Adaptive switching traffic signal timing optimization method and system

Abstract

The invention discloses an adaptive switching traffic signal timing optimization method and system, and belongs to the technical field of intelligent traffic control. The method comprises the following steps: firstly, exploring by using a global optimization algorithm, and calculating a diversity index of a population in real time; when the population diversity index is continuously lower than a preset threshold value for multiple generations, triggering algorithm switching, and realizing smooth conversion of the population through an elitist retention and population mapping strategy; and then fine search is carried out by using a local optimization algorithm to obtain a final optimal signal timing scheme. According to the invention, through adaptive switching and intelligent mapping, deep collaboration of different optimization algorithms is realized, and the optimization precision, convergence speed and intersection traffic efficiency of a signal timing scheme are effectively improved.

Description

Technical Field

[0001] This invention belongs to the field of intelligent traffic control system technology, specifically relating to an adaptive switching traffic signal timing optimization method and system. Background Technology

[0002] Traffic efficiency at urban road intersections is a key bottleneck restricting the performance of the entire road network, and optimizing traffic signal control strategies has always been an important research topic in the field of intelligent transportation. In practical applications, timed signal control remains the most widely deployed basic control mode due to its high stability, low implementation cost, and ease of coordination. However, this strategy relies on historical traffic flow data and lacks the ability to perceive and adapt to real-time traffic fluctuations. When facing tidal traffic, sudden events, or highly random traffic flows, the control effect decreases significantly, leading to increased vehicle delays, longer queues, and reduced traffic efficiency.

[0003] To overcome the limitations of fixed timing, intelligent optimization algorithms have been introduced into the complex nonlinear optimization problem of traffic signal control. Among them, global optimization algorithms have attracted widespread attention due to their strong exploration capabilities. They extensively search the solution space through operations such as population crossover and mutation, avoiding premature entrapment in local optima. However, global optimization algorithms often face challenges such as insufficient local exploration capabilities and slower convergence speed in the later stages of the search. Meanwhile, local optimization algorithms typically exhibit high accuracy and efficiency in the fine-grained search phase, demonstrating their unique value. However, the global exploration capabilities of these algorithms are often relatively limited, and their solution quality largely depends on the initial population distribution.

[0004] Furthermore, existing hybrid schemes often rely on fixed iteration counts or simple performance thresholds as conditions for algorithm switching, lacking dynamic perception and intelligent decision-making capabilities regarding the algorithm's inherent search state (such as population diversity), making it difficult to achieve a dynamic and precise balance between global exploration and local development.

[0005] The above situation indicates that single intelligent optimization algorithms generally suffer from inherent limitations in simultaneously achieving both global exploration and local exploitation capabilities. This consensus has led to hybrid optimization strategies becoming an important direction for technological development. However, in this field, existing hybrid schemes mostly rely on fixed iteration counts or simple performance thresholds as conditions for algorithm switching, and apply them to the optimization of traffic signal parameters (such as signal cycle and green light ratio). This fixed switching rule lacks dynamic perception and intelligent decision-making capabilities regarding the algorithm's internal search state (such as population diversity), making it difficult to achieve a dynamic and precise balance between global exploration and local exploitation during the optimization process. This results in the hybrid strategy's performance not being fully utilized, thus limiting further improvements in intersection traffic efficiency.

[0006] Therefore, there is an urgent need in this field to explore new optimization methods that can adaptively perceive the search state and intelligently decide when to switch, so as to more effectively balance global exploration and local development capabilities, thereby meeting the higher requirements for traffic signal control accuracy, convergence speed and robustness in modern complex and dynamic traffic scenarios. Summary of the Invention

[0007] To address the aforementioned issues, this invention discloses an adaptive switching traffic signal timing optimization method and system. Through adaptive switching and intelligent mapping, it achieves deep collaboration among different optimization algorithms, effectively improving the optimization accuracy, convergence speed, and intersection traffic efficiency of the signal timing scheme. It also exhibits good versatility and scalability.

[0008] To achieve the above objectives, the technical solution of the present invention is as follows:

[0009] An adaptive traffic signal timing optimization method includes the following steps:

[0010] S1: Obtain traffic status information of the target intersection; the traffic status information includes traffic flow, lane occupancy rate and queue length of each approach lane;

[0011] S2: Based on the traffic state information, the information is preprocessed, and a global optimization algorithm is used to explore signal timing schemes; candidate schemes are evaluated through traffic simulation or actual system operation, and performance indicators generated by traffic simulation or actual operation are collected to calculate their fitness, thus obtaining a preliminary optimized solution set; wherein, the parameters optimized by the signal timing optimization scheme include the total duration of the signal cycle and the green light time of each phase;

[0012] S3: During the iterative process of the global optimization algorithm, the population diversity index is calculated in real time. ;

[0013] S4: When the population diversity index is... For multiple consecutive generations, the value was below the preset threshold. When, trigger algorithm switching

[0014] S5: After triggering the algorithm switch, the current population of the global optimization algorithm is transformed through the elite preservation and nonlinear mapping strategy, and the parameters representing the timing scheme of the signal are mapped to the solution space of the local optimization algorithm to form the initial population of the local optimization algorithm.

[0015] S6: Use the local optimization algorithm to perform a fine search on the initial population; evaluate candidate schemes through traffic simulation or actual system operation, and collect output performance data to calculate their fitness, thereby obtaining the optimal signal timing optimization scheme;

[0016] S7: Apply the optimal signal timing scheme to the signal control of the traffic scenario.

[0017] Furthermore, in step S1, the traffic state information is obtained through a traffic simulation system or roadside sensing equipment, and the bidirectional data interaction between it and the optimization algorithm process is achieved through a data interaction interface.

[0018] Furthermore, in steps S2 and S6, the fitness function is constructed based on the average vehicle delay time, average queue length, and traffic capacity indicators, and adopts a weighted combination form. The fitness function is constructed in a weighted form, as shown below:

[0019]

[0020] in: An indicator used to characterize the average vehicle delay time; An indicator characterizing the average queue length; Indicators representing traffic capacity; This is a weighting factor for the average vehicle delay time. This is the weighting coefficient for the average queue length. This is the weighting coefficient for traffic capacity.

[0021] Furthermore, the traffic performance indicators such as average vehicle delay time, average queue length, and traffic capacity are dimensionless indicators after normalization.

[0022] Furthermore, the population diversity index in step S3 The calculation formula is as follows:

[0023]

[0024] in: Population size; For the first The individual in the first In the traffic signal optimization application of this invention, each individual corresponds to a candidate signal timing scheme, and each dimension ( Each of these corresponds to a signal control parameter to be optimized. For example, for a typical four-phase intersection, the parameters to be optimized may include: 1) total signal cycle duration, 2) green light time for the first phase, 3) green light time for the second phase, 4) green light time for the third phase, and 5) green light time for the fourth phase. Therefore, Representing the The total cycle of each timing scheme The phase-green light time represents the phase of this scheme, and so on;

[0025] For the current population in the th The arithmetic mean of the dimensional features, calculated using the following formula. .

[0026] Furthermore, in steps S2 and S5, the global optimization algorithm is a genetic algorithm, and the local optimization algorithm is a sand cat swarm optimization algorithm.

[0027] Furthermore, the population mapping in step S5 employs a strategy combining elite preservation and nonlinear mapping, specifically including:

[0028] Individuals with the top K% fitness ranking are directly passed to the local optimization algorithm;

[0029] The remaining individuals are mapped to the solution space using a sigmoid function;

[0030] Ensure that the initial population of the local optimization algorithm contains both high-quality solutions and maintains sufficient diversity.

[0031] Furthermore, the nonlinear mapping in step S5 maps the chromosome encoding of the global optimization algorithm to the solution space of the local optimization algorithm through a sigmoid function.

[0032] Furthermore, the sigmoid function is the sigmoid function.

[0033] The present invention also provides a traffic signal timing optimization system for implementing the above method, comprising:

[0034] The data acquisition module is used to acquire traffic status information at the target intersection;

[0035] The global exploration module is used to run global optimization algorithms and explore signal timing schemes;

[0036] The status monitoring and switching judgment module is used to calculate the population diversity index in real time and trigger algorithm switching when it falls below a preset threshold.

[0037] The population mapping module is used to transform the population of the global exploration module through an elite retention and nonlinear mapping strategy after the algorithm switch is triggered.

[0038] The local search module is used to receive the initial population processed by the population mapping module and run a local optimization algorithm to perform a fine-grained local search.

[0039] The control output module is used to send the optimal signal control scheme obtained by the local search module to the signal control unit.

[0040] The system constitutes a complete closed-loop optimization process: the data acquisition module transmits traffic status information to the global exploration module; the status monitoring and switching judgment module analyzes the operating status of the global exploration module in real time and triggers algorithm switching; after the population mapping module completes the population conversion, the local search module starts working; finally, the control output module applies the optimal solution to signal control.

[0041] The beneficial effects of this invention are as follows:

[0042] This invention achieves intelligent collaboration and deep complementarity between different algorithms through an introduced adaptive switching mechanism. The core of this mechanism lies in intelligent decision-making through state awareness: firstly, by real-time monitoring of the population diversity index... It senses the search status of the global optimization algorithm in the solution space; when the diversity index is below the threshold for several consecutive generations, it determines that the global exploration potential has been exhausted, and then triggers the algorithm switch at the best time; this move fundamentally overcomes the blindness of the fixed switching rules in the traditional hybrid strategy.

[0043] By employing elite retention and nonlinear mapping strategies, a smooth transition and complementary advantages are achieved during the optimization phase. After the trigger switch, the high-quality solution set obtained from global exploration regarding parameters such as signal period and green light time is smoothly converted into the starting point for fine-grained search by the local optimization algorithm. This not only preserves the high-quality solution set and transforms it into a superior starting point for local search, effectively avoiding the risk of the local optimization algorithm getting stuck in local optima due to a poor initial population, but also greatly compensates for the lack of local development capabilities of the global optimization algorithm.

[0044] The aforementioned fusion approach based on intelligent decision-making allows different algorithms to leverage their strengths, thereby improving overall optimization accuracy and convergence speed. This invention is based on the following technological insight: population diversity index. It can effectively reflect the search status of the algorithm; when When the value is high, the algorithm is in an effective global exploration phase; when When the value is low and persists for many generations, the algorithm tends to converge. At this point, switching can achieve precise complementarity of the algorithm's advantages. After optimization by the method and system, key performance indicators such as average vehicle delay time, average queue length, and number of passing vehicles can be effectively optimized, thereby significantly improving the traffic efficiency of the intersection.

[0045] The method and system architecture are clear, and the module functions are well-defined. In particular, the status monitoring and switching judgment module and the population mapping module constitute a complete closed-loop process of data acquisition, intelligent optimization and control execution. This invention not only supports full verification and parameter tuning in a simulation environment, but also has the potential to be integrated with actual traffic signal control systems through interfaces due to its modular design, demonstrating good versatility and practicality. Attached Figure Description

[0046] Figure 1 This is a flowchart of the traffic signal optimization method of the present invention.

[0047] Figure 2 This is a principle block diagram of the GA-SCSO adaptive hybrid optimization strategy provided in the embodiments of the present invention.

[0048] Figure 3 This is a schematic diagram illustrating the interaction between the optimization system and the simulation platform / transportation system provided in this embodiment of the invention. Detailed Implementation

[0049] The present invention will be further illustrated below with reference to the accompanying drawings and specific embodiments. It should be understood that the following specific embodiments are for illustrative purposes only and are not intended to limit the scope of the invention.

[0050] This embodiment uses the SUMO simulation platform as an example to explain in detail the implementation process of the technical solution of the present invention. The experimental scenario is a typical four-phase signal-controlled intersection with independent left-turn and straight-through phases. The hardware basis for the simulation is a general-purpose computing device, and the software environment is SUMO and Python; the optimization algorithm is implemented in Python and co-simulated with SUMO through the TraCI interface.

[0051] 1. Simulation Environment Setup

[0052] First, road network data processing is performed. Real road network data for the target intersection area is obtained from the OpenStreetMap open-source map platform and exported as a standard .osm file. Then, the netconvert command from the SUMO official toolchain is used to convert the .osm file into a high-precision road network file (.net.xml format) recognizable by the SUMO simulation kernel. This conversion process defines the number of lanes, lane connections, and generates traffic signal logic containing basic signal phases and default timing parameters, which together constitute the static scenario for subsequent traffic simulation. The optimization goal of this invention is to replace the aforementioned default timing parameters to find the optimal signal control scheme. It should be noted that the default timing parameters and the target parameters to be optimized in this invention refer to the total signal cycle duration of the intersection and the green light time allocated to each phase. During simulation initialization, these parameters are set to a default fixed value as the starting point for the optimization algorithm.

[0053] Secondly, traffic flow generation and route planning are performed. To simulate dynamic traffic flow, the randomTrips.py script from the SUMO toolchain is used to generate random vehicle travel demands within a specified simulation time window. This script first generates an intermediate file containing vehicle departure times, starting locations, and destination locations. To realize realistic vehicle behavior within the road network, the script internally calls SUMO's route planning tool, duarouter, to calculate a specific travel path conforming to traffic rules for each vehicle based on the generated travel demand file and the converted road network file (.net.xml). The final output of this process is a standard vehicle routing file (in .rou.xml format), which defines the generation time, departure ID, and complete travel path for each vehicle within the simulation period.

[0054] Finally, the configuration file is integrated. The path information from the generated road network file (.net.xml) and vehicle routing file (.rou.xml) is combined and written into a single SUMO configuration file (.sumocfg). This configuration file serves as the main entry point for the simulation, guiding the SUMO simulator to load the correct road network and traffic flow data at runtime, thereby completing the construction of the entire simulation environment.

[0055] 2. Specific parameter configuration examples

[0056] To enable those skilled in the art to implement this invention, a set of exemplary parameter configuration schemes is provided below. These schemes are based on conventional practice settings for optimization algorithms and can serve as a feasible starting point.

[0057] Table 1. Examples of Core Algorithms and Adaptive Switching Parameters

[0058] Module Parameter name Example value settings illustrate Global Optimization (GA) Algorithm Population size 50 Balancing search breadth with computational efficiency. Crossover probability 0.8 Control the frequency of information exchange between individuals. Mutation probability 0.1 Introduce new genes to maintain population diversity. Choose strategy Roulette selection method Probabilistic selection is based on fitness. Termination conditions Maximum 100 generations As a backup measure, it prevents an infinite loop when the switch is not triggered. Adaptive switching mechanism Population diversity threshold ( ) 0.02 Used to determine whether a population has lost its exploratory ability due to excessive convergence. Sustained Algebra 3rd generation Ensure that switching decisions are based on stable trends, rather than single-generation fluctuations. Local Optimization Algorithm (SCSO) Population size 50 It is inherited from the GA population through mapping, and remains consistent with the global exploration phase. Termination conditions Maximum 50 generations Control the depth of the local search. Population mapping strategy Elite retention rate 20% Ensure that high-quality solutions are directly retained for the next stage. Nonlinear mapping function Sigmoid function Achieve a smooth, bounded transformation of the solution space from GA to SCSO.

[0059] Genetic algorithm parameters: population size can be set to 50; crossover probability can be set to 0.8; mutation probability can be set to 0.1; selection strategy can be roulette wheel selection; the iterative process of this algorithm uses the aforementioned adaptive switching mechanism as the main exit method, while setting a maximum number of iterations (e.g., 100 generations) as a minimum termination condition to ensure the feasibility of the optimization process.

[0060] Adaptive switching monitoring parameters: To achieve intelligent switching, the key threshold settings are based on the following principles: Population diversity index threshold. A small positive value can be selected to determine whether the differences between individuals in the population have significantly decreased; for example, the threshold can be set to 0.02. To enhance the stability of the switching decision and prevent misjudgments caused by single-generation fluctuations, it can be stipulated that the above conditions must be maintained for three consecutive generations before switching can be triggered, ensuring that the change in the algorithm state is based on a trend rather than chance.

[0061] Sand cat swarm optimization algorithm parameters: its population size is obtained by mapping the population of the self-genetic algorithm, and the number is 50; the algorithm termination condition is set to a maximum of 50 iterations.

[0062] Population mapping strategy parameters: Elite retention rate =20%; the nonlinear mapping function uses the Sigmoid function, which has the following form:

[0063]

[0064] in: The position in the solution space of the sand cat swarm algorithm after mapping corresponds to the actual signal timing parameters; The chromosome encoding value in the genetic algorithm is usually a real number vector, where each dimension corresponds to a signal timing parameter to be optimized. and These are the upper and lower bounds of the parameters, ensuring that the generated timing scheme is within a physically feasible range. For example, the feasible range for green light time is usually 20 to 90 seconds.

[0065] Fitness function definition: The optimization objective of the algorithm is guided by the fitness function. This function implements the weighted sum form described in the claims, and in this embodiment, it is constructed as follows:

[0066] To eliminate the influence of dimensions, the key performance indicators are first normalized:

[0067]

[0068]

[0069]

[0070] in: This represents the average vehicle delay time. This represents the average queue length. Traffic capacity refers to the number of vehicles that can pass through. , , These are reference values ​​for delay time, queue length, and number of vehicles passing through, obtained through a fixed timing scheme based on operational benchmarks.

[0071] Therefore, the specific expression for the fitness function is:

[0072]

[0073] in: This represents the normalized average vehicle delay time. The normalized average queue length Normalized traffic capacity; This is a weighting factor for the average vehicle delay time. This is the weighting coefficient for the average queue length. These are the weighting coefficients for traffic capacity. In this embodiment, these three parameters are set to 0.5, 0.3, and 0.2, respectively.

[0074] It should be noted that all algorithm parameters given in this embodiment (including population size, crossover rate, mutation rate, number of iterations, weight coefficients, and adaptive thresholds) are a set of exemplary settings. In practical applications, those skilled in the art can adjust these parameters near or within a reasonable range of conventional parameters according to the specific size of the target intersection, the complexity of traffic flow, and different requirements for computational efficiency and accuracy. All such adjustments should be considered within the scope of protection of this invention.

[0075] 3. Execution of the hybrid optimization process

[0076] refer to Figure 1 The overall flowchart shown illustrates the hybrid optimization process, which is executed according to the following steps:

[0077] (S1) Initialization and Data Acquisition: Start the SUMO simulation and initialize the TraCI connection on the Python side. After the simulation starts, the algorithm reads the traffic status of the intersection in real time through the TraCI interface as the basis for optimization.

[0078] (S2) Genetic Algorithm (GA) Global Exploration Phase: An initial population of, for example, 50 individuals is randomly generated. In each generation iteration, each individual in the population is deployed to the SUMO simulation environment via the TraCI interface and run for a sufficient simulation time (e.g., 300 seconds) to obtain stable traffic flow performance, and its fitness value is calculated from the simulation output. Subsequently, selection, crossover, and mutation operations are performed to generate a new generation of the population.

[0079] (S3) Status monitoring: During the iteration process of the genetic algorithm (S2), the diversity index of the population is calculated in real time. Among them, the population diversity index The calculation formula is:

[0080]

[0081] Through monitoring The system can sense the search state of the genetic algorithm by observing changes in the value: when When the value remains consistently low, it indicates insufficient population diversity, increasing the risk of the algorithm falling into premature convergence.

[0082] (S4) Adaptive judgment: Monitor the population diversity index Changes in the population diversity index Three consecutive generations below the threshold When the value reaches (0.02), it is determined that the genetic algorithm has completed the global exploration phase, and the benefits of continuing to run it are limited. The algorithm switch is immediately triggered, and step S5 is executed. If this switch condition is not met, the algorithm returns to step (S2) to continue iterating until a minimum number of iterations (e.g., 100 generations) is reached.

[0083] (S5) Algorithm Switching and Population Mapping: After triggering the algorithm switch, the population is not simply transferred. First, an elite retention strategy is implemented, directly retaining the top 20% of the individuals in the current GA population (i.e., excellent signal timing schemes, including combinations of period and green light time parameters). Then, a nonlinear mapping is performed on the remaining 80% of individuals, mapping their chromosome codes (i.e., the encoded values ​​of parameters such as signal period and green light time) to the corresponding parameter positions in the solution space of the Sand Cat Group Algorithm through the Sigmoid function. During this mapping process, by setting the upper and lower bounds of the function parameters, it is ensured that the newly generated signal period and green light time parameters are always within the reasonable physical range of traffic engineering. The elite individuals and the mapped individuals together constitute the initial population of SCSO.

[0084] (S6) Local Fine-Search Phase of Sand Cat Group Optimization (SCSO): The Sand Cat Group Optimization algorithm is used to perform a local fine-search on the initial population (i.e., the population transformed by S5). In each iteration, for all individuals in the SCSO population, their new timing schemes are deployed to the SUMO simulation environment through the TraCI interface, and a sufficient simulation time (e.g., 300 seconds) is run to obtain stable traffic flow performance. The fitness value is calculated from the simulation output. Subsequently, the SCSO algorithm updates the position of each individual in the population (i.e., fine-tunes the timing scheme parameters) by simulating the search and attack behavior of sand cats based on the fitness value. This cycle is repeated until the termination condition of 50 generations is reached.

[0085] (S7) Results Output and Application: When the SCSO algorithm terminates, the individual with the highest fitness value is selected from its final population and determined as the optimal signal timing scheme. This final scheme is set as a fixed control scheme in the SUMO simulation environment through the TraCI interface, and long-term simulations (e.g., 3600 seconds) can be performed to comprehensively evaluate the stability and superiority of its control effect.

[0086] After the local fine-grained search (S6), an optional secondary switching decision can be made. The trigger condition is: monitoring the relative improvement in the fitness value of the optimal individual in the local optimization algorithm; if this improvement is lower than a preset threshold for several consecutive generations, a switching is triggered. The formula for calculating the relative improvement is as follows:

[0087]

[0088] in, This represents the optimal fitness value for the current generation. This is the optimal fitness value of the previous generation. If there are M consecutive generations (e.g., M=3), this relative improvement... All are less than a preset gain threshold. (For example If the local search yields insufficient results, a second algorithm switch is triggered, mapping the population back to the global algorithm and restarting from step S2, forming a deep optimization process of "global → local → global".

[0089] 4. Analysis of the effectiveness of the plan

[0090] To verify the effectiveness of this method, the GA-SCSO hybrid optimization scheme described in this invention was run and compared with the single algorithm optimization process. The results show that:

[0091] First, regarding stability optimization, this invention effectively avoids premature convergence caused by population diversity depletion in a single global optimization algorithm (GA) through an adaptive switching mechanism, while also overcoming the strong dependence of a single local optimization algorithm (SCSO) on the initial solution set. This mechanism fundamentally ensures that the optimization process can stably search for signal timing schemes in high-performance regions, significantly improving the reliability of the solution.

[0092] Secondly, regarding efficiency and accuracy optimization, the intelligent switching and population mapping mechanism based on population diversity monitoring ensures efficient transition from global exploration to local development. The global exploration phase provides a high-quality initial population for local search, enabling the proposed solution to achieve better solution accuracy with fewer iterations compared to a single algorithm. This significantly improves the overall convergence speed and the performance ceiling of the final solution.

[0093] The above analysis confirms that the adaptive hybrid optimization method provided by this invention has an inherent mechanism that ensures its significant advantages over traditional single algorithms or fixed switching strategies in terms of optimization stability, accuracy, and efficiency.

[0094] It should be noted that the above content merely illustrates the technical concept of the present invention and should not be construed as limiting the scope of protection of the present invention. For those skilled in the art, various improvements and modifications can be made without departing from the principle of the present invention, and all such improvements and modifications fall within the scope of protection of the claims of the present invention.

Claims

1. A method for adaptive switching traffic signal timing optimization, characterized by: The method comprises the following steps: S1: obtaining traffic state information of a target intersection; the traffic state information comprises vehicle flow, lane occupancy and queue length of each entrance lane; S2: based on the traffic state information, pre-processing the information, using a global optimization algorithm to explore a signal timing scheme; evaluating the candidate scheme through traffic simulation or actual system operation, collecting performance indicators generated by the traffic simulation or actual operation to calculate the fitness of the candidate scheme, and obtaining a preliminary optimization solution set; wherein the parameters optimized by the signal timing optimization scheme include total signal cycle length and green time of each phase; S3: calculating the population diversity index in real time in the iteration process of the global optimization algorithm ; S4: when the population diversity index is below a preset threshold for consecutive generations algorithm switch is triggered S5: after triggering algorithm switching, converting the current population of the global optimization algorithm through an elite reservation and nonlinear mapping strategy, mapping the parameters representing the signal timing scheme to the solution space of the local optimization algorithm, and forming an initial population of the local optimization algorithm; S6: using the local optimization algorithm to perform fine search on the initial population; evaluating the candidate scheme through traffic simulation or actual system operation, collecting output performance data to calculate the fitness of the candidate scheme, and obtaining an optimal signal timing optimization scheme; S7: applying the optimal signal timing scheme to signal control of the traffic scene.

2. The adaptive switching traffic signal timing optimization method of claim 1, wherein: In step S1, the traffic state information is obtained through a traffic simulation system or a roadside perception device, and bidirectional data interaction between the traffic state information and the optimization algorithm process is realized through a data interaction interface.

3. The adaptive switching traffic signal timing optimization method of claim 1, wherein: In steps S2 and S6, the fitness function is constructed based on vehicle average delay time, average queue length and traffic capacity indicators, and a weighted combination form is adopted; the fitness function is constructed in a weighted form as follows: wherein: is an index representing the average delay time of vehicles; is an index representing the average queue length; is an index representing the traffic capacity; is a weight coefficient of the average delay time of vehicles, is a weight coefficient of the average queue length, is a weight coefficient of the traffic capacity.

4. The adaptive switching traffic signal timing optimization method of claim 3, wherein: The vehicle average delay time, average queue length and traffic capacity indicators are dimensionless indicators after normalization processing.

5. The adaptive switching traffic signal timing optimization method of claim 1, wherein: The population diversity index in step S3 The formula for calculating the population diversity index is as follows: wherein: is the population size; is the value of the jth individual on the ith characteristic, representing the ith parameter in a candidate signal timing scheme; is the value of the jth individual on the ith characteristic, representing the ith parameter in a candidate signal timing scheme; is the value of the jth individual on the ith characteristic, representing the ith parameter in a candidate signal timing scheme; is the value of the jth individual on the ith characteristic, representing the ith parameter in a candidate signal timing scheme; is the arithmetic mean of the current population on the ith characteristic, calculated as is the arithmetic mean of the current population on the ith characteristic, calculated as is the arithmetic mean of the current population on the ith characteristic, calculated as 6. The adaptive switching traffic signal timing optimization method of claim 1, wherein: In steps S2 and S5, the global optimization algorithm is a genetic algorithm, and the local optimization algorithm is a sand cat swarm optimization algorithm.

7. The adaptive switching traffic signal timing optimization method of claim 1, wherein: In step S5, the population mapping adopts an elite reservation and nonlinear mapping combined strategy, which specifically includes: Reserving the top K% individuals in terms of fitness to directly pass to the local optimization algorithm; Mapping the remaining individuals to the solution space through an S-shaped function.

8. The adaptive switching traffic signal timing optimization method of claim 1, wherein: In step S5, the nonlinear mapping maps the chromosome coding of the global optimization algorithm to the solution space of the local optimization algorithm through an S-shaped function; the S-shaped function is a Sigmoid function.

9. The adaptive switching traffic signal timing optimization method of claim 1, wherein: After step S6, secondary switching judgment is performed: the triggering condition is that the relative improvement amplitude of the fitness value of the optimal individual of the local optimization algorithm is monitored, and if the improvement amplitude is continuously lower than a preset threshold for multiple generations, switching is triggered; the calculation formula of the relative improvement amplitude is as follows: wherein, is the optimal fitness value of the current generation; is the optimal fitness value of the previous generation; if the relative improvement amplitude is less than a preset gain threshold , it is determined that the local search is insufficient, triggering the second algorithm switching, mapping the population back to the global algorithm, and starting from step S2 to execute again, forming a "global → local → global" deep optimization process.

10. A traffic signal optimization system for implementing the method of any one of claims 1 to 9, characterized by The method comprises the following steps: A data acquisition module for acquiring traffic state information of a target intersection; A global exploration module for running a global optimization algorithm and exploring a signal timing scheme; A state monitoring and switching judgment module for calculating a population diversity index in real time and triggering algorithm switching when the population diversity index is lower than a preset threshold; A population mapping module for converting the population of the global exploration module through an elite reservation and nonlinear mapping strategy after triggering algorithm switching; and A local optimization module for performing fine search on the initial population of the local optimization module. The local search module is configured to receive the initial population processed by the population mapping module and run a local optimization algorithm to perform a local fine search; The control output module is configured to issue the optimal signal control scheme obtained by the local search module to a signal control unit. The system constitutes a complete closed-loop optimization process: the data acquisition module transmits traffic state information to the global exploration module; the state monitoring and switching judgment module analyzes the running state of the global exploration module in real time and triggers algorithm switching; after the population mapping module completes population conversion, the local search module starts working; finally, the control output module applies the optimal scheme to signal control.

Images