Multi-intersection variable guide lane traffic signal control method based on path selection

Abstract

The invention relates to the technical field of traffic signal control, and discloses a multi-intersection variable guide lane traffic signal control method based on path selection, and the method comprises the steps: firstly constructing a road network model, and initializing parameters; secondly, establishing a bilevel planning model, optimizing a lane function, a signal period and a green time ratio by the upper layer model by taking minimization of the total impedance of a road network as a target, and performing traffic flow distribution on the basis of a Logit model and a road section impedance function by the lower layer model so as to obtain balanced road section flow; then, a hybrid heuristic algorithm is adopted for solving, an outer layer searches for control parameters through a non-dominated sorting genetic algorithm, and an inner layer loads flow through a continuous averaging method; and finally, analyzing the optimal scheme, and driving the variable lane and the annunciator by adopting asynchronous logic after consistency verification. According to the invention, deep cooperation of time and space resources is realized, road network flow distribution is optimized in combination with driver path selection behaviors, and traffic efficiency and system safety are effectively improved.

Description

Technical Field

[0001] This invention relates to the field of traffic signal control technology, specifically a traffic signal control method for multi-intersection variable guidance lanes based on path selection. Background Technology

[0002] Traffic signal control systems play a crucial role in urban traffic operations, and the quality of their control strategies directly determines the travel costs and traffic efficiency of the entire road network. With the increasing demand for urban traffic, tidal traffic and uneven turning flow are becoming more frequent. Variable guidance lanes, as a flexible spatial resource management measure, are widely used to alleviate traffic congestion caused by fluctuations in turning flow.

[0003] However, existing traffic control technologies have limitations in optimizing signal control strategies and allocating the functions of variable guidance lanes. Specifically, existing control methods calculate optimal signal timing or lane attributes based on a fixed traffic flow distribution pattern, or conversely, existing traffic flow distribution models assume that both signal control strategies and lane functional attributes remain unchanged. This static approach, which separates control strategies from traffic flow distribution, ignores the strong coupling between the two in the actual traffic system.

[0004] In actual urban road network operation, drivers following the user equilibrium principle typically choose the route with the lowest cost based on their travel experience. Once lane functions or signal control strategies change within the road network, it inevitably leads to changes in road segment impedance, inducing drivers to alter their route choices and causing a redistribution of traffic flow across the urban road network. Existing technologies, failing to fully consider this dynamic interaction mechanism, struggle to achieve deep coordination of spatiotemporal resources, easily resulting in single-dimensional control schemes that cannot match actual traffic demands, leading to a waste of time and space resources. Furthermore, by ignoring the flow transfer effect caused by changes in control strategies, local optimizations fail to translate into improvements in overall road network efficiency, and may even create new congestion points due to passive flow transfers. This fails to minimize total driver travel time at the regional road network level, making it difficult to meet the demands of modern urban transportation for efficient and balanced operation. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a traffic signal control method for variable guidance lanes at multiple intersections based on path selection. This method solves the problems in existing traffic control technologies where lane function division and signal control strategy optimization are disconnected, and the dynamic impact of changes in control schemes on driver path selection behavior and road network traffic distribution is ignored, resulting in mismatch of road network spatiotemporal resource allocation and low overall traffic efficiency.

[0006] To achieve the above objectives, the present invention provides the following technical solution: This invention provides a traffic signal control method for variable guidance lanes at multiple intersections based on path selection. This method constructs a road network topology model and generates an origin-destination demand matrix and initializes road segment impedance function parameters based on collected traffic data. Then, the road network topology model, origin-destination demand matrix, and road segment impedance functions are input into a pre-constructed two-layer planning model for processing.

[0007] The two-layer planning model is the core decision-making logic of this invention, comprising an upper-layer model and a lower-layer model. The upper-layer model is responsible for formulating specific control strategies, setting lane function variables, signal cycle variables, and green light ratio variables; the lower-layer model is responsible for simulating traffic flow response behavior, receiving the various variables set by the upper-layer model and combining them with the origin-destination demand matrix to allocate traffic flow.

[0008] To solve the aforementioned two-layer programming model, this invention employs a hybrid heuristic algorithm. By searching for combinations of lane function variables, signal cycle variables, and green ratio variables in the upper-layer model, and calculating the corresponding equilibrium segment traffic flow in the lower-layer model, the algorithm iterates repeatedly until a convergent optimal control scheme is obtained. Finally, the system analyzes the optimal control scheme, decoding the numerical vectors contained within it into lane attribute commands and signal timing schemes, which are then sent to the variable lane indicator and signal controller drive equipment, respectively.

[0009] Furthermore, in constructing the bi-level programming model, this invention defines specific objective functions and constraints. The upper-level model uses minimizing the total network impedance as the objective function, which is calculated by accumulating the products of equalization segment traffic flow and segment travel time. Simultaneously, the upper-level model sets multiple constraints, including lane function conservation constraints to ensure a constant total number of lanes at approach points, signal cycle duration constraints setting maximum and minimum signal cycle values, and green light ratio constraints to ensure minimum green light time for each phase. These constraints ensure that the generated control scheme is physically feasible and complies with traffic safety regulations.

[0010] Furthermore, this invention establishes a coupling relationship between the upper-level and lower-level models through road segment capacity parameters. The road segment capacity is defined as a function of lane function variables and green light ratio variables. Its mechanism is as follows: when the upper-level model adjusts lane function variables (such as changing the number of left-turn or straight lanes) or green light ratio variables (adjusting the green light time ratio), it directly changes the road segment capacity parameters, thereby altering the road segment travel time calculated by the road segment impedance function. The road segment impedance function parameters include intersection delay, which is jointly determined by lane function variables, signal cycle variables, and green light ratio variables. This coupling mechanism accurately reflects the direct impact of spatiotemporal resource adjustments on road segment impedance.

[0011] Furthermore, in the lower-level model, this invention utilizes a Logit model to describe the probability of drivers choosing different paths, thus reflecting the randomness of drivers' perceived path impedance. Under the premise of satisfying flow conservation and non-negativity constraints, the lower-level model allocates the traffic demand in the origin-destination demand matrix to the effective paths of the road network until a stochastic user equilibrium state is reached, i.e., all drivers cannot reduce perceived impedance by changing paths. The flow distribution obtained at this point is the equilibrium road segment flow.

[0012] Furthermore, to address the solution complexity of the aforementioned model, the hybrid heuristic algorithm is designed as a nested outer loop and inner loop structure. The outer loop primarily handles the optimization problem involving a mixture of discrete and continuous variables. It utilizes a non-dominated sorting genetic algorithm to process the upper-level model, encoding lane function variables as integer sequences and signal periodicity and green ratio variables as real number sequences, generating a candidate solution set that is then passed to the inner loop. The inner loop focuses on the balanced calculation of traffic flow assignment. For each candidate solution set passed from the outer loop, it uses a continuous averaging method to solve the lower-level model. The inner loop iteratively loads auxiliary traffic flow to update the equilibrium segment traffic flow, and at the end of each iteration, it calculates the average traffic flow using a sliding window until the relative error of the average traffic flow is less than a preset convergence threshold. After receiving the equilibrium segment traffic flow returned by the inner loop, the outer loop substitutes it into the objective function of the upper-level model to calculate the fitness value, and performs selection, crossover, and mutation operations on the candidate solution set based on the fitness value to generate the next generation of candidate solution sets, until the iteration termination condition is met and the optimal control scheme is output.

[0013] Furthermore, to ensure the effective implementation of the control scheme, this invention incorporates a specific decoding and verification mechanism. The parsing process includes: mapping lane function variables in the optimal control scheme to a guide arrow bitmap based on a pre-defined lane function mode library, generating lane attribute instructions; rounding the signal period variable in the optimal control scheme, calculating the green light start and end times for each phase based on the green light ratio variable, and generating a signal timing scheme. After generating the instructions, a consistency check is performed to determine whether the green light time for each phase in the signal timing scheme covers the enabled guide lane direction in the lane attribute instructions. If a lane function is enabled but the corresponding phase's green light time is missing, the system will intercept the lane attribute instructions and revert to the default safety scheme to prevent traffic conflicts.

[0014] Furthermore, this invention employs asynchronous execution logic in the device driving stage to address safety hazards during variable lane switching. Specifically, the variable lane indicator device is prioritized to display the new directional arrow and enter a warning buffer period. During this buffer period, the signal controller maintains the original timing scheme to clear vehicles queuing according to the old lane function. After confirming the queue is cleared, the signal controller then loads and executes the new signal timing scheme, thereby achieving a smooth transition in the switching of spatiotemporal resources.

[0015] This invention provides a traffic signal control method for variable guidance lanes at multiple intersections based on path selection. It has the following advantages: 1. This invention achieves deep collaborative optimization of spatiotemporal resources in traffic control. By jointly solving the lane function division of variable directional lanes at multiple intersections with the signal control scheme, it utilizes variable lanes to adjust the spatial capacity of road segments and uses signal timing to adjust time right-of-way, eliminating the resource waste caused by single-dimensional control, ensuring accurate matching between intersection capacity and traffic demand in time and space, and improving the traffic efficiency of nodes and road segments.

[0016] 2. This invention introduces a traffic flow dynamic redistribution mechanism based on path impedance. During the optimization process, the system calculates the path impedance between all origin-end point pairs in the regional road network based on the alternative lane and signal schemes, and simulates the driver's path selection behavior accordingly. This control strategy based on traffic flow distribution feedback can predict the flow shift caused by changes in the control scheme, thereby formulating a scheme that can actively guide the balanced distribution of flow and prevent the shift of road network congestion caused by local control adjustments.

[0017] 3. This invention takes minimizing the total travel time of all drivers in the road network as the global optimization objective. By constructing a two-level programming model to process the road network topology and traffic data, it completes the optimal allocation of traffic flow while searching for the optimal control parameters. Compared with single-point or single-line control, this method can reduce the total impedance of the system at the regional road network level, reduce the average travel time and delay of vehicles in the road network, and improve the operation and service level of the entire transportation network. Attached Figure Description

[0018] Figure 1 A flowchart of a multi-intersection variable guidance lane traffic signal control method based on path selection provided in an embodiment of the present invention; Figure 2 This is a structural block diagram of a multi-intersection variable guidance lane traffic signal control system based on path selection provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the interaction logic between the upper and lower layers in the two-layer planning model provided in this embodiment of the invention; Figure 4 This is a flowchart of the hybrid heuristic algorithm for solving the problem provided in an embodiment of the present invention; Figure 5 The flowchart for decoding and security verification of the optimal control scheme provided in this embodiment of the invention is as follows; Figure 6 The convergence curve of the hybrid heuristic algorithm for iterative optimization provided in the embodiments of the present invention is shown. Figure 7 A comparison chart of the operational indicators of the collaborative control method and the single control method provided in the embodiments of the present invention; Figure 8 The simulation diagram shows the dynamic response and stability of the system under a sudden traffic disturbance scenario provided in the embodiments of the present invention. Detailed Implementation

[0019] The technical solutions in 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.

[0020] See attached document Figure 2 This invention provides a traffic signal control method for variable guidance lanes at multiple intersections based on path selection. The method includes a traffic signal control system, which mainly consists of a traffic data acquisition unit, a data transmission unit, a collaborative control calculation unit, and a roadside execution unit.

[0021] Traffic data acquisition units are deployed at the entrances of various intersections and key road sections of the target road network. These units consist of traffic flow detection equipment configured to collect real-time traffic operation status data of the road network. This traffic operation status data includes, but is not limited to, traffic volume, average vehicle speed, lane occupancy, and queue length at each entrance. Regarding the specific selection of traffic flow detection equipment, those skilled in the art can choose devices such as microwave radar, video detectors, or geomagnetic coils based on actual road conditions. Their installation and debugging are well-known technologies in the field and will not be elaborated upon here.

[0022] The data transmission unit connects to the traffic data acquisition unit, the collaborative control computing unit, and the roadside execution unit via a fiber optic network or a wireless communication network. The data transmission unit is configured to upload the collected traffic operation status data to the collaborative control computing unit and to send the control commands generated by the collaborative control computing unit to the roadside execution unit.

[0023] The cooperative control computing unit is equipped with memory and a processor. The memory stores computer programs and road network topology data, and the processor executes the computer programs to achieve coordinated optimization of lane functions and signal timing. The cooperative control computing unit receives real-time data from the traffic data acquisition unit and runs a cooperative optimization algorithm based on two-level programming based on this data.

[0024] In the collaborative control computing unit, a directed graph of the road network is pre-built. .in, This represents the set of nodes in the road network, corresponding to intersections in the actual traffic road network; This represents the set of road segments in the road network, corresponding to the roads connecting intersections. For any logical road segment in the road network... The collaborative control computing unit defines the traffic impedance properties associated with it.

[0025] To achieve coordinated control, the coordinated control computing unit defines three types of key control variables: lane function variables. Signal periodic variables and the green credit ratio variable Among them, lane function variables This variable characterizes the functional attributes of variable-direction lanes at intersection approach lanes. Its value directly determines the number of available lanes and corresponding physical capacity for each flow direction (e.g., left turn, straight, right turn) at the approach lanes. Signal cycle variable. The duration of a complete cycle for an intersection signal control scheme. Green ratio variable. Characterizing the effective green light time and signal periodicity variables for a specific phase The ratio of .

[0026] The collaborative control computational unit is configured to construct and solve a two-level programming model. In the two-level programming model, lane function variables... Signal periodic variables Compared with green credit variables These three variables are used as interconnected decision variables. The collaborative control computing unit sets the algorithm logic so that they work together to calculate intersection delays, thereby affecting the total impedance of the road segment. Changes in road segment impedance are further used to simulate the path selection behavior of traffic flow in the road network.

[0027] Specifically, the upper-level model logic configuration in the collaborative control computing unit is to search for the optimal [model] with the goal of minimizing the total impedance of the road network. , and The upper-level model uses a combination of parameters; the lower-level model is logically configured based on the principle of random user equilibrium, allocating traffic flow according to the impedance parameters set by the upper-level model. The upper and lower-level models interact and iterate through two parameters, road segment impedance and road segment flow, until a convergent control scheme is obtained.

[0028] The roadside execution unit includes a signal controller and a variable lane indicator. The signal controller is connected to the traffic lights at the intersection and is used to control the switching of red and green lights. The variable lane indicator is installed above the guide lanes at the intersection approach and is used to display the lane's guidance function.

[0029] The roadside execution unit receives the optimal control parameter vector output from the cooperative control calculation unit. When a new lane function variable is received... When a command is received, the variable lane indicator switches the displayed directional arrows, thereby changing the traffic function of the physical lane; when a new signal cycle variable is received... Compared with green credit variables When given a command, the signal controller adjusts the current timing scheme. Through the coordinated execution of the aforementioned hardware, the system achieves synchronous and dynamic adjustment of road network spatial resources (lanes) and temporal resources (signals).

[0030] See attached document Figure 1 This invention provides a traffic signal control method for variable guidance lanes at multiple intersections based on path selection. The method consists of four steps: digital modeling, construction of a two-level programming model, solution of a hybrid algorithm, and strategy execution, to achieve deep collaborative optimization of the spatiotemporal resources of the traffic network.

[0031] Step S1: Construct a road network topology model and initialize traffic parameters.

[0032] First, the physical transportation network is abstracted as a set of nodes. and road segment collection The directed graph formed In this system, nodes represent intersections, and road segments represent roads connecting intersections. The system reads the basic geometric data of the road network, including road segment lengths, design speeds, baseline values ​​for the number of lanes, and the phase configuration structure of each intersection. Based on this, it initializes traffic demand data and impedance function parameters. It then reads the current origin-destination (OD) flow matrix. This matrix reflects all starting points in the road network. To the finish line The total travel demand is calculated. Simultaneously, the free-flow time, traffic capacity baseline value, and congestion coefficient are set in the road segment impedance function (such as the BPR function). For the specific parameter calibration in the impedance function, those skilled in the art can determine it using field surveys or least squares regression. The parameter calibration process is well-known in the field and will not be elaborated upon here.

[0033] Step S2: Construct a two-layer planning model that coordinates lane function and signal control.

[0034] In this step, a mathematical model is established to describe the interaction between managerial decisions and traveler behavior. The upper-level model is defined as the road network system performance optimization layer. The decision variables of the upper-level model are set as lane function variables. Signal periodic variables and the green credit ratio variable The objective function is set to minimize the total impedance of the road network, which is the sum of the products of traffic flow and travel time on all road segments. In the upper-level model, constraints are set to ensure physical feasibility, including: lane function conservation constraints (ensuring that the total number of lanes at the approach remains unchanged), signal cycle duration constraints (setting the maximum and minimum range of the cycle), and green light ratio constraints (ensuring that the minimum green light time for each phase meets the pedestrian crossing safety requirements).

[0035] The lower-level model is defined as the user path selection equalization layer. It receives lane and signal schemes determined by the upper-level model, which alter road segment capacity and intersection delays, thereby changing road segment impedance. The lower-level model employs a stochastic user equalization principle, using a Logit model to describe the probability of drivers choosing different paths. Network traffic reaches equilibrium when all drivers cannot reduce their perceived impedance by unilaterally changing their paths.

[0036] The key to this step lies in establishing the coupling relationship between the upper and lower level models: a collaborative control scheme. The change directly modifies the parameters of the road segment impedance function in the lower-level model; the road segment flow after the lower-level model is balanced. Conversely, these parameters are used as input parameters to calculate the objective function value of the upper-level model.

[0037] Step S3: Solve the model using a hybrid algorithm based on non-dominated sorting genetic algorithm and continuous averaging method.

[0038] Since the aforementioned bilevel programming model is a non-convex, nonlinear, NP-hard problem, this step employs a nested hybrid heuristic algorithm for solution. In the outer loop, a non-dominated sorting genetic algorithm (NSGA-II) with an elitist strategy is used to optimize the upper-level model. Lane function variables are then considered. Signal periodic variables Compared with green credit variables The algorithm encodes candidate solutions. New candidate solutions (i.e., a set of control schemes) are generated using selection, crossover, and mutation operators. In the inner loop, for each candidate solution (i.e., a specific control scheme) generated by the outer loop, the Continuous Average Method (MSA) is called to solve the lower-level model. The MSA algorithm iterates multiple times, updating the traffic distribution in the road network based on the current impedance, until the stochastic user equilibrium (SUE) segment traffic under that control scheme is obtained. The segment traffic calculated in the inner loop is substituted into the upper-level objective function to calculate the fitness value of the candidate solution. The algorithm repeats the above inner and outer loops until the preset number of iterations or convergence accuracy is met, finally outputting the Pareto optimal solution set or the weighted optimal solution.

[0039] Step S4: Decoding the control scheme and execution by roadside equipment.

[0040] The optimal solution vector output in step S3 is decoded into specific engineering parameters. For lane function variables... This is mapped to specific lane attribute commands (e.g., defining the second inner lane of the north entrance as "left turn") and sent to the corresponding variable lane indicator device, driving the LED screen to display the corresponding directional arrows. For signal periodic variables... Compared with green credit variables This is then converted into a specific signal timing scheme (including the start and end times of the green light for each phase) and sent to the signal controller for execution. The system continuously monitors the road network status, and when it detects that the change in the OD demand matrix exceeds a preset threshold, steps S1 to S4 are retried to achieve dynamic closed-loop control.

[0041] In order to digitally describe the traffic network and support subsequent two-level planning calculations, this embodiment first establishes a topological model of the road network.

[0042] The collaborative control computing unit abstracts the actual traffic network into a directed graph. .in, Represents a set of nodes, each node This corresponds to a controlled intersection or traffic flow entrance / exit in the road network; Represents a set of road segments, each road segment A one-way roadway that connects two nodes.

[0043] For each road segment The system defines its inherent physical attribute parameters, including road segment length. Free flow velocity And the baseline configuration for the number of lanes. Specifically, to provide a more refined description of the turning behavior at intersections, the road segment set... The road segments in the code not only include the road segments between intersections, but also logically include the turning lanes of the intersection entrances.

[0044] Based on this topology, the system defines the origin-destination (OD) demand matrix. Let... The set of originating points in the road network. This is the set of destinations for each trip. (in , ), Indicates starting from the origin To the finish line The total travel demand. For obtaining the OD matrix, those skilled in the art can use checkpoint identification data to infer its origin or extract it from mobile phone signaling data. The specific matrix generation process is well-known in the field and will not be elaborated here.

[0045] Based on directed graphs Total travel demand The system defines the traffic conversion relationship between paths and road segments, which is the structural basis for subsequent traffic allocation calculations.

[0046] set up For OD The set of all valid paths between them. Indicates OD pair Between Traffic flow on the route (of which) To establish the mapping between path flow and segment flow, a correlation variable is defined. Its value selection rules are as follows: If OD is... The The route passes through the following sections ,but ;otherwise .

[0047] Based on the above definition, road segment Traffic flow on the road section The calculation formula is: ; Here, represents the cumulative summation symbol.

[0048] This formula clarifies that the traffic flow of a road segment is the sum of the traffic flows of all paths through all OD pairs passing through that road segment.

[0049] Simultaneously, the flow conservation constraint must be satisfied in the topology network, meaning the sum of the flow along all paths between any two OD pairs must equal the total travel demand for that OD pair. Its mathematical expression is: ; And it satisfies the nonnegativity constraint: ; After defining the static topology and traffic flow relationships, the system parameterizes the dynamic attributes of road segments. Unlike traditional static road networks, in this embodiment, road segments... Traffic capacity It is not a fixed value, but a function defined as a control variable. Specifically, for the road segment at the intersection approach, its capacity... Associated with lane function variables Compared with green credit variables When the collaborative control computing unit adjusts (For example, increasing the number of left-turn lanes) or adjusting When the green light time is increased, the traffic capacity of this road segment in the topology model. This will be dynamically updated, thus affecting the impedance calculation of that road segment. This dynamic mapping mechanism is pre-set in the digital modeling stage to support the dynamic evolution of supply and demand relationships in the subsequent two-level planning model.

[0050] To achieve synchronous optimization of lane and signal resources in the mathematical model, the cooperative control computing unit parameterizes the control elements of the physical intersection into three core decision variables: lane function variables. Signal periodic variables and the green credit ratio variable .

[0051] Lane function variables This is used to characterize the physical lane division scheme for intersection approach lanes. For each controlled approach lane, the system predefines its total number of lanes and the location of reversible lanes. Lane function variables. Defined as a discrete state vector, each value corresponds to a specific lane function combination (e.g., left turn, left turn, straight, straight / right). In the road network topology model, different... The value directly changes the saturation flow rate in the corresponding flow direction. For example, when the lane function variable... When an additional left-turn lane is added, the physical capacity of the left-turn flow increases accordingly.

[0052] Signal periodicity Defined as the total time required for a traffic signal controller to complete a full phase sequence, measured in seconds. Green ratio variable. Defined as the effective green light time and signal period variable for a specific phase or flow direction. The ratio of green light ratio to time loss ratio. For each signal-controlled intersection in the road network, the sum of the green light ratios of all phases and the sum of the time loss ratios should equal 1.

[0053] In the collaborative control logic of this invention, the three variables mentioned above do not act independently, but rather jointly determine the traffic capacity of a road segment. Coupled with intersection delays.

[0054] Specifically, for a road segment in the road network that represents a specific flow direction Its traffic capacity It is no longer a static constant, but a function of the aforementioned control variables. (Passage capacity) With saturation flow rate (by (determined) and green ratio variable Proportional. When the collaborative control computing unit adjusts or At that time, the road section Traffic capacity Changes occur, which in turn alter the vehicle's driving behavior on that section of road.

[0055] This influence is reflected through the road segment travel time function (BPR function). The system uses the following formula to calculate the road segment travel time excluding intersection delays. : ; in, For road section Free-flow time; Traffic flow on the road segment; and The congestion coefficient (in this embodiment, the value is taken as...) , As can be seen from this formula, the lane function variable... Compared with green credit variables By changing This directly affects the travel time on the road section. Size.

[0056] In addition, these three variables also directly participate as independent variables in intersection delays. The calculation. In the objective function constructed subsequently, the intersection delay is explicitly expressed as... This means that even if the traffic flow on the road segment... Keep it unchanged, only change the lane function variable. Signal periodic variables or green credit ratio variable Any one or more of these parameters will directly cause a non-linear change in the calculated delay value. Through this parameterized definition, this invention constructs a dynamic correlation mechanism between lanes, signals, and traffic flow at the mathematical level.

[0057] After constructing the road network topology and defining the collaborative control variables, the collaborative control computing unit further initializes and configures the constant parameters and basic boundary conditions required by the two-level planning model.

[0058] The system loads origin-destination (OD) traffic demand matrix data. Let... The OD demand matrix has the following elements. Represents starting point To the finish line The total travel demand. This matrix serves as the input benchmark for subsequent traffic allocation in lower-level models. For The numerical values ​​can be obtained by those skilled in the art through historical traffic survey data or by feedback based on real-time checkpoint flow data. The specific data cleaning and matrix estimation process is a well-known technology in the field and will not be described in detail here.

[0059] The system calibrates key coefficients in the road segment impedance function. In this embodiment, the BPR function is used to describe the nonlinear relationship between road segment travel time and traffic flow. Initialization parameters include free-flow time. Congestion coefficient and Free-flow time Based on the physical length of the road segment With the design speed of the road section The calculation shows that, ; Congestion coefficient and This determines the steepness of the impedance function curve. According to a specific embodiment of the present invention, to conform to the traffic flow characteristics of urban roads, these two parameters are initialized to... and Based on the above initialization parameters, the road segment travel time function is specified as follows: ; in, For road section Free-flow time; Traffic flow on the road segment This refers to the traffic capacity of the road segment. It should be noted that, although... , and It is set as a constant during the initialization phase, but the denominator in the formula In subsequent optimization processes, it will be treated as a controlled variable (a lane function variable). Compared with green credit variables (adjustment), thereby making the calculated It can dynamically reflect the effectiveness of collaborative control strategies.

[0060] In addition, the system needs to initialize the boundary parameters of the constraints, specifically including signal control constraints and lane geometry constraints. For signal control constraints, the system sets the minimum permissible signal cycle. and maximum signal period In this embodiment, considering both the driver's waiting tolerance and the intersection's traffic efficiency, the following is determined: Set to 60 seconds. Set to 180 seconds. Simultaneously, set the minimum green light ratio for each phase. This value is primarily set based on pedestrian safety requirements, ensuring that pedestrians have sufficient time to cross the crosswalk within a given cycle. In this embodiment, the value is set as follows: That is, the effective green time for any phase must not be less than 15% of the cycle.

[0061] For lane geometry constraints, the system reads the total number of lanes for each approach lane. The total number of lanes Depending on the actual physical construction conditions of the intersection, for example, at a typical urban arterial road intersection, this value is usually between 3 and 5 (roads). The system relies on the read data... Values ​​to define lane function variables This dimension ensures that the lane function schemes generated subsequently will not exceed the actual physical red line range of the road.

[0062] See attached document Figure 3 This embodiment defines a unified road segment impedance function to deeply couple the allocation of resources in physical space (lane function) with the allocation of resources in the time dimension (signal control) at the mathematical level.

[0063] The collaborative control computing unit guides and regulates traffic flow based on road segment travel time (i.e., impedance). For any road segment in the road network... The system defines its segment impedance. It consists of three parts: road segment travel time, intersection delay, and intersection clearance time. The formula for calculating this impedance is defined as: ; in, The travel time on a section of road with flowing water. Delays at the intersection; The travel time for a vehicle to cross the physical area of ​​the intersection; For lane function variables; The green credit ratio is the variable; For signal periodicity variables; The physical length of the road segment; This represents the average speed of the vehicle as it crosses the intersection.

[0064] In this formula, It is the core coupling node for achieving coordinated lane and signal control. This indicates that intersection delays are not solely dependent on traffic flow. Instead, it is determined by lane function variables. Green credit ratio variable and signal periodic variables A multivariate function jointly determined.

[0065] Specifically, lane function variables The lane function variable determines the spatial capacity of the approach lane. When the status changes (e.g., converting a straight lane into a left-turn lane), the saturation flow rate in the corresponding direction changes abruptly. However, spatial capacity alone is insufficient to determine actual traffic efficiency; time resources must also be considered. (Green ratio variable) This determines the effective utilization rate of the space's accessibility over time. Signal periodicity variable. This determines the frequency of right-of-way switching and the phase difference between vehicle arrival and traffic light display.

[0066] The collaborative control computing unit established the following collaborative mechanism using the above formula: First, spatial resources constrain time resources. When lane function variables... When the number of lanes is relatively small, the saturation flow rate in that direction is low. In this case, if the traffic flow in the road segment... The value is relatively large, so a larger green credit ratio variable must be configured. Only then can Keep it within a reasonable range; conversely, if adjustments are made... Increasing the number of lanes allows for a suitable reduction in the green light ratio variable for that phase, while maintaining the same level of delay. The saved time and resources will be allocated to other congested phases.

[0067] Second, the compensation of time resources for space resources. When physical road space is limited and adjusting lane function variables is not feasible... When the number of lanes is increased, the cooperative control calculation unit increases the signal period variable. or green credit ratio variable This increases the time right-of-way in that direction, thereby mathematically reducing intersection delays. The value.

[0068] In addition, the first term in the formula This also implies the impact of collaborative control. As mentioned before, The BPR function is used for calculation, and its denominator is the traffic capacity of the road segment. In this embodiment, Lane function variables (Determining the number of lanes) and the green ratio variable The product function (which determines the reduction factor). Therefore, for and Any adjustment will simultaneously change the first item. Second item The numerical value was thus used to precisely quantify the effect of the control strategy on the road segment impedance. The combined impact.

[0069] Based on the above mechanism, the model of this invention can quantitatively evaluate the combined efficiency of spatial resources (lanes) and temporal resources (signals), ensuring the efficiency of lane function variables. With signal control variables At current traffic To achieve optimal matching under load, avoid increased impedance caused by mismatch between the number of lanes and the green ratio (such as resource idleness caused by "multiple lanes and low green ratio" or traffic bottleneck caused by "few lanes and high green ratio").

[0070] In the two-layer planning architecture, the lower-layer model is configured to simulate the route selection behavior of travelers in the traffic network. Its core logic is to calculate the equilibrium traffic flow distributed across each road segment under given road network supply conditions (i.e., lane functions, signal cycles, and green light ratios determined by the upper layer).

[0071] This embodiment uses the Stochastic User Equilibrium (SUE) model as the lower-level model. This model is based on the following behavioral assumptions: drivers strive to minimize their perceived travel impedance when choosing a route, and due to the incompleteness of drivers' knowledge of road network information, their perceived impedance is a random variable.

[0072] The lower-level model is defined as a function of road segment flow, and its mathematical expression is a convex programming problem. To construct this model, the cooperative control computational unit first defines the generalized path impedance.

[0073] For any origin-end point OD pair Between, in the The first iteration calculation A path, its path impedance The path impedance is defined as the sum of the impedances of all constituent segments along the path. The system calculates the path impedance using the following formula: ; in, Indicates the first In the next iteration step, the road segment Traffic flow on the affected road sections; This indicates the actual impedance of the road segment under this flow rate (including the travel time of the road segment and the delay at the intersection). For path and road segment association variables, when road segment Belongs to OD pair The The value is 1 if there is a path, otherwise it is 0; Represents a set of road segments Each section of the road Perform a traversal and summation.

[0074] Based on path impedance The collaborative control computational unit uses the Logit discrete choice model to calculate the path selection probability. For OD pairs The driver selected the first Path selection probability of a path The calculation formula is: ; in, Indicates the starting point of the connection. and the end point The set of all valid paths; For path indices in the collection; This is a discrete parameter used to measure the driver's sensitivity to impedance differences. In this embodiment, The value range is set to 0.1 to 1.0. As a preferred implementation, the value is set to... ,when When the value is large, it indicates that drivers are very sensitive to time costs, and traffic will be highly concentrated on the shortest path; Represented by natural constant Exponential functions with base 0; This indicates an inverse relationship where the greater the impedance, the smaller the utility (the lower the probability of selection); This represents the sum of the utility values ​​of all valid paths between the OD pair, and serves as a normalization factor to ensure that the sum of the probabilities of choosing all paths equals 1.

[0075] The lower-level model strictly adheres to flow conservation and non-negativity constraints during the calculation process.

[0076] The flow conservation constraint requires the total travel traffic demand between each pair of origin and destination (OD) to be... It must be fully allocated to all its valid paths, which can be mathematically described as follows: ; in, Indicates OD pair Between Traffic flow on the route, and satisfying ; Indicates OD pair The set of all valid paths between them; Indicates the relationship between OD and Iterate through and sum all valid paths between them.

[0077] The non-negativity constraint requires that the flow of any path must not be negative: ; Ultimately, the traffic flow on the road segment The following is obtained by superimposing the traffic flows of all paths passing through this road segment: ; in, This indicates that the system needs to perform multiple traversals and accumulations, sequentially traversing the set of all travel origins in the road network. The collection of all travel destinations and the set of all valid paths under each OD pair. .

[0078] By constructing the aforementioned lower-level model, the system establishes a closed-loop mapping relationship: the control parameters determined at the upper level are implicitly contained within... The function structure thus determines the path impedance. and path selection probability This ultimately leads to increased traffic flow on road sections within the road network. The redistribution of .

[0079] In the two-level planning architecture, the goal of the upper-level model is to minimize the operating cost of the entire transportation network from the perspective of global optimization by adjusting collaborative control variables.

[0080] This embodiment constructs the upper-level model as a nonlinear optimization problem. Unlike the lower-level model, which simulates individual drivers maximizing their own interests (i.e., minimizing their personal travel time), the upper-level model aims to find an optimal combination of control parameters, i.e., lane function variables. Green credit ratio variable and signal periodic variables This minimizes the total impedance of the road network.

[0081] The objective function of the upper-level model is defined as minimizing the total system travel time. The cooperative control computational unit describes this objective function using the following formula: ; in, This represents the operation of finding the minimum value of the objective function; Indicates the total system travel time; Represents a set of road segments Each section of the road Perform a traversal and summation; Indicates road segment Regarding the traffic flow on the affected road sections, it is particularly important to emphasize that... These are not variables directly independent of the upper-level model, but rather equilibrium solutions calculated by the lower-level model (SUE model) based on the control scheme given by the upper-level model. Implicitly dependent on ; Indicates road segment The impedance per unit travel time, which varies with the flow rate of the road segment. Changes are more directly controlled by lane function variables. (Determines the physical capacity of the road segment) and signal parameters (Determines the capacity of road sections during specific time periods); Indicates road segment Total time consumed (data volume multiplied by unit time).

[0082] To ensure that the generated control scheme is engineering-feasible and complies with traffic regulations, the upper-level model must meet a series of physical and operational constraints.

[0083] First, there are constraints on signal control parameters. This applies to signal periodic variables. The set duration boundary must be met: ; in, and These represent the minimum allowable value (e.g., 60 seconds) and the maximum allowable value (e.g., 180 seconds) of the signal period, respectively. For the green signal ratio variable... The minimum green light time requirement and the conservation relationship within the cycle must be satisfied: ; ; in, Indicates an intersection The The green light ratio of the phase; This represents the minimum green light ratio threshold required to ensure safe pedestrian crossing and minimal motor vehicle traffic. It typically ranges from 0.10 to 0.25, with the specific value depending on the physical scale of the intersection and pedestrian walking speed. Indicates an intersection The percentage of total lost time within a cycle. The formula means that the sum of the effective green light ratio and the percentage of lost time for all phases must be strictly equal to 1.

[0084] Secondly, there are lane function geometric constraints. Lane function variables. The adjustments cannot change the total physical width of the road; that is, the sum of the number of functional lanes of each approach road must equal the total number of physical construction lanes of that approach road. ; in, Indicates lane function variables The number of lanes for a specific function (such as left turn or straight ahead); This represents the total physical number of lanes at this entrance. Furthermore, Integer values ​​must be used, and the lane layout must conform to topological continuity (for example, left-turn lanes are usually located on the innermost side).

[0085] Finally, there is the two-layer coupling constraint. The solution of the upper-layer model is strictly constrained by the equilibrium behavior of the lower-layer users, i.e., the objective function... It must be the lower-level model with the current control parameters. The following equilibrium solution: ; This constraint clarifies the master-slave game relationship in the two-level planning: the upper-level manager formulates the strategy first. Lower-level travelers respond by making route choices based on the road network conditions generated by this strategy. Then, upper management evaluates the effectiveness of the strategy based on this response.

[0086] See attached document Figure 4 Given that the lane and signal coordinated bi-level planning model constructed in this invention has nonlinear, nonconvex, and mixed integer characteristics, conventional gradient descent methods are difficult to solve. To address this, this embodiment designs a nested hybrid heuristic algorithm architecture, using an improved non-dominated sorting genetic algorithm (NSGA-II) as the upper-level master solver and embedding the continuous average method (MSA) as the lower-level sub-solver. The two form a closed loop through parameter transfer and fitness feedback.

[0087] The hybrid algorithm architecture mainly consists of two core modules: an upper-level evolutionary search loop and a lower-level traffic allocation loop.

[0088] In the upper-level evolutionary search loop, the cooperative control computational unit utilizes the powerful global search capability of the NSGA-II algorithm to handle the combinatorial optimization problem of the control variables. The system first performs hybrid encoding on the cooperative control variables. Due to the lane function variables... For discrete integers, while the signal periodic variable... Compared with green credit variables For consecutive real numbers, this embodiment employs a hybrid chromosome strategy combining real number encoding and integer encoding. The algorithm initializes and generates chromosomes containing... The candidate solution set of entities, where each entity mathematically represents a potential and complete set of cooperative control schemes. .

[0089] In each iteration of the algorithm, accurately evaluating the merits of each control scheme in the candidate solution set is crucial for guiding the search direction. However, the total system travel time... The calculation depends on the traffic flow of the road segment. ,and It is not a known quantity; it is determined by the traveler's response to the current control scheme.

[0090] This introduces a lower-level traffic assignment cycle. For each candidate solution in the upper-level candidate solution set (i.e., each determined set of...) The system passes this as an input parameter to the lower-level model. At this point, the supply state of the road network is fixed, and the road segment resistance function... The parameters were determined. The Cooperative Control Computing Unit then initiated the MSA algorithm to simulate the driver's route selection behavior. Through continuous iterative loading, the MSA algorithm gradually brought the road network traffic distribution closer to the Stochastic User Equilibrium (SUE) state.

[0091] Once the MSA algorithm meets the convergence criterion, it outputs the segment flow rate under this control scheme. At this point, the system will Feedback is sent back to the upper-level model and combined with the current collaborative control scheme. Substitute into the formula of the upper objective function: ; Calculated The value is the objective function value for that individual. Subsequently, the NSGA-II algorithm performs optimization and iterative operations on the current candidate solution set. The system first considers the... The system performs fast non-dominated sorting and crowding distance calculation, using an elitist retention strategy to select high-quality solutions with excellent performance and uniform distribution. Subsequently, by applying operators such as selection, crossover (recombination), and mutation (perturbation), the system performs genetic operations on the retained high-quality solutions, thereby deriving and generating the next generation of candidate control schemes. This optimization closed loop of evaluation, selection, and derivation will be continuously executed until the preset maximum number of iterations or solution set convergence threshold is reached, ultimately outputting the optimal cooperative control scheme with the lowest total system impedance.

[0092] For the conventional genetic operator operations in the NSGA-II algorithm (such as simulated binary crossover, polynomial mutation, etc.) and the basic iterative principles of the MSA algorithm, those skilled in the art can refer to relevant classic algorithm literature for implementation. These are well-known technologies in the fields of evolutionary computation and traffic engineering, and will not be elaborated here.

[0093] The inner loop, as a sub-solver of the hybrid algorithm, has the core task of solving the numerical solution of the lower-level stochastic user equilibrium (SUE) model under the premise that the upper-level cooperative control scheme (lane function variables, signal period variables, and green ratio variables) remains unchanged, that is, determining the traffic distribution state of the road network under specific supply conditions.

[0094] This embodiment uses the continuous average method (MSA) for iterative solution. This algorithm can effectively overcome the flow oscillation problem caused by Logit probability calculation in the SUE model, and ensure the convergence and stability of the solution.

[0095] The specific execution steps of the inner loop are as follows: First, perform initialization settings. The system sets the number of iterations for the inner layer. and the initial traffic flow of all road segments in the road network. Set to zero (or initial allocation flow based on free-flow time).

[0096] Next, the iterative update process begins. In the... In this iteration, the collaborative control computing unit first bases its calculations on the current road segment flow. Update the segment impedance. At this point, use the impedance function. Calculate the actual travel time for each road segment. It is important to note that during this process, , and For fixed constants passed in from the upper layer, only It is a variable.

[0097] Subsequently, random loading is performed to obtain auxiliary traffic. Based on the updated segment impedance, the system uses the constructed Logit model to calculate the selection probability of each path. and the total travel demand This traffic flow is allocated to the road network, resulting in a new set of road segment traffic distributions. This traffic flow is called auxiliary traffic flow, denoted as... For auxiliary traffic The specific calculation process involves shortest path search and probability allocation. Those skilled in the art can implement it using Dijkstra's algorithm or Dial's algorithm. The specific programming implementation is a well-known technology in this field and will not be elaborated here.

[0098] Obtain auxiliary traffic Then, the system updates the road segment flow rate using the core iterative formula of the continuous averaging method. This embodiment uses the following formula to calculate the... Traffic flow in the next iteration : ; in, This represents the updated road segment used in the next iteration of the calculation. Traffic flow on the road section; Indicates the current number The road segment in the next iteration Traffic flow on the road section; This represents the auxiliary flow rate calculated based on the current impedance, which represents the driver's full responsiveness to the current road network conditions. Indicates the index of the current iteration count ( ); This represents the step size sequence of the MSA algorithm.

[0099] However, considering the randomness of the search trajectory in the MSA algorithm, directly using the first... Traffic flow in the next iteration Convergence testing cannot guarantee a monotonically decreasing error. Therefore, this embodiment introduces a flow smoothing mechanism. Before performing a convergence check, the system first calculates the nearest... The average flow rate of the next iteration is denoted as... The calculation formula is as follows: ; in, The span of the sliding window is a preset integer constant (e.g., 3) used to define the time range of the smoothing process; Indicates the first In the next iteration, the road segment after sliding window smoothing... The traffic flow on the road segment will be used as the basis for the final convergence determination. Indicates in the current iteration (when ) and the past The road segment calculated in the next iteration Traffic flow on the road section.

[0100] After completing the traffic smoothing calculation, the system executes a convergence criterion based on average traffic flow. The collaborative control computing unit calculates the relative root mean square error of the road network traffic distribution and compares it with a pre-defined convergence accuracy threshold. Compare the results. The convergence condition is defined as: ; in, Represents a set of road segments All road sections Perform a traversal and summation operation; Indicates the first The road segment calculated in the second iteration Traffic flow on the road section; Indicates the first The road segment calculated in the second iteration Average flow rate; It represents the total smoothed traffic flow of the entire road network under the current iteration, and serves as a normalization factor in the denominator, making the criterion applicable to road networks of different sizes; This indicates the preset convergence accuracy threshold (e.g., 10). -4When the relative error is less than this value, the algorithm is considered to have converged.

[0101] If the above inequality holds, it indicates that the road network traffic distribution has achieved a monotonically decreasing state and reached a stable equilibrium. The system then terminates the inner loop and calculates the final average traffic flow. As the traffic flow of the road section Report to the upper-level model; otherwise, if the convergence condition is not met, let Then, return to the impedance update step to continue the next iteration. This improved mechanism ensures the robustness of solving the bilevel programming model.

[0102] As the top-level decision module of the hybrid solution architecture, the outer loop is responsible for global optimization. Its essence is to use a heuristic search strategy to select and iterate the optimal lane and signal coordination scheme that minimizes the total system impedance in a multi-dimensional and nonlinear solution space.

[0103] To adapt to the heterogeneous data characteristics of the decision variables in this invention—that is, lane function is a discrete attribute while signal parameters are continuous attributes—this embodiment constructs a hybrid encoding mechanism to digitally represent each potential control scheme. Specifically, the system encodes lane function variables... The mapping is performed as an integer encoded sequence, where each value in the sequence corresponds to a specific lane function division index for the approach lane (e.g., index "1" represents "straight + left turn", index "2" represents "left turn + straight + straight"); simultaneously, the signal period variable is... Compared with green credit variables It is directly mapped to a real-number encoded sequence. Based on this hybrid encoding strategy, the collaborative control computing unit will randomly generate... A set of initial control schemes that satisfy the basic physical constraints, this The group of solutions constitutes the initial candidate solution set in the algorithm logic, providing a basic sample for subsequent iterative optimization.

[0104] After establishing the foundation of the candidate solution set, how to accurately evaluate each candidate solution, i.e., each set of potential collaborative control schemes? The quality of these evaluations becomes crucial for algorithm advancement. This evaluation process relies on close data interaction between the outer and inner loops. The system passes candidate solution parameters from the candidate solution set as environmental constraints to the inner MSA module, which then completes the game equilibrium calculation and feeds back the steady-state road segment flow. Then, the outer loop immediately calls the objective function formula. Conduct fitness assessment.

[0105] Obviously, the calculated total system travel time The lower the value, the better the control scheme represented by the individual performs in alleviating congestion and improving efficiency, and the stronger its survival competitiveness in subsequent evolution.

[0106] Based on the fitness evaluation results, the system immediately initiates an optimization iteration process to generate new solutions with better performance. To explore new solution spaces while preserving existing high-quality features, this embodiment introduces a simulated evolutionary mechanism: a binary tournament selection operator is used to select better-performing solutions from the current candidate solution set as a foundation, and differentiated operator operations are implemented to generate new derived solutions. For real-number sequences representing signal parameters, simulated binary crossover and polynomial mutation algorithms are applied for fine-tuning; while for integer sequences representing lane parameters, discrete crossover and uniform mutation algorithms are applied for recombination. It is worth noting that to ensure the newly generated derived solutions are practically feasible in engineering (e.g., avoiding a sum of green ratios exceeding 1 or lane numbers violating physical limitations), the system specifically embeds a constraint correction mechanism in the computation stage to forcibly correct or eliminate non-compliant parameter combinations through a penalty function mechanism.

[0107] Once a new set of derived solutions is generated, the algorithm enters the update and selection phase. The system merges the candidate solution set from the previous round with the newly generated set of derived solutions, and selects the best solution based on the merits of each solution. The algorithm performs fast non-dominated sorting and crowding distance calculation. By comprehensively considering the performance level of the solutions and their sparsity in the solution space, the top-performing solutions are selected. The optimal solutions form the next generation of candidate solutions. This elite retention mechanism not only ensures the monotonicity of the algorithm's convergence direction but also effectively maintains the diversity of the search, preventing it from getting trapped in local optima. As the above evaluation, generation, and selection processes continuously iterate, when the preset termination condition is met (such as the number of iterations reaching a threshold or the objective function value stabilizing), the system will terminate the outer loop and output the set with the best fitness. The parameter combination is the final solution for lane and signal coordination optimization.

[0108] See attached document Figure 5 When the hybrid algorithm based on NSGA-II and MSA satisfies the termination condition and outputs the solution vector with the best fitness, the numerical result at the mathematical level cannot be directly recognized by the roadside equipment. Therefore, it needs to go through a strict decoding and conversion process to translate it into lane attribute instructions and signal timing schemes.

[0109] In this embodiment, the optimal solution output by the hybrid algorithm is essentially a multidimensional vector sequence containing heterogeneous data. The collaborative control computing unit first segments this vector, decomposing it into signal control parameter vectors according to predefined encoding rules. With lane function parameter vector Two parts.

[0110] For signal control parameter vector The decoding process mainly involves the conversion of physical dimensions from normalized values ​​to specific time values.

[0111] For the optimal signal period Since it is retained as a real number during the optimization process, the system directly rounds it down or to the nearest integer seconds, which is used as the uniform operating cycle length of the intersection.

[0112] For the optimal green light ratio The system is based on the formula The green light ratio of each phase is converted into a specific green light duration (in seconds), whereby... Intersection middle The green light duration for the phase. During this transition, the system will simultaneously perform a minimum green light time check; if the calculated... If the time is less than a safe threshold (such as the minimum time required for pedestrians to cross the street), the system will forcibly adjust it to the minimum value and correspondingly compress the green light time of non-critical phases to ensure the conservation of cycle length. The above process generates an executable signal timing scheme.

[0113] For lane function parameter vector The decoding process involves mapping and searching from discrete indexes to topological structures.

[0114] Lane function variables in the lane function parameter vector Each value is an integer index that points to a pre-defined "feasible lane function mode library" within the system. For example, when the lane function variable of a certain approach lane... At that time, the system can query the database to find that the physical configuration corresponding to the index is "left turn + straight + straight".

[0115] To transform this abstract configuration into driver-visible guidance information, the cooperative control computing unit generates corresponding lane attribute instructions (manifested as bitmap control instructions at the hardware level) based on the mapping results. These instructions are then sent to the variable message sign controller on the roadside, driving the LED screen to display the corresponding directional arrows (such as switching the original straight arrow to a left turn arrow), thereby completing the dynamic reconstruction of lane functions in physical space.

[0116] More importantly, after independently decoding the signal and lane, the system will perform a phase and lane consistency logic check. Since changes to lane attribute instructions (such as adding a left-turn lane) must be coordinated with corresponding signal release times to be effective, the decoding module will check the green light duration in the signal timing scheme. Does it cover all enabled lane directions? If it finds that a lane function has been changed but the corresponding signal phase is not activated (e.g., the left-turn lane is open but the left-turn phase green ratio is 0), the system will trigger an alarm and revert to the default safety scheme to prevent traffic conflicts caused by parameter mismatch.

[0117] The data structure parsing and basic logic judgment algorithms involved in the above process can be implemented by those skilled in the art according to conventional programming standards. These are well-known technologies in the field of computer data processing and will not be elaborated here.

[0118] In the process of converting the decoded optimal control scheme into physical actions, the system does not adopt an instantaneous switching mode, but follows a time-axis advancement logic with a safety buffer mechanism to smoothly transition the road network state and avoid traffic chaos caused by sudden changes.

[0119] In this embodiment, the cooperative control computing unit employs an asynchronous clearing and switching execution mechanism. Considering that changes to lane function involve alterations to vehicle queuing attributes within the physical space, and that their response lag is far greater than signal timing adjustments, the system prioritizes initiating the Variable Lane Sign (VMS) update process. Specifically, when it is determined that a new lane scheme needs to be implemented... During this transition period, the system will set a warning buffer period. The roadside VMS display will first switch to the new directional arrows or enter a flashing alert state to inform subsequently entering vehicles that the lane function has changed; simultaneously, the traffic signal controller will maintain the old timing scheme during this transition period. The core purpose of this operation is to give vehicles already queuing under the old lane function sufficient time to pass through the intersection, and to prevent left-turning vehicles from remaining in the intersection after, for example, "the original left-turn lane is changed to a straight lane", thus conflicting with the new straight-through release logic.

[0120] Once the warning buffer period ends and video or radar detectors confirm that the remaining queue in the relevant lanes has been cleared, the system immediately triggers the dynamic loading of signal control parameters for the new signal cycle. Compared to the new green credit When the green light cycle takes effect, the traffic signal controller employs a smooth transition strategy rather than a hard cut-off. If there is a significant difference between the old and new cycle lengths (e.g., from 120 seconds to 150 seconds), the controller will gradually adjust the cycle over the next 2 to 3 cycles by progressively stretching or compressing the green light duration of the common phase (usually the through phase of the main flow direction). This gradual adjustment strategy effectively maintains the relative stability of the green wave band on arterial roads, avoiding disruption of network-level phase difference coordination due to drastic cycle jumps.

[0121] Furthermore, to address discrepancies between model predictions and actual road conditions, the system incorporates a real-time closed-loop monitoring mechanism during the execution phase. The collaborative control computing unit continuously tracks traffic flow across road segments. The system monitors the real-time rate of change and the overflow of queue lengths at intersections. If, within the preset observation window after the control strategy is executed, the saturation of the target road segment not only fails to decrease but also experiences an abnormal surge (e.g., a deadlock occurs), the system will trigger the circuit breaker protection logic, immediately suspending the execution of the current strategy and rapidly rolling back the lane and signal parameters to fault-oriented safe default values. Regarding the underlying communication protocols of the signal controllers (such as NTCIP) and the VMS hardware driver interface technology involved in the above process, those skilled in the art can implement them using existing standard industrial control protocols. These are well-known technologies in the field of intelligent transportation facility control and will not be elaborated upon here.

[0122] Specific application examples: To verify the effectiveness and superiority of the multi-intersection variable guidance lane traffic signal control method based on path selection proposed in this invention, this embodiment selects a typical road network area in the core business district of a city for simulation experiments. This area includes 4 core signal-controlled intersections and 12 key road segments.

[0123] A road network environment was built using VISSIM simulation software, and MATLAB was configured as the collaborative control computing unit to achieve bidirectional data interaction through the COM interface.

[0124] First, construct a directed graph according to step S1. Set the design speed for the road section. The speed is 50 km / h. Initialize the road section impedance function parameters according to the formula in the instruction manual: ; Among them, the congestion coefficient Take 0.15, Choose 4.

[0125] Set signal control constraint parameters: minimum allowable value of signal period The maximum allowable value of the signal period Minimum green ratio threshold .

[0126] The optimization algorithm parameters are set as follows: the size of the candidate solution set is 50 groups, the maximum number of iterations is 50, and the parameter perturbation probability is 0.1.

[0127] The cooperative control computing unit runs a hybrid algorithm based on NSGA-II and MSA. The upper-level model uses the total system travel time. Minimize as the objective: ; During the solution process, the objective function value of the optimal solution in the candidate solution set is recorded in each iteration.

[0128] See attached document Figure 6 The horizontal axis represents the iteration number, and the vertical axis represents the total travel time Z of the system (unit: vehicle-hour).

[0129] It should be noted that the appendix Figure 6 The data label "1.85e+0.4" represents the value 1.85 x 10^4. 4 That is, the total system impedance in the initial state is 18,500 vehicle-hours; the data label "1.42e+0.4" in the figure represents the value 1.42 x 10⁻⁶. 4 That is, the total impedance of the system in the convergent state is 14200 car-hours.

[0130] Experimental data show that in the initial iteration phase (round 0), due to the use of a randomly generated initial control scheme set, the total system travel time is reduced. At 1.85X10 4 (i.e., 18500 car-hours) is the high point. As the iterative computation progresses, the algorithm quickly eliminates poorly performing solutions through parameter recombination, random perturbation, and elite retention strategies. In rounds 0-20, the curve shows a clear downward trend, indicating that the algorithm has good search efficiency and successfully guides the solution set towards the optimal solution space. After round 40, the curve tends to stabilize and eventually converges to 1.42 x 10⁻⁶. 4 (i.e., 14200) car-hours.

[0131] Data analysis shows that: Through the hybrid algorithm optimization of this invention, the total system impedance was reduced from 18,500 vehicle-hours in the initial state to 14,200 vehicle-hours, a decrease of approximately 23.2%. This specific numerical change strongly demonstrates that the algorithm can effectively escape local extrema and find lane function variables. Signal periodic variables Compared with green credit variables The optimal collaborative combination vector.

[0132] To further demonstrate the necessity of lane and signal coordination, this example sets up four comparative schemes for ablation experiments: Initial solution: Maintain the status quo, no optimization.

[0133] Option A (Signal Only): Lane function is fixed; only signal cycle variables are optimized. Compared with green credit variables .

[0134] Option B (Lane Only): Fixed signal timing, only optimizing lane function variables. .

[0135] Invention Solution: Lane Function Variable With signal periodic variables Compared with green credit variables All elements are optimized collaboratively.

[0136] We selected the most congested key intersections in the road network and statistically analyzed two indicators: average delay and maximum queue length. (See attached document.) Figure 7 The data comparison shown: Regarding average latency: the initial scheme had a latency as high as 58.4 seconds; while schemes A and B showed some improvement, decreasing to 45.2 seconds and 42.1 seconds respectively; however, after adopting the collaborative scheme of this invention, the average latency was reduced to 32.6 seconds. Compared to the initial scheme, this represents a reduction of 44.1%; compared to the single-dimensional schemes A and B, performance was improved by 27.8% and 22.5% respectively.

[0137] Regarding maximum queue length: the initial scheme resulted in a queue length of 180 meters, posing an overflow risk; the proposed solution successfully reduced this to 85 meters, a reduction of 52.7%, significantly better than scheme A (145 meters) and scheme B (110 meters). Result analysis: While adjusting signals or lanes individually is effective, bottlenecks exist. For example, when left-turn traffic is excessive, simply increasing the green light ratio (scheme A) leads to congestion in other phases, while simply increasing the left-turn lane (scheme B) without coordinating with the redistribution of green light time results in a waste of time and space resources. This invention utilizes a formula... It achieves dual dynamic matching of spatiotemporal resources, thereby obtaining the optimal congestion reduction effect.

[0138] To verify the robustness of the system, simulations were performed during simulation time. Minutes later, a sudden, massive traffic surge was injected into key road sections (traffic volume increased by 30%).

[0139] See attached document Figure 8 As shown, the horizontal axis represents the simulation time (minutes), and the vertical axis represents the saturation of key road sections (V / C ratio).

[0140] No dynamic control scheme (dashed line): In After the sudden influx of traffic, due to the lack of a response mechanism, the saturation level quickly rose and exceeded 0.9 (reaching an oversaturated state). Subsequently, the system fell into deadlock oscillation, and the saturation level remained at a high level of 0.95 to 1.0, preventing the road section from returning to normal operation.

[0141] The dynamic cooperative control scheme of this invention (solid line): when detected... After a sudden change in flow rate, the system triggers a re-optimization mechanism. Although, due to physical inertia, saturation... The green light ratio briefly rose to a peak (approximately 0.95) but then quickly fell back. This is because the roadside execution unit rapidly switched the variable lane function (such as adding an approach lane) and simultaneously extended the green light ratio.

[0142] Restoring steady state: With approximately 10 minutes, the solution of this invention can stably control the saturation at an ideal level of around 0.82, successfully mitigating the risk of congestion.

[0143] The experimental data above fully demonstrate that the lane and signal coordinated control method based on bi-layer planning proposed in this invention is not only superior to the traditional single-dimensional control method in terms of static indicators (delay, queuing), but also has the dynamic adjustment capability to cope with sudden traffic demands, effectively preventing road network deadlock and improving the resilience of urban transportation systems.

Claims

1. A traffic signal control method for multi-intersection variable guidance lanes based on path selection, characterized in that, Includes the following steps: Construct a road network topology model and collect traffic data, and generate a demand matrix for origin and destination points and an initial road segment impedance function based on the traffic data; The road network topology model, the origin-destination demand matrix, and the road segment impedance function are input into a pre-constructed two-layer planning model. The two-layer planning model includes an upper-layer model and a lower-layer model. The upper-layer model sets lane function variables, signal cycle variables, and green ratio variables. The lower-layer model receives the variables set by the upper-layer model and combines them with the origin-destination demand matrix to perform traffic flow allocation. A hybrid heuristic algorithm is used to solve the two-level programming model. The algorithm searches for the combination of lane function variables, signal period variables and green ratio variables in the upper-level model and calculates the corresponding equilibrium road segment flow in the lower-level model until a convergent optimal control scheme is obtained. The optimal control scheme is analyzed, and the numerical vectors contained in the optimal control scheme are decoded into lane attribute commands and signal timing schemes, which are then sent to the variable lane indicator and the signal controller drive equipment to take action, respectively.

2. The traffic signal control method for multi-intersection variable guidance lanes based on path selection according to claim 1, characterized in that, In the two-level programming model, the objective function of the upper-level model is set to minimize the total impedance of the road network. The minimized total impedance of the road network is composed of the sum of the products of the equalization segment flow and the segment travel time. The upper-level model sets constraints, including lane function conservation constraints to ensure the total number of lanes at the entrance remains unchanged, signal cycle duration constraints that set the maximum and minimum signal cycle values, and green light ratio constraints to ensure the minimum green light time for the phase.

3. The traffic signal control method for multi-intersection variable guidance lanes based on path selection according to claim 1, characterized in that, In the two-layer planning model, the upper-layer model and the lower-layer model are coupled through road segment capacity, which is defined as a function of the lane function variable and the green ratio variable. The upper-level model changes the road segment capacity parameters by adjusting the lane function variables or the green ratio variables, thereby changing the road segment travel time calculated by the road segment impedance function; The road segment impedance function parameters include intersection delay, which is determined by the lane function variable, the signal cycle variable, and the green ratio variable.

4. The traffic signal control method for multi-intersection variable guidance lanes based on path selection according to claim 3, characterized in that, The lower-level model uses the Logit model to describe the probability of a driver choosing different paths; Under the premise of satisfying the flow conservation constraint and the non-negativity constraint, the lower-level model allocates the traffic demand in the origin-destination demand matrix to the effective paths of the road network until all drivers can no longer reduce perceived impedance by changing paths, thereby obtaining the balanced road segment flow.

5. The traffic signal control method for multi-intersection variable guidance lanes based on path selection according to claim 1, characterized in that, The hybrid heuristic algorithm includes nested outer and inner loops; The outer loop uses a non-dominated sorting genetic algorithm to process the upper-level model, encodes the lane function variables as integer sequences, encodes the signal period variables and the green ratio variables as real number sequences, and generates a candidate solution set which is then passed to the inner loop.

6. The traffic signal control method for multi-intersection variable guidance lanes based on path selection according to claim 5, characterized in that, The inner loop uses the continuous averaging method to solve the lower-level model for each candidate solution set passed in by the outer loop. The inner loop updates the balanced road segment flow by loading auxiliary flow multiple times. At the end of each iteration, the average flow is calculated using a sliding window until the relative error of the average flow is less than a preset convergence threshold.

7. The traffic signal control method for multi-intersection variable guidance lanes based on path selection according to claim 6, characterized in that, The outer loop receives the balanced road segment flow returned by the inner loop, and substitutes the balanced road segment flow into the objective function of the upper model to calculate the fitness value; The outer loop performs selection, crossover, and mutation operations on the candidate solution set based on the fitness value to generate the next generation of candidate solution sets until the iteration termination condition is met and the optimal control scheme is output.

8. The traffic signal control method for multi-intersection variable guidance lanes based on path selection according to claim 1, characterized in that, The process of analyzing the optimal control scheme includes: Based on the preset lane function mode library, the lane function variables in the optimal control scheme are mapped to guide arrow bitmaps to generate the lane attribute instructions; The signal period variable in the optimal control scheme is rounded down, and the green light start and end times for each phase are calculated based on the green light ratio variable to generate the signal timing scheme.

9. The traffic signal control method for multi-intersection variable guidance lanes based on path selection according to claim 8, characterized in that, After generating the lane attribute command and the signal timing scheme, a consistency check is performed: Determine whether the green light time of each phase in the signal timing scheme covers the direction of the directional lane enabled in the lane attribute instruction; If a lane function is enabled but the corresponding phase green light time is missing, the lane attribute command will be intercepted and the system will revert to the default safety scheme.

10. The traffic signal control method for multi-intersection variable guidance lanes based on path selection according to claim 1, characterized in that, The variable lane indicator and the signal controller use asynchronous execution logic: The variable lane indicator is prioritized to display a new directional arrow and enter a warning buffer period. During the warning buffer period, the signal controller maintains the original timing scheme to clear the queued vehicles. After confirming that the queue of vehicles has been cleared, the signal controller loads and executes the new signal timing scheme.

Images