A traffic signal light control method based on pulse neural network

Abstract

A traffic light control method based on a pulse neural network includes the following steps: first, a mathematical model of the traffic network load state is constructed, and the weighted sum of the number of waiting vehicles and the stop time in the lane is used as the load evaluation index. Secondly, the load state is converted into a pulse time series using a nonlinear and integral converter. Then, a pulse neural network model of a single traffic light is established to process the input pulse signal and control the phase switching. Furthermore, the pulse neural network parameters are binary-encoded for the multi-traffic light scenario, and a genetic algorithm is used to optimize the search using a distance metric. Finally, a genetic algorithm is applied to coordinate the pulse neural network parameters, and the fitness function is used to evaluate the index to maximize the overall traffic efficiency. This method realizes centralized and adaptive traffic light phase adjustment, significantly improves the road network throughput and reduces the delay time, while having the advantages of low computing and storage resource consumption.

Description

Technical Field

[0001] The present invention relates to a road signal light control technology, and in particular to a traffic signal light control method based on a pulse neural network. Background Art

[0002] Traffic light control technology refers to the use of sensors to collect traffic network load status information, such as the number of vehicles in a given lane, waiting time, average speed, etc., and adjust the traffic light cycle and phase based on this information to improve road network throughput and reduce delay time.

[0003] Existing traffic light control technologies mainly include three categories: the first category is operations research methods based on mixed integer programming and dynamic programming. These methods build relatively accurate system dynamics models based on the topological structure of the traffic network and the statistical characteristics of traffic flow, construct optimization problems, and solve the optimal traffic light period and phase duration. However, these methods are highly model-dependent and have low scalability for large-scale traffic light control; the second category is control methods based on linear quadratic regulators, maximum pressure control, and model predictive control. These methods use real-time feedback, prediction, and other means to make traffic light timing adaptive to real-time changes in traffic flow, but these methods emphasize the balance and stability of the system rather than global optimality; the third category is artificial intelligence methods based on reinforcement learning, including Q networks, policy gradients, actor-critic algorithms, etc. These algorithms are based on second-generation neural networks. For large-scale reinforcement learning models, second-generation neural networks consume relatively high computing and storage resources.

[0004] The applicant previously proposed a traffic light control method based on distributed reinforcement learning. This method divides a large-scale traffic network into several small-scale sub-networks and constructs a sub-network signal light control model based on a Q-network. Finally, a maximum plus algorithm is used to coordinate the output decisions of all Q-networks to optimize the overall traffic quality of the road network. The large-scale traffic network signal light control method based on this method, which uses deep Q-networks to control the road network sub-units, consumes a large amount of computing and storage resources.

[0005] It should be noted that the information disclosed in the above background technology section is only used to understand the background of this application, and therefore may include information that does not constitute prior art known to ordinary technicians in this field. Summary of the Invention

[0006] The main purpose of the present invention is to solve the problems existing in the above-mentioned background technology and provide a traffic light control method based on pulse neural network.

[0007] To achieve the above object, the present invention adopts the following technical solutions:

[0008] In a first aspect of the present invention, a traffic light control method based on a spiking neural network comprises the following steps:

[0009] Step 1: Construct a mathematical model of the urban traffic network load state and define the weighted sum of the number of waiting vehicles in the lane and the queue stop time as the evaluation index of the lane's current load;

[0010] Step 2: Convert the traffic network load state into a pulse time series, and implement pulse encoding of the state through a nonlinear function converter and an integral converter;

[0011] Step 3: Construct a spiking neural network model for single traffic light control, including the integral conversion of input pulse signals and pulse output, as well as the phase switching mechanism;

[0012] Step 4: For the multi-signal light control scenario, perform binary encoding of the pulse neural network parameters and use distance measurement to determine the order of binary features to prepare for genetic algorithm optimization.

[0013] Step 5: Use genetic algorithms to coordinate multiple pulse neural network parameters to optimize the overall traffic efficiency of the traffic network, and maximize the global traffic efficiency through fitness function evaluation indicators.

[0014] In a second aspect of the present invention, a method for jointly controlling traffic lights in an urban traffic network based on a spiking neural network comprises the following steps:

[0015] Real-time traffic status monitoring and evaluation: Real-time monitoring of traffic conditions at multiple intersections in the city, and use of monitoring data for evaluation, providing real-time decision support for traffic light control;

[0016] State vector pulse coding: The actual traffic state information monitored is converted into a pulse time series through a nonlinear function converter and an integral converter, which serves as the input of the pulse neural network;

[0017] Pulse neural network control model construction and implementation: Build a pulse neural network model for single traffic light control, implement the integral conversion of input pulse signals and pulse output, as well as the phase switching mechanism;

[0018] Parameter optimization and genetic algorithm coordination: Parameter optimization for multi-signal light control scenarios is carried out through binary encoding of pulse neural network parameters and genetic algorithm coordination to optimize road network traffic efficiency;

[0019] Global traffic efficiency optimization: Utilize genetic algorithms to coordinate the parameters of multiple pulse neural networks and evaluate indicators through fitness functions to maximize the overall traffic efficiency of the transportation network.

[0020] In a third aspect of the present invention, a computer-readable storage medium stores a computer program, which, when executed by a processor, implements the method for jointly controlling urban traffic network signal lights based on a pulse neural network.

[0021] In a fourth aspect of the present invention, a computer program product includes a computer program, which, when executed by a processor, implements the urban traffic network signal light joint control method based on a pulse neural network.

[0022] The present invention has the following beneficial effects:

[0023] The present invention proposes an innovative method for joint control of urban traffic network signal lights based on a pulse neural network. By constructing a pulse neural network model for phase switching of signal lights at a single intersection and optimizing multiple pulse neural network parameters using a genetic algorithm, efficient control of traffic lights is achieved. Compared with the signal light control method based on the Q network, the pulse neural network model designed by the present invention has the characteristics of event-driven and low energy consumption, which significantly reduces the consumption of computing and storage resources. In addition, the present invention adopts a distributed optimization concept and realizes real-time communication between vehicles and road facilities through wireless communication technology, so that the central controller can perform centralized and adaptive signal light phase adjustment according to real-time traffic network dynamic information, effectively improving the throughput of the road network and reducing the delay time. The low-power artificial intelligence model constructed by the present invention not only improves the intelligence level of traffic management, but also optimizes the overall traffic efficiency of the urban traffic network, demonstrating significant technological progress and application potential in the field of urban traffic control.

[0024] Other beneficial effects of the embodiments of the present invention will be further described below. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 This is a flowchart of a traffic light control method based on a pulse neural network according to an embodiment of the present invention.

[0026] Figure 2 An intersection (left) and a standard eight-phase diagram of a traffic light (right) according to an embodiment of the present invention.

[0027] Figure 3 Schematic diagram of traffic network load status pulse coding according to an embodiment of the present invention.

[0028] Figure 4 This is a pulse neural network model diagram for single intersection signal light control according to an embodiment of the present invention.

[0029] Figure 5 This is a pulse neural network parameter coding diagram of an embodiment of the present invention.

[0030] Figure 6A global optimization graph of a traffic network is implemented using a genetic algorithm according to an embodiment of the present invention. DETAILED DESCRIPTION

[0031] The following is a detailed description of the embodiments of the present invention. It should be emphasized that the following description is only exemplary and is not intended to limit the scope of the present invention and its application.

[0032] This paper proposes a method for the joint control of traffic lights in urban traffic networks based on a spiking neural network. This method is specifically applied in urban traffic networks, which utilize wireless communication technology to enable real-time communication between vehicles and between vehicles and road infrastructure. By establishing a central controller for the traffic network, the controller can implement centralized, adaptive adjustments to traffic light phases based on real-time dynamic information received from the traffic network, such as vehicle speed, location, and travel time. This significantly improves network throughput and reduces latency.

[0033] See Figure 1 The embodiment of the present invention provides a traffic light control method based on a pulse neural network, comprising the following steps:

[0034] Step 1: Construct a mathematical model of the urban traffic network load state and define the weighted sum of the number of waiting vehicles in the lane and the queue stop time as the evaluation index of the lane's current load.

[0035] See Figure 2 In some embodiments, step one specifically includes:

[0036] The number of waiting vehicles and the queue stop time in each lane are weighted and summed to form an evaluation index for the lane's current load. Based on the relative positions of the lanes and traffic flow directions, state variables are defined, including the load conditions of the uplink and downlink lanes. A signal light phase diagram is designed to distinguish between the main phase and other phases, and corresponding access rights are assigned to each lane. For each signal light phase, the corresponding lane load state is calculated, including a combination of maximum and average values ​​to represent the lane load condition in a specific phase. The overall state of the current intersection is defined, and the traffic load differences at the intersection are evaluated by comparing the load states in different phases.

[0037] Step 2: Convert the traffic network load state into a pulse time series, and realize the pulse encoding of the state through a nonlinear function converter and an integral converter.

[0038] See Figure 3 In some embodiments, step 2 specifically includes:

[0039] A nonlinear function converter, such as a square function, is used to process the traffic network load state to enhance the signal's discrimination and sensitivity. When a specific signal light phase is activated, the state variables are dynamically adjusted according to changes in the lane load state, such as when the load state of one phase decreases while the load state of another phase increases. Threshold crossing conditions are set for the state variables, and when the state variables cross these thresholds for the first time, a pulse signal is generated to reflect the sudden change in the traffic load state. The state variables are detected at preset time points, and based on the comparison results between the state variables and the thresholds, a corresponding pulse signal sequence is generated as the input of the pulse neural network, providing a basis for subsequent signal light control decisions.

[0040] Step 3: Construct a pulse neural network model for single traffic light control, including the integral conversion and pulse output of the input pulse signal, as well as the phase switching mechanism.

[0041] See Figure 4 In some embodiments, step three specifically includes:

[0042] A pulse neural network model is constructed for single-intersection traffic light control. The model processes input pulses encoded by a nonlinear converter; the encoded pulse signal is converted into a continuous signal by an integral converter for further processing and analysis; the processed continuous signal is compared with a preset upper limit value, and when the continuous signal exceeds the upper limit value, the pulse converter outputs a pulse signal; based on the pulse signal and the output of the integral converter, the traffic light is controlled to gradually transition from one phase to another to ensure the continuity and safety of traffic flow; a lock time is set to keep the traffic light in the current phase for a period of time after the pulse signal is output to prevent frequent switching of the traffic light phase; before the end of a phase of the traffic light, a decision is made whether to continue to maintain the current phase or switch to another phase based on continuous monitoring and evaluation of the traffic load status.

[0043] Step 4: For the multi-signal light control scenario, perform binary encoding of the pulse neural network parameters, and determine the order of binary features through distance measurement methods (such as Hamming distance) to prepare for genetic algorithm optimization.

[0044] See Figure 5 In some embodiments, step 4 specifically includes:

[0045] Through the joint optimization of pulse time parameters, the parameter affecting the pulse occurrence time, namely the phase transition time parameter, is adjusted to achieve global efficiency optimization of the transportation network. The pulse time parameters of each intersection are encoded as binary vectors, which together form a long binary vector used in the optimization process of the genetic algorithm. The ordering of binary vectors is defined using Hamming distance, so that vectors with more 1-bit bits are ranked higher to reflect the magnitude of the parameter value. The pulse time parameters are uniformly discretized from the lowest value to the highest value, forming a one-to-one correspondence with the binary vectors. The binary vectors of all intersections are spliced ​​into a long binary vector to complete the parameter encoding and provide input for the optimization of the genetic algorithm.

[0046] Step 5: Use genetic algorithms to coordinate multiple pulse neural network parameters to optimize the overall traffic efficiency of the traffic network, and maximize the global traffic efficiency through fitness function evaluation indicators.

[0047] In some embodiments, step five specifically includes:

[0048] The average number of vehicles entering and exiting the traffic network per unit time is calculated, and based on this, a network throughput evaluation index is defined, which is a weighted combination of the maximum and average numbers of vehicles entering and exiting the network. For each vehicle entering and exiting the traffic network, the difference between its actual travel time and the ideal travel time without interference from other vehicles is calculated, and this is defined as the delay time. The delay times of all vehicles are aggregated, and the weighted sum of the maximum delay time and the average delay time is calculated to obtain the delay time evaluation index of the entire traffic network. The network throughput and network delay time are combined to construct an overall traffic efficiency evaluation index of the traffic network, which is a weighted combination of the above two evaluation indicators. The traffic efficiency evaluation index is defined as a fitness function, which is used to evaluate the effectiveness of different traffic light control strategies and select the optimal strategy. A genetic algorithm is used to optimize the traffic light control parameters, and the iterative evolution of the parameters is guided by the fitness function to maximize the traffic network traffic efficiency.

[0049] See Figure 6 In some embodiments, the genetic algorithm specifically includes:

[0050] A population containing multiple samples is randomly generated, each sample is a binary vector, representing a set of potential pulse neural network parameters; the binary vector is converted into the pulse generation parameters of the integral converter, which is used to simulate the pulse generation in the signal light control process; each parameter vector is Simulation experiment to simulate traffic flow in time period ΔT Condition, and calculate the fitness function value, evaluateThe performance of each parameter vector; based on the fitness function value, the roulette wheel method is used to select samples with better performance, and a new generation of population is generated through crossover and mutation operations; the selection, crossover, and mutation operations are repeated until the number of iterations reaches the preset maximum value or the population parameters converge to a certain range; through the iterative evolution of the genetic algorithm, the pulse neural network parameter vector that can optimize the overall traffic indicators of the traffic network is determined.

[0051] An embodiment of the present invention further provides a method for jointly controlling urban traffic network signal lights based on a pulse neural network, comprising the following steps:

[0052] Real-time traffic status monitoring and evaluation: Real-time monitoring of traffic conditions at multiple intersections in the city, and use of monitoring data for evaluation, providing real-time decision support for traffic light control;

[0053] State vector pulse coding: The actual traffic state information monitored is converted into a pulse time series through a nonlinear function converter and an integral converter, which serves as the input of the pulse neural network;

[0054] Pulse neural network control model construction and implementation: Build a pulse neural network model for single traffic light control, implement the integral conversion of input pulse signals and pulse output, as well as the phase switching mechanism;

[0055] Parameter optimization and genetic algorithm coordination: Parameter optimization for multi-signal light control scenarios is carried out through binary encoding of pulse neural network parameters and genetic algorithm coordination to optimize road network traffic efficiency;

[0056] Global traffic efficiency optimization: Utilize genetic algorithms to coordinate the parameters of multiple pulse neural networks and evaluate indicators through fitness functions to maximize the overall traffic efficiency of the transportation network.

[0057] Through these implementation steps, the present invention realizes an efficient traffic light control strategy. By applying pulse neural networks and genetic algorithms, it optimizes the traffic light control of urban traffic networks, improves the traffic efficiency and load balance of the road network, and reduces the consumption of computing resources.

[0058] The present invention proposes a low-power neural network model for the joint control of traffic lights in urban traffic networks. Based on the pulse neural network model, a phase switching mechanism for signal lights at a single intersection is constructed; a genetic algorithm is used to coordinate the parameters of multiple pulse neural networks to maximize the overall traffic efficiency of the traffic network. In the present invention, a single-intersection phase switching model is constructed based on a pulse neural network, and a genetic algorithm is used to coordinate all pulse neural network parameters in the traffic network to achieve optimal road network traffic efficiency. The present invention adopts a distributed optimization concept. In the control of road network subunits, compared with the applicant's prior method, a pulse neural network is used to replace the Q network. It has the characteristics of event-driven and low energy consumption, and the consumption of computing and storage resources is significantly reduced.

[0059] The following further describes specific embodiments of the present invention and examples of algorithm implementation thereof.

[0060] Based on spiking neural networks, this paper proposes a framework and method for the coordinated control of traffic lights in urban traffic networks, improving network load balance and reducing transport delays. The method primarily includes the following components: 1. Constructing a mathematical description of the traffic network load state; 2. Designing an encoding method to convert the traffic network load into pulse signals; 3. Designing a single traffic light control method based on spiking neural networks; 4. Designing a binary encoding method for spiking neural network parameters for multi-traffic light control scenarios; and 5. Designing a method for the coordinated control of multiple traffic lights in a traffic network.

[0061] like Figure 1 As shown, the implementation process of the present invention mainly includes the following steps:

[0062] (1) Constructing a mathematical description of the traffic network load state

[0063] Figure 2 The left part of the diagram is a crossroads diagram, with the light blue track indicating the direction of traffic allowed at the intersection. Each lane has its own unique number. represents the number of vehicles waiting in lane i and queue stop time The weighted sum of is the evaluation index of the existing load of the lane, which is

[0064]

[0065] Among them, ρ1 and ρ2 are weights, which usually have and Transformed into functions of the same order of magnitude.

[0066] Figure 2The right side of the diagram shows a standard eight-phase signal light diagram. G represents a green light, with uppercase letters indicating that the corresponding lane has priority. g also represents a green light, with lowercase letters indicating that the corresponding lane yields to G. r represents a red light, indicating that the corresponding lane is closed. y represents a yellow light, indicating that the corresponding lane is about to turn red. For example, phases GGgGrrGGgGrr represent the phases corresponding to lanes 1, 2, 3, 7, 8, 9, 14, 15, 16, 20, 21, and 22, respectively. Phases 1 and 5 are primary phases. When these phases occur, right-turn lanes 1, 7, 14, and 20 are in G. Therefore, the phase shift has a minimal impact on right-turning vehicles, with the primary impact concentrated on through-travel and left-turn lanes.

[0067] The traffic lanes corresponding to phase 1 are 2, 3, 15, and 16, and the traffic lanes corresponding to phase 2 are 8, 9, 21, and 22. Let the load states corresponding to phase 1 and phase 2 be:

[0068]

[0069] Here, k=1 represents the load state corresponding to phase 1, where i=2, 3, 15, 16. k=5 represents the load state corresponding to phase 5, where i=8, 9, 21, 22.

[0070] Define the current state of the intersection as

[0071]

[0072] (2) Traffic network load state pulse coding

[0073] like Figure 3 As shown in Figure 2, the traffic network load state is encoded into a time-series pulse signal using the following encoding method:

[0074] Define nonlinear function transformer

[0075] f(s)=s 2

[0076] When the intersection signal is set to phase 1, Reduce, Increase, if Then s t Decreases, and no pulse is generated during this period; if Then s t Increase, at t0, t1...times s t Perform the test, if s t The first time you cross a segment point ( Figure 3 In the nonlinear converter, the red point shows the value), a pulse is generated, otherwise no pulse is generated, such as Figure 3 As shown, s is detected at t0, t1, and t3. tThe first time a segmentation point is crossed, pulses are generated at t0, t1, and t3.

[0077] (3) Pulse neural network structure design

[0078] Figure 4 The upper left figure shows a pulse neural network model for single intersection signal control, with input V i (t) is the pulse encoded by the nonlinear converter, Convert it to V(t), and then compare V(t) with the set upper limit In comparison, if The pulse converter outputs a pulse. Figure 4 The lower left and right diagrams show a cycle of the signal light conversion: Let t0 be the time and Gradually increasing and gradually decreases, so In the nonlinear converter, it approaches the upper left in sequence, because This is not the first time to cross the segment point, so V i (t) generates pulses at t0, t1, and t3, and the output of the integral converter V(t) gradually accumulates until after t3, when V(t) exceeds V o (t) Output pulse, the phase gradually transitions from 1 to 2-4 and then to 5, t lock This is the lock time after the output pulse, during which V(t) remains at 0 to prevent the main phase from frequently switching between 1 and 5. t4, t5, and t6 are after the end of phase 1. Gradually increasing and Gradually decreases, but This means that Phase 5 needs to continue, so V i (t) No pulse is output until time t7, Beyond V i (t) starts to output pulses, and after t7 and t 10 , t 11 At this moment, V(t) exceeds V o (t) Output pulse, the phase gradually transitions from 5 to 1 through 6-8.

[0079] (4) Pulse Neural Network Parameter Coding

[0080] Integral Generator In this paper, τ is a parameter that affects when the pulse occurs, that is, when the phase changes. The present invention achieves the best global efficiency of the traffic network by jointly optimizing the pulse time parameters τ of multiple intersections. To use the genetic algorithm for optimization, all pulse time parameters τ1, τ2, ..., τ n Encoded as a binary vector.

[0081] Figure 5 Shows the spiking neural network parameter encoding.

[0082] τ i Is a continuous variable. Assume that it corresponds to a P-bit binary vector. The lowest value of the vector is all 0 and the highest value is all 1. Using the definition of Hamming distance, the more 1-bits the intermediate vector contains, the higher the ranking. For vectors with the same 1-bit value, they are sorted by their numerical value. There are a total of Y binary vectors from the lowest value to the highest value. i From the lowest value to the highest value, the discretization is uniformly divided into Y points, which form a one-to-one correspondence with the value of the P-bit binary vector. n according to The above method constructs n binary vectors and concatenates them into a long binary vector, and the encoding is completed.

[0083] (5) Collaborative control of traffic network signal lights

[0084] Define the average number of vehicles entering a traffic network per unit time within a certain period of time as N. in , the average number of vehicles leaving the traffic network is N out , define the network throughput evaluation index:

[0085] N net =αmax{N in ,N out}+(1-α)mean{N in ,N out}

[0086] For vehicles entering and exiting a traffic network, the difference between their actual travel time and the ideal travel time (travel time without interference from other vehicles) is defined as the delay time D. q , where q is the vehicle number. The delay time of the traffic network is defined as:

[0087]

[0088] D net It is the weighted sum of the maximum and average delay times of all vehicles entering and exiting the traffic network within a specified time period.

[0089] Define the evaluation index of the overall traffic efficiency of the transportation network

[0090] E net =γ1Nnet +γ2D net

[0091] γ1 and γ2 are weights, which usually have the function of N net and D net Transformed into a function of the same order of magnitude. net (ΔT) is defined as the fitness function, where ΔT is a selected period of time. If there are multiple measured samples, E net (ΔT) is the average value of historical samples.

[0092] Figure 6 It shows that the global optimization of the transportation network can be achieved through genetic algorithm.

[0093] like Figure 6 As shown, the embodiment of the present invention uses a genetic algorithm to solve the pulse neural network parameter vector that can optimize the overall traffic index of the traffic network. The specific steps are as follows:

[0094] Step 1: Randomly initialize the population, i.e., a binary vector set containing m samples, convert it into the pulse generation parameters of the integrating converter, and conduct a simulation experiment for ΔT time to obtain the fitness function value;

[0095] Step 2: Based on the fitness function value, the roulette wheel method is used to select the population, and crossover and mutation operations are performed on the selected samples to obtain the next generation population;

[0096] Step 3: Check whether the number of iterations has reached the maximum value, or whether the population has converged to a certain area. If so, stop the calculation; if not, go to step 1 to continue the calculation.

[0097] In summary, the embodiment of the present invention proposes a traffic light control method based on a pulse neural network, wherein: (1) a pulse coding method is proposed to convert the load state of the urban traffic network into a pulse time sequence. (2) a pulse neural network structure is proposed for single signal light switching time control, which includes an integral converter that converts the input pulse signal into a continuous wave signal, and a pulse generator that converts the continuous signal into a pulse output. (3) a pulse neural network is proposed. A binary encoding method for network parameters is proposed, which uses Hamming distance to determine the order of binary features. (4) Genetic algorithms are used to coordinate the parameters of multiple pulse neural networks to improve the global traffic indicators of urban transportation networks.

[0098] The urban traffic network signal light joint control method based on pulse neural network proposed in the present invention has the characteristics of event-driven and low energy consumption compared with the signal light control method based on Q network, and consumes significantly less computing and storage resources.

[0099] An embodiment of the present invention further provides a storage medium for storing a computer program, which at least performs the above method when executed.

[0100] An embodiment of the present invention further provides a control device, comprising a processor and a storage medium for storing a computer program; wherein the processor is configured to execute at least the method described above when executing the computer program.

[0101] An embodiment of the present invention further provides a processor, which executes a computer program and at least performs the method described above.

[0102] The storage medium can be implemented by any type of non-volatile storage device, or a combination thereof. Among them, the non-volatile memory can be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), a magnetic random access memory (FRAM), a flash memory, a magnetic surface memory, an optical disc, or a compact disc read-only memory (CD-ROM); the magnetic surface memory can be a magnetic disk memory or a magnetic tape memory. The storage medium described in the embodiments of the present invention is intended to include, but is not limited to, these and any other suitable types of memory.

[0103] In the several embodiments provided by the present invention, it should be understood that the disclosed systems and methods can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as: multiple units or components can be combined, or can be integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the components shown or discussed can be through some interfaces, and the indirect coupling or communication connection of the devices or units can be electrical, mechanical or other forms.

[0104] The units described above as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one place. can also be distributed to multiple network units Yuan Shang; you can choose some of them according to actual needs or all units to achieve the purpose of the embodiment.

[0105] In addition, all functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may be separately used as a unit, or two or more units may be integrated into one unit; the above-mentioned integrated units may be implemented in the form of hardware or in the form of hardware plus software functional units.

[0106] Those skilled in the art will appreciate that all or part of the steps of the above-mentioned method embodiments may be implemented by hardware associated with program instructions, and the aforementioned program may be stored in a computer-readable storage medium. When the program is executed, the program executes the steps of the above-mentioned method embodiments. The aforementioned storage medium includes various media that can store program codes, such as mobile storage devices, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical disks.

[0107] Alternatively, if the above-mentioned integrated unit of the present invention is implemented in the form of a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiment of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a number of instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as mobile storage devices, ROM, RAM, magnetic disks or optical disks.

[0108] The methods disclosed in the several method embodiments provided by the present invention can be arbitrarily combined without conflict to obtain new method embodiments.

[0109] The features disclosed in several product embodiments provided by the present invention can be arbitrarily combined without conflict to obtain new product embodiments.

[0110] The features disclosed in several method or device embodiments provided by the present invention can be arbitrarily combined without conflict to obtain new method embodiments or device embodiments.

[0111] The above is a further detailed description of the present invention in conjunction with specific preferred embodiments, and the specific implementation of the present invention should not be considered to be limited to these descriptions. Those skilled in the art will recognize that, without departing from the scope of the present invention, several equivalent substitutions or obvious variations can be made, and the performance or use of the same should be considered to fall within the scope of protection of the present invention.

Claims

1. A traffic light control method based on pulse neural network, characterized in that: The steps include: Step 1: Construct a mathematical model of the urban traffic network load state and define the weighted sum of the number of waiting vehicles in the lane and the queue stop time as the evaluation index of the lane's current load; Step 2: Convert the traffic network load state into a pulse time series, and implement pulse encoding of the state through a nonlinear function converter and an integral converter; Step 3: Construct a spiking neural network model for single traffic light control, including the integral conversion of input pulse signals and pulse output, as well as the phase switching mechanism; Step 4: For the multi-signal light control scenario, perform binary encoding of the pulse neural network parameters and determine the order of binary features using a distance metric to prepare for genetic algorithm optimization. Step 5: Use genetic algorithms to coordinate the parameters of multiple spiking neural networks to optimize the overall traffic efficiency of the traffic network, and maximize the global traffic efficiency through fitness function evaluation indicators; Step 5 specifically includes: Calculate the average number of vehicles entering and exiting the traffic network per unit time, and use this to define a network throughput evaluation index, which is a weighted combination of the maximum and average numbers of vehicles entering and exiting the network. For each vehicle entering and exiting the traffic network, the difference between its actual travel time and the ideal travel time without interference from other vehicles is calculated, which is defined as the delay time; Summarize the delay time of all vehicles, calculate the weighted sum of the maximum delay time and the average delay time, and obtain the delay time evaluation index of the entire traffic network; Combining network throughput and network delay time, we construct an evaluation index for the overall traffic efficiency of the transportation network. This index is a weighted combination of the above two evaluation indicators. The traffic efficiency evaluation index is defined as a fitness function to evaluate the effects of different traffic light control strategies and select the optimal strategy; Genetic algorithms are used to optimize traffic light control parameters, and the iterative evolution of parameters is guided by fitness functions to maximize traffic network efficiency. The genetic algorithm specifically includes: Randomly generate a population of multiple samples, each of which is a binary vector representing a set of potential spiking neural network parameters; Convert the binary vector into pulse generation parameters of the integrating converter, which is used to simulate the pulse generation in the process of signal light control; Conduct simulation experiments for each parameter vector to simulate traffic flow conditions within a time period ∆T, and calculate the fitness function value to evaluate the performance of each parameter vector; According to the fitness function value, the roulette wheel method is used to select samples with better performance, and a new generation of population is generated through crossover and mutation operations; Repeat the selection, crossover, and mutation operations until the number of iterations reaches the preset maximum value or the population parameters converge to a certain range; Through the iterative evolution of genetic algorithms, the pulse neural network parameter vector that can optimize the overall traffic indicators of the transportation network is determined.

2. The traffic signal light control method based on pulse neural network according to claim 1, characterized in that: Step 1 specifically includes: The number of waiting vehicles in each lane and the queue stop time are weighted and summed to form an evaluation index of the lane's current load; Based on the relative positions of the lanes and the traffic flow direction, state variables are defined, including the load conditions of the uplink and downlink lanes; Design the phase diagram of the traffic light, distinguish the main phase from other phases, and assign corresponding access rights to each lane; For each signal phase, the corresponding lane load status is calculated, including a combination of maximum and average values ​​to represent the lane load condition at a specific phase; Define the overall state of the current intersection and evaluate the traffic load difference of the intersection by comparing the load states in different phases.

3. The traffic light control method based on pulse neural network according to claim 1 or 2, characterized in that: Step 2 specifically includes: Adopting nonlinear function converter to process traffic network load status to enhance signal discrimination and sensitivity; When a specific signal phase is activated, the state variables are dynamically adjusted according to the changes in lane load status; Setting threshold crossing conditions for state variables, when the state variables cross these thresholds for the first time, generating pulse signals to reflect the sudden change of traffic load status; The state variables are detected at preset time points, and based on the comparison results between the state variables and the threshold, a corresponding pulse signal sequence is generated as the input of the pulse neural network, providing a basis for subsequent traffic light control decisions.

4. The traffic signal light control method based on pulse neural network as claimed in claim 3, characterized in that: The change in the lane load state includes a load state decreasing in one phase and a load state increasing in another phase.

5. The traffic light control method based on pulse neural network according to any one of claims 1 to 2, characterized in that: Step three specifically includes: A spiking neural network model is constructed for single intersection signal control, which processes input pulses encoded by a nonlinear converter. The coded pulse signal is converted into a continuous signal through an integrating converter for further processing and analysis; Comparing the processed continuous signal with a preset upper limit value, when the continuous signal exceeds the upper limit value, the pulse converter outputs a pulse signal; According to the pulse signal and the output of the integral converter, the traffic light is controlled to gradually transition from one phase to another to ensure the continuity and safety of traffic flow; Set a lock time to keep the signal light at the current phase for a period of time after outputting the pulse signal to prevent the signal light from switching frequently. Before a phase of a signal light ends, a decision is made based on continuous monitoring and evaluation of the traffic load status whether to continue with the current phase or switch to another phase.

6. The traffic light control method based on pulse neural network according to any one of claims 1 to 2, characterized in that: Step 4 specifically includes: By jointly optimizing the pulse time parameters, the parameters that affect the pulse generation time, namely the phase transition time parameters, are adjusted to achieve global efficiency optimization of the transportation network; Encoding the pulse time parameters of each intersection into binary vectors, which together form a long binary vector for use in the optimization process of the genetic algorithm; The Hamming distance is used to define the order of binary vectors, so that vectors with more 1-bits are ranked higher to reflect the size of the parameter value; The pulse time parameter is discretized uniformly from the lowest value to the highest value to form a one-to-one correspondence with the binary vector; The binary vectors of all intersections are concatenated into a long binary vector to complete the parameter encoding and provide input for the optimization of the genetic algorithm.

7. A method for joint control of urban traffic network signal lights based on pulse neural network, characterized in that: include: Real-time traffic status monitoring and evaluation: Real-time monitoring of traffic conditions at multiple intersections in the city, and use of monitoring data for evaluation, providing real-time decision support for traffic light control; State vector pulse coding: The actual traffic state information monitored is converted into a pulse time series through a nonlinear function converter and an integral converter, which serves as the input of the pulse neural network; Pulse neural network control model construction and implementation: Build a pulse neural network model for single traffic light control, implement the integral conversion of input pulse signals and pulse output, as well as the phase switching mechanism; Parameter optimization and genetic algorithm coordination: Parameter optimization for multi-signal light control scenarios is carried out through binary encoding of pulse neural network parameters and genetic algorithm coordination to optimize road network traffic efficiency; Global traffic efficiency optimization: Using genetic algorithms to coordinate the parameters of multiple spiking neural networks and evaluating indicators through fitness functions to maximize the overall traffic efficiency of the transportation network; specifically, Calculate the average number of vehicles entering and exiting the traffic network per unit time, and use this to define a network throughput evaluation index, which is a weighted combination of the maximum and average numbers of vehicles entering and exiting the network. For each vehicle entering and exiting the traffic network, the difference between its actual travel time and the ideal travel time without interference from other vehicles is calculated, which is defined as the delay time; Summarize the delay time of all vehicles, calculate the weighted sum of the maximum delay time and the average delay time, and obtain the delay time evaluation index of the entire traffic network; Combining network throughput and network delay time, we construct an evaluation index for the overall traffic efficiency of the transportation network. This index is a weighted combination of the above two evaluation indicators. The traffic efficiency evaluation index is defined as a fitness function to evaluate the effects of different traffic light control strategies and select the optimal strategy; Genetic algorithms are used to optimize traffic light control parameters, and the iterative evolution of parameters is guided by fitness functions to maximize traffic network efficiency. The genetic algorithm specifically includes: Randomly generate a population of multiple samples, each of which is a binary vector representing a set of potential spiking neural network parameters; Convert the binary vector into pulse generation parameters of the integrating converter, which is used to simulate the pulse generation in the process of signal light control; Conduct simulation experiments for each parameter vector to simulate traffic flow conditions within a time period ∆T, and calculate the fitness function value to evaluate the performance of each parameter vector; According to the fitness function value, the roulette wheel method is used to select samples with better performance, and a new generation of population is generated through crossover and mutation operations; Repeat the selection, crossover, and mutation operations until the number of iterations reaches the preset maximum value or the population parameters converge to a certain range; Through the iterative evolution of genetic algorithms, the pulse neural network parameter vector that can optimize the overall traffic indicators of the transportation network is determined.

8. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method for jointly controlling urban traffic network signal lights based on a pulse neural network as described in any one of claims 1 to 7 is implemented.

9. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the method for jointly controlling urban traffic network signal lights based on a pulse neural network as described in any one of claims 1 to 7 is implemented.

Images