signals is achieved.

[0140] Step S4, phase dynamic optimization logic.

[0141] In terms of routine optimization, a strategy evaluation is performed after each signal cycle.

[0142] During the evening rush hour, the saturation of the eastbound lane was detected to reach 85%, triggering an adjustment to the green light duration in the signaling scheme.

[0143] Based on real-time traffic conditions, the green light duration for the left-turn phase at the east entrance has been increased from 30 seconds to 35 seconds, effectively reducing vehicle queuing time.

[0144] In terms of emergency response, when the vehicle-mounted V2X sends a request for an ambulance to pass urgently, the system immediately initiates the interrupt service procedure, inserts the priority release phase, maintains the green light in the direction of the ambulance's travel, locks the red light in the conflicting direction, and ensures that the ambulance passes through the intersection quickly, with the response time controlled within 1 second.

[0145] In the dynamic phase sequence adjustment strategy, when the left-turn and straight-through traffic flow at the north entrance surges during the morning rush hour, the original composite phase (straight-through + left-turn at the north entrance) is split into independent phases for separate control. At the same time, a simultaneous release phase at the south entrance is inserted into the symmetrical release phases (straight-through and left-turn at the north and south entrances), which improves the traffic efficiency in this direction by 20%.

[0146] When a phase has no vehicles passing for several consecutive cycles during off-peak hours, that phase is skipped to reduce idle green light time and lower signal cycle loss. The green light duration for each phase is adaptively adjusted based on traffic flow.

[0147] For example, during the morning rush hour tidal flow, the green light duration for inbound phases is increased and the green light duration for outbound phases is decreased, which effectively reduces queuing time for high-flow phases and reduces idle time for low-flow phases, thus reducing the green light time allocation error by about 30%.

[0148] By using reinforcement learning to sort the outputs, the system automatically adjusts the priority of high-demand phases in the tidal flow direction. For example, when there is a large flow of traffic in the inbound direction during the morning rush hour, the priority of the inbound phase is increased, which improves the traffic capacity during peak hours by 35%, reduces the queuing difference to less than 10%, and controls the green light vacancy rate to less than 3%.

Claims

1. A traffic signal adaptive control system, characterized in that, include: The sensing and transmission module collects multimodal traffic flow data at intersections in real time and compresses and transmits it. The spatiotemporal graph module abstracts intersections into dynamic spatiotemporal graphs based on traffic flow data: nodes represent physical lanes, and edges connect upstream and downstream lanes or adjacent time steps of the same lane. The weight of the edges is dynamically calculated and updated in real time based on historical turning probabilities, phase constraint influence factors, and real-time queue length influence factors. The decision module, based on the spatiotemporal diagram, generates an action space containing phase switching timing according to the state space, so as to maximize the number of vehicles passing through per unit time and minimize the difference in maximum waiting time in each direction. The phase dynamic optimization module receives the action space, responds to the conventional optimization trigger conditions, and dynamically adjusts the phase green light duration and switching sequence.

2. The traffic signal adaptive control system according to claim 1, characterized in that, The phase dynamic optimization module includes: responding to emergency event requests, inserting a priority phase and locking the conflict direction red light; and executing at least one dynamic adjustment strategy among phase splitting, merging, skipping, or priority reordering.

3. The traffic signal adaptive control system according to claim 1, characterized in that, The weights of the edges are obtained by summing the historical turning probability, phase constraint influence factor, and real-time queue length influence factor after applying their respective weight coefficients: the weight coefficient of the historical turning probability is dynamically adjusted according to the sudden event; the weight coefficient of the phase constraint influence factor is dynamically adjusted according to the phase conflict matrix; and the weight coefficient of the real-time queue length influence factor is dynamically adjusted according to the current traffic.

4. The traffic signal adaptive control system according to claim 1, characterized in that, The conventional optimization trigger condition is as follows: when the lane saturation is greater than the first threshold or the difference in queue length between the lane and the upstream and downstream lanes is greater than the second threshold, the green light duration is adjusted.

5. A traffic signal adaptive control system according to claim 2, characterized in that, The dynamic adjustment strategy includes: splitting a single-direction flow surge in a composite phase into an independent phase; inserting a co-release phase into a symmetrical release phase; skipping a phase when there are no vehicles passing for multiple consecutive cycles; and merging multiple independent phases into a composite phase when there is a serious lack of traffic.

6. A traffic signal adaptive control system according to claim 2 or 5, characterized in that, The dynamic adjustment strategy includes: when the proportion of flow in the tidal flow direction is greater than the third threshold, increasing the phase priority of that direction and extending the green light duration.

7. A traffic signal adaptive control system according to claim 1, characterized in that, In the spatiotemporal graph module, the attributes of the nodes include current flow, average speed, and saturation. The spatiotemporal graph module reconstructs the graph structure at regular intervals to update the real-time status.

8. A traffic signal adaptive control system according to claim 1, characterized in that, The sensing and transmission module uses a Kalman filter algorithm to correct the drift of the high-precision positioning data collected by the vehicle terminal and matches it with the high-precision map to lane-level accuracy.

9. A traffic signal adaptive control system according to claim 1 or 2, characterized in that, In the decision-making module, the state space includes the current phase combination, the number of vehicles in each direction, the saturation of each direction, and the remaining time of the phase, while the action space includes the green light duration adjustment amount, the red light duration adjustment amount, and the phase priority adjustment coefficient.

10. A traffic signal adaptive control method, based on the traffic signal adaptive control system according to any one of claims 1-9, characterized in that, Includes the following steps: S1 collects vehicle dynamic data, lane-level traffic parameters and historical traffic flow data in real time to build a perception network; S2, construct a dynamic spatiotemporal graph based on the sensing network and the current phase state; S3 generates a signal control strategy based on the state data of the spatiotemporal diagram; S4, execute the strategy and adjust the traffic lights, and feed back the execution results to the perception network for re-collection.

Images