A method, system, device and medium for coordinating signal timing optimization of a traffic artery

Abstract

The application provides a traffic trunk coordination signal timing optimization method, system, device and medium, belongs to the field of traffic control, and comprises the following steps: a dynamic model of traffic signal control is established by using a Markov decision process; a value function is obtained according to the traffic signal control model, and precise description of signal timing is realized. The value function is introduced into the brightness function and position updating function of the glowworm swarm optimization algorithm, and an adaptive light intensity absorption coefficient and a dynamic random disturbance factor are introduced to improve the glowworm swarm optimization algorithm, so that an improved glowworm swarm optimization algorithm is obtained, and the global search ability of the glowworm swarm optimization algorithm is improved. The adaptive neuro-fuzzy inference system is optimized and trained by using the improved glowworm swarm optimization algorithm, and a signal timing optimization model is obtained. According to the adjustment amount of the signal timing parameter output by the signal timing optimization model, adaptive intelligent control of the traffic signal timing is realized. The method can be applied to complex and variable traffic trunk environments, and can efficiently and accurately provide signal timing optimization strategies for traffic trunks.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of traffic control, and particularly relates to a traffic trunk coordination signal timing optimization method, system, device and medium. BACKGROUND

[0002] As the backbone of urban traffic network, trunk roads have large traffic flow and high traffic density, and the optimization of traffic signal control is of great significance to improve overall traffic efficiency. Traditional traffic signal timing methods are mostly based on fixed cycle or simple vehicle sensing, which are difficult to adapt to complex and variable traffic environment, leading to increased vehicle delay and queue length and reduced road capacity. The complexity and dynamics of urban traffic network make traffic signal control one of the key means to improve road traffic efficiency and alleviate traffic congestion.

[0003] In recent years, intelligent optimization algorithms have received extensive attention in the field of traffic signal control. As a new bionic intelligent algorithm, the firefly algorithm has strong global search ability and simple parameter adjustment, and has been applied to traffic signal timing optimization. However, the standard firefly algorithm is prone to local optimum and slow convergence when dealing with high-dimensional, nonlinear and multi-objective optimization problems, which limits its application in complex traffic environment. On the other hand, Markov Decision Process (MDP) is a powerful tool for handling stochastic dynamic systems, which can effectively describe the state transition and decision-making process in traffic signal control. By representing traffic state as Markov state, MDP can develop optimal signal timing strategies based on consideration of traffic flow randomness. However, the computational complexity of MDP in high-dimensional state and action space is high, making it difficult to be directly applied to actual traffic systems. SUMMARY

[0004] In order to solve the problem of traffic trunk coordination signal timing optimization in complex traffic environment, the application provides a traffic trunk coordination signal timing optimization method, system, device and medium.

[0005] In order to achieve the above purpose, the application provides the following technical scheme:

[0006] A traffic trunk coordination signal timing optimization method, specifically comprising the following steps:

[0007] Obtain traffic trunk coordination signal timing parameters and traffic flow parameters, and establish a traffic signal control model based on Markov Decision Process (MDP) according to the traffic trunk coordination signal timing parameters and traffic flow parameters, and obtain a value function according to the traffic signal control model.

[0008] An adaptive light intensity absorption coefficient and a dynamic random disturbance factor are introduced into the firefly algorithm, and the value function is introduced into the brightness function and position updating function of the firefly algorithm to form an improved firefly algorithm; the adaptive neural fuzzy inference system is trained based on the improved firefly algorithm to obtain a signal timing optimization model;

[0009] The real-time traffic trunk data is input into the signal timing optimization model, and an adjustment amount of the signal timing parameter is output.

[0010] Preferably, a traffic signal control model based on Markov decision process (MDP) is established according to the traffic trunk coordinated signal timing parameters and traffic flow parameters, and specifically includes the following steps:

[0011] A state space vector S of the traffic trunk is obtained, and the state space vector includes traffic states of each intersection, i.e., queue length and delay time, and a state vector formula s t Specifically,

[0012] s t =[L q,1 (t),L q,2 (t),...,L q,N (t)];

[0013] Wherein, t is time; N is the number of intersections on the trunk; L q,N (t) represents the queue length;

[0014] An action space vector A of the traffic trunk is obtained, and the action space vector includes signal timing strategies, i.e., green light time, red light time and phase difference, and an action vector a t The formula is specifically:

[0015] a t =[G1(t),G2(t),...,G N (t),θ1(t),θ2(t),...,θ N (t)];

[0016] Wherein, θ N (t) is the phase difference of the Nth intersection; G N (t) is the effective green light time of the Nth intersection;

[0017] A state transition probability function is established, which represents the probability that the system transitions to the next state s t under the condition of given state s t and action a t+1 , and is specifically:

[0018]

[0019] In the state st Take action a t The obtained immediate reward is a reward function, which is specifically:

[0020]

[0021] Wherein, T d,i (t) represents the delay time of the i-th intersection at time t; L q,i (t) represents the queue length; Q(t) represents the traffic capacity;

[0022] From state s t , the expected value of future cumulative reward is a value function, which is specifically:

[0023]

[0024] Wherein, γ∈(0,1) is a discount factor; E is an expected value;

[0025] The calculation of the value function V(s t ) uses the Bellman equation, which is specifically:

[0026]

[0027] The value iteration algorithm is used to solve the value function, which is specifically:

[0028]

[0029] Preferably, a target function and a constraint condition are further included for the traffic signal control model, the target function is specifically:

[0030]

[0031] The constraint condition is specifically to constrain the signal cycle, green ratio, phase difference, minimum green time and saturation, which is specifically:

[0032]

[0033]

[0034] Wherein, g i is the green ratio of the i-th intersection; θ i is the phase difference of the i-th intersection; G i is the effective green time of the i-th intersection; R i is the red light time of the i-th intersection; λ i (t) represents the saturation of the i-th intersection at time t, α represents the average vehicle length; μ represents the average vehicle speed.

[0035] Preferably, the value function is introduced into the brightness function and the position updating function of the firefly algorithm, and an adaptive light intensity absorption coefficient and a dynamic random disturbance factor are introduced to form an improved firefly algorithm, specifically by the following steps:

[0036] The value function is introduced into the brightness function, specifically as follows:

[0037]

[0038] Wherein, F i is the objective function value of the i-th firefly, is the value function of the state corresponding to the i-th firefly, and ε is a minimum value to avoid zero;

[0039] The attraction degree function in the firefly algorithm is specifically as follows:

[0040]

[0041] Wherein, γ ij is the value function of the state corresponding to the i-th firefly; β0 is the initial attraction degree of the firefly; r ij represents the Euclidean distance between the i-th firefly and the j-th firefly;

[0042] The position updating function is specifically as follows:

[0043]

[0044] Wherein, is the value of the i-th firefly of the l-th generation in the k-th dimension; ε k is a random number obeying a standard normal distribution; α (l) is a dynamic random disturbance factor;

[0045] The parameters are set to be adaptively adjusted, wherein the adaptive light intensity absorption coefficient γ ij is adjusted according to the following formula, and the random disturbance factor α (l) is adjusted according to the following formula:

[0046] α (l) = α max · e -ηl ;

[0047] Wherein, η is a decay coefficient;

[0048] In the position updating, the influence of the current state and the action on the value function is considered, and the direction that makes the value function maximum is selected for movement;

[0049] The value gradient of the value function is obtained by calculating the partial derivative of the value function to the action and the partial derivative of the action to the decision variable, and is specifically:

[0050]

[0051] The guidance based on the value function is specifically shown in the following formula:

[0052]

[0053] Wherein, λ is a learning rate.

[0054] Preferably, the adaptive neuro-fuzzy inference system is trained based on the improved glowworm swarm algorithm to obtain a signal timing optimization model, and the method specifically comprises the following steps:

[0055] Collecting coordinated intersection traffic data, including queue length L q,i (t), vehicle speed v i (t), saturation λ i (t) and the corresponding optimal green time adjustment amount ΔG i (t);

[0056] Randomly initializing a glowworm population to obtain glowworm population parameters;

[0057] Inputting the glowworm population parameters, calculating the fitness function, and the fitness function is specifically:

[0058]

[0059] Wherein, M is the number of samples; is the actual output; is the output predicted by the ANFIS model;

[0060] According to the fitness function value, the position of the glowworm, i.e. the parameter of the ANFIS model, is updated;

[0061] The model is trained until the convergence condition is met to obtain a signal timing optimization model.

[0062] Preferably, it also includes establishing a fuzzy rule in the adaptive neuro-fuzzy inference system, and the fuzzy rule is specifically: when the vehicle queue length is long and the traffic flow is large, the green light time is extended; when the vehicle queue length is short and the traffic flow is small, the green light time is shortened.

[0063] The application also provides a traffic trunk coordination signal timing optimization system, which specifically comprises:

[0064] The coordination control module is configured to acquire traffic trunk coordination signal timing parameters and traffic flow parameters, establish a traffic signal control model based on a Markov decision process (MDP) according to the traffic trunk coordination signal timing parameters and the traffic flow parameters, and obtain a value function according to the traffic signal control model.

[0065] The model construction module is configured to introduce the value function into a brightness function and a position updating function of a glowworm swarm optimization algorithm, introduce an adaptive light intensity absorption coefficient and a dynamic random disturbance factor, and construct an improved glowworm swarm optimization algorithm; and train an adaptive neuro-fuzzy inference system based on the improved glowworm swarm optimization algorithm to obtain a signal timing optimization model.

[0066] The decision output module is configured to input real-time traffic trunk data into the signal timing optimization model and output an adjustment amount of the signal timing parameters.

[0067] The application further provides a computer device, which comprises a memory, a processor and a computer program stored in the memory, and the processor executes the computer program to realize the steps in the traffic trunk coordination signal timing optimization method.

[0068] The application further provides a computer readable storage medium, which stores a computer program, and the computer program can realize the steps in the traffic trunk coordination signal timing optimization method when loaded by a processor.

[0069] The traffic trunk coordination signal timing optimization method has the following beneficial effects:

[0070] The application establishes a dynamic model of traffic signal control by using a Markov decision process to obtain a value function. The value function is introduced into a brightness function and a position updating function of a glowworm swarm optimization algorithm, an adaptive light intensity absorption coefficient and a dynamic random disturbance factor are introduced, and an improved glowworm swarm optimization algorithm is obtained to improve the global search ability and processing efficiency of the glowworm swarm optimization algorithm. The adaptive neuro-fuzzy inference system is trained according to the improved glowworm swarm optimization algorithm to obtain a signal timing optimization model. The signal timing optimization model can dynamically adjust the signal timing of the traffic trunk according to real-time traffic conditions, and the optimization method fully considers the dynamic characteristics and uncertainty of traffic flow to improve the intelligentization and adaptive level of the signal timing. BRIEF DESCRIPTION OF DRAWINGS

[0071] In order to more clearly illustrate the embodiments of the application and the design scheme thereof, the drawings required by the embodiments will be briefly introduced as follows. The drawings in the following description are only some embodiments of the application, and other drawings can be obtained by those skilled in the art without creative labor on the basis of the drawings.

[0072] In order to more clearly illustrate the embodiments of the application and the design scheme thereof, the drawings required by the embodiments will be briefly introduced as follows. The drawings in the following description are only some embodiments of the application, and other drawings can be obtained by those skilled in the art without creative labor on the basis of the drawings.Figure 1 A flow chart of a traffic trunk coordination signal timing optimization method of the present application.

[0073] Figure 2 A Markov state transition matrix diagram in an embodiment of the present application.

[0074] Figure 3 An improved glowworm swarm optimization algorithm diagram in an embodiment of the present application.

[0075] Figure 4 An output surface diagram of an adaptive neuro-fuzzy inference system in an embodiment of the present application.

[0076] Figure 5 A traffic flow distribution diagram after optimization in an embodiment of the present application.

[0077] Figure 6 A queue length diagram after optimization in an embodiment of the present application.

[0078] Figure 7 A comparison diagram of different algorithms in an embodiment of the present application.

[0079] Figure 8 A fuzzy rule activation degree heat map in an embodiment of the present application.

[0080] Figure 9 A queue length change with iteration number diagram in an embodiment of the present application. DETAILED DESCRIPTION

[0081] In order to make the technical solution of the present application better understood by those skilled in the art and to be implemented, the present application will be described in detail below in conjunction with the drawings and specific embodiments. The following embodiments are only used to more clearly illustrate the technical solution of the present application, and cannot be used to limit the protection scope of the present application.

[0082] EMBODIMENT

[0083] In view of the problems of insufficient optimization and poor adaptability in current urban traffic signal control, the present application proposes a traffic trunk coordination signal timing optimization method. A dynamic model of traffic signal control is established by using Markov decision process to provide a mathematical model basis for traffic signal timing optimization. The value function of Markov decision process is integrated into glowworm swarm optimization algorithm, and the value function information is introduced in the brightness calculation and position updating process to overcome the problem that traditional optimization algorithms are prone to fall into local optimum. The parameters, state transition probability and reward function of Markov decision process and improved glowworm swarm optimization algorithm are set. An adaptive signal timing optimization model is constructed by combining adaptive neuro-fuzzy inference system with improved glowworm swarm optimization algorithm and Markov decision process. By integrating the learning ability of neural network and the reasoning ability of fuzzy logic, the ability of the model to handle nonlinear and uncertain problems is enhanced, such as Figure 1As shown, specifically comprising the following steps:

[0084] S1, establish a traffic signal control model based on Markov Decision Process (MDP), define system state, action, transition probability and reward function, express traffic state as Markov state, as shown in the formula: Figure 2 As shown, a multi-objective optimization model is constructed to minimize vehicle delay time, queue length and maximize mainline traffic capacity.

[0085] S11: define mainline coordinated signal timing parameters, specifically: N is the number of intersections on the mainline; C is the signal cycle (seconds), the signal cycle of the intersections on the mainline in the present application is the same; G i is the effective green time (seconds) of the i-th intersection; R i is the red light time (seconds) of the i-th intersection; θ i is the phase difference (seconds) of the i-th intersection; g i is the green ratio of the i-th intersection, is the red ratio of the i-th intersection,

[0086] S12: define traffic flow parameters, symbols and definitions as follows: v i (t) represents the traffic flow (vehicles / hour) of the i-th intersection at time t, which is regarded as a random variable in the present application; s i represents the saturation flow (vehicles / hour) of the i-th intersection; λ i (t) represents the saturation degree of the i-th intersection at time t, α represents the average vehicle length (meters), including the headway; μ represents the average speed (meters / second).

[0087] S13: establish a Markov Decision Process (MPD) to describe the random process of mainline coordinated traffic signal control, the definition and formula are as follows: first define the state space S, mainly including the traffic state of each intersection, which refers to the queue length and delay time in the present application. The state vector formula is as follows:

[0088] s t =[L q,1 (t),L q,2 (t),...,L q,N (t)];

[0089] Wherein, t is the time; N is the number of intersections on the mainline; L q,N (t) represents the queue length.

[0090] Define the action space A, mainly including signal timing strategies, which refers to green time, red time and phase difference in the present application. The action vector at The formula is as follows:

[0091] a t = [G1(t), G2(t),..., G N (t), θ1(t), θ2(t),..., θ N (t)].

[0092] The state transition probability P(s t+1 |s t , a t ) represents the probability of the system transitioning to the next state s t under the condition of given state s t and action a t+1 . The formula of the state transition probability is as follows:

[0093]

[0094] The reward function R(s t , a t ) is set to represent the immediate reward obtained by taking action a t in state s t . The formula of the reward function is as follows:

[0095]

[0096] Where T d,i (t) represents the delay time of the i-th intersection at time t; L q,i (t) represents the queue length; Q(t) represents the traffic capacity.

[0097] Under the MDP framework, the value function V(s t ) is used to evaluate the expected cumulative reward that can be obtained by taking a certain action in a specific state. Specifically, the value function V(s t ) is set to represent the expected value of future cumulative rewards starting from state s t , and the specific formula is as follows:

[0098]

[0099] Where γ ∈ (0, 1) is the discount factor; E is the expected value.

[0100] The calculation of the value function V(s t ) uses the Bellman equation, and its formula is as follows:

[0101]

[0102] To solve the value function, the value iteration algorithm is used, and the specific formula is as follows:

[0103]

[0104] S14: Construct the traffic flow probability density function, to describe the random type of traffic flow, assume that traffic flow obeys normal distribution, the formula is as follows:

[0105]

[0106] Where: μ i Indicates the average traffic flow; Indicates the variance.

[0107] Considering the delay time, queue length and traffic capacity, the target is to maximize the cumulative reward, that is, to minimize the comprehensive traffic cost, the objective function is constructed as follows:

[0108]

[0109] S15: Set the constraint conditions, respectively, signal cycle constraint, green ratio constraint, phase difference constraint, minimum green time constraint and saturation constraint, the specific formula is as follows:

[0110] Signal cycle constraint:

[0111] Green ratio constraint:

[0112] Phase difference constraint:

[0113] Minimum green time constraint:

[0114] Saturation constraint:

[0115] S2, use improved firefly algorithm to optimize arterial coordination signal timing, as Figure 3 shown, the main improvements of improved firefly algorithm (IFA) are two points.

[0116] 1) Introduce adaptive light absorption coefficient γ and dynamic random disturbance factor α to improve the convergence speed and global search ability of the algorithm.

[0117] 2) Combine the value function of MDP, the value function V(s t ) is used to evaluate the pros and cons of traffic signal timing strategy, and then guide the optimization process of improved firefly algorithm (IFA), introduce the information of value function in brightness calculation and position update.

[0118] S21: Define algorithm parameters, symbols and definitions are as follows: Lmax : maximum iteration number; γ: light intensity absorption coefficient, adaptive adjustment; β0: initial attractiveness of firefly; α: random disturbance factor, dynamic adjustment.

[0119] S22: luminance function part, on the basis of traditional luminance function, combined with MDP value function, luminance function formula is as follows:

[0120]

[0121] Where: F i is the objective function value of the i th firefly, containing the information of value function, and ε is the minimum value to avoid zero.

[0122] The objective function value formula is as follows:

[0123]

[0124] Where: is the value function of the i th firefly corresponding state.

[0125] The attractiveness function value formula is as follows:

[0126]

[0127] Where: γ ij is the value function of the i th firefly corresponding state; r ij represents the Euclidean distance between the i th firefly and the j th firefly.

[0128] The position update formula is set as follows:

[0129]

[0130] Where: is the value of the i th firefly in the l th generation in the k th dimension; ε k is a random number obeying standard normal distribution; α (l) is a dynamic random disturbance factor.

[0131] Then set the parameter adaptive adjustment, random disturbance factor α (l) Adjustment formula is as follows:

[0132] α( l )=α max ·e -ηl ;

[0133] Where: η is the attenuation coefficient.

[0134] The improved firefly algorithm needs to combine MDP for position update. In the position update, the current state and action The influence on the value function is selected to move in the direction that makes the value function The maximum.

[0135] S23: When calculating the value gradient of the value function, the partial derivative of the value function with respect to the action and the partial derivative of the action with respect to the decision variable need to be calculated, and the specific formula is as follows:

[0136]

[0137] Guided based on the value function, the specific formula is as follows:

[0138]

[0139] Where: λ is the learning rate.

[0140] S24: The penalty function method is used to punish the objective function value of the solution that violates the constraint, and the penalty function formula is as follows:

[0141]

[0142] Where: M is the penalty coefficient; M c is the number of constraint conditions.

[0143] S3, in order to ensure the effectiveness and accuracy of the algorithm, the algorithm parameters, traffic parameters and model parameters are set and initialized in detail. The algorithm parameters and traffic parameters are set as follows:

[0144] The algorithm parameters of the firefly algorithm are set as follows: population size n: the value is 50, that is, 50 candidate solutions are optimized at the same time to ensure the diversity of the population and the global search ability of the algorithm; the maximum iteration number L max : set to 500 to ensure that the algorithm has enough iteration times to converge to the optimal solution; the initial attraction degree β0: the value is 1, indicating the maximum attraction of the firefly in the initial state; the range of light intensity absorption coefficient γ: the minimum value γ min : set to 0.1 to ensure that the attraction degree does not decay too fast, which is beneficial to global search, and the maximum value γ max : set to 1 to prevent the attraction degree from being too concentrated, which is beneficial to local search; the maximum value of the random disturbance factor α max : set to 0.5 to increase the search diversity in the early stage; the decay coefficient η: set to 0.01 to make α decay exponentially with the iteration number; the learning rate λ: set to 0.1 for step size control of the value function gradient; the penalty coefficient M is set to 10 6 , which is used for the punishment degree of the constraint condition.

[0145] For the Markov decision process parameters, the algorithm parameters are set as follows: the discount factor λMDP : value is 0.9, indicating the degree of emphasis on future rewards; state space S and action space A: discretized according to the actual traffic system, the discrete granularity is determined according to the computing power and the balance effect; time step Δt: set to 15 seconds, corresponding to the minimum unit of the signal period.

[0146] Traffic parameters are set as follows:

[0147] Arterial road parameters, number of intersections N: set to 3, representing 3 consecutive intersections on the arterial road that need to be coordinated; signal period C: set to 120 seconds; minimum green time G min : set to 10 seconds to ensure the safe passage of pedestrians and vehicles; maximum saturation λ max : set to 0.9 to avoid traffic collapse caused by excessive saturation.

[0148] Traffic flow parameters, traffic flow variance: set according to the fluctuation of historical data Saturation flow s i : value is 1800 pcu / h; average vehicle length α: value is 7.5 m, including the headway; average speed μ: set to 55 km / h; average traffic flow μ: set according to the average flow of each intersection based on historical data, parameters as follows:

[0149]

[0150] S4, introduce an adaptive neuro-fuzzy inference system (ANFIS) combined with an improved firefly algorithm to build a signal timing optimization model. This model can adaptively adjust the signal timing strategy in complex traffic environments, improve the efficiency and stability of arterial traffic, and realize the deep integration of ANFIS with the improved firefly algorithm and Markov decision process.

[0151] Adaptive neuro-fuzzy inference system (ANFIS) is a hybrid intelligent system combining neural networks and fuzzy logic, with learning ability and fuzzy reasoning ability. It can automatically generate fuzzy rules by learning input and output data, and realize modeling and control of complex nonlinear systems.

[0152] According to the characteristics of traffic flow, fuzzy rules between input and output are established, as shown in Figure 8 The activation degree of fuzzy rules is shown in the heat map. The two rules are as follows:

[0153] Rule 1: If the vehicle queue length is long and the traffic flow is large, extend the green light time.

[0154] Rule 2: If the vehicle queue length is short and the traffic flow is small, shorten the green light time.

[0155] The input variables x1, x2, x3 of the rule are the vehicle queue length L q,i (t), the vehicle speed v i (t), and the time saturation λ i (t) of the intersection, respectively. i The output variable y of the rule is the green light time adjustment amount ΔG i (t), as shown in the following equation.The integration of the ANFIS model and the improved firefly algorithm optimizes the parameters of the ANFIS model, including the membership function parameters c i ,σ ki and the rule output function parameters p k , r q,i , through the integration optimization, which includes the following steps:

[0156] S41: Collect traffic data at the coordinated intersection, including L i (t), v i (t), λ i (t), and the corresponding optimal green light time adjustment amount ΔG i (t).

[0157] S42: Initialize the firefly population, randomly initialize the firefly population, and each firefly represents a parameter set of the ANFIS model.

[0158] S43: Calculate the fitness function, as shown in the following equation:

[0159]

[0160] Where: M is the number of samples; is the actual output; is the output predicted by the ANFIS model.

[0161] S44: Update the firefly position, use the improved firefly algorithm to update the position of the firefly, i.e. the parameters of the ANFIS model, according to the brightness (fitness function value).

[0162] S45: Train the ANFIS model, repeat S43 and S44 until the convergence condition is met, i.e. the maximum number of iterations is 1000, then perform model optimization and signal timing adjustment. As shown in the following figure, the queue length changes with the number of iterations in the traffic trunk. Figure 9

[0163] Adjust the green light time ΔG i (t) according to the output of the ANFIS model, and the green light time adjustment amount ΔG i(t) represents the green time that needs to be increased or decreased for each intersection i under the current traffic state, and the green time updating formula is as follows:

[0164] G i (t+1)=G i (t)+ΔG i (t);

[0165] Under the new green time, other signal timing parameters are calculated to ensure that the constraint condition: red time R i (t+1)=C-G i (t+1) is met.

[0166] Phase difference adjustment: according to the Markov decision process, the phase difference θ i (t+1) is adjusted according to the current traffic state. The adjusted phase difference is fused with the Markov decision process at each decision time t i The green time adjustment amount ΔG i (t) predicted by the ANFIS model is used to update the system state st. The state transition probability P(s t+1 |s t ,a t ) of the Markov decision process is determined by the output of the ANFIS model. The specific composition of the system state includes a variety of key indicators reflecting the current state of the traffic system, including: traffic flow, vehicle queue length, saturation and signal timing parameters.

[0167] Wherein, the input and output variables of the ANFIS model come from the arterial coordination model, ensuring the data consistency between the models, and the traffic signal control model provides macroscopic traffic state information, which, combined with the improved firefly algorithm, provides the basis for the training and prediction of the ANFIS model. The improved firefly algorithm is not only used to optimize the signal timing parameters, but also used to optimize the parameters of the ANFIS model, realizing the unity and cooperation of the algorithm and processing different optimization objectives under the same algorithm framework. The parameter setting module provides the initial algorithm parameters, traffic parameters and model parameters, laying the foundation for the training of the ANFIS model and the operation of the improved firefly algorithm.

[0168] The innovation of the application lies in that (1) a dynamic model of traffic signal control is established by using Markov decision process (MDP). By representing the traffic state as Markov state, the description of traffic signal timing is realized. (2) An improved firefly algorithm is proposed and applied to traffic signal optimization, and an improved firefly algorithm (IFA) with adaptive light intensity absorption coefficient is designed. The value function of Markov decision process is integrated into the algorithm, and the value function information is introduced in the brightness calculation and position updating to improve the global search ability of the algorithm. (3) The adaptive neuro-fuzzy inference system is combined with the improved firefly algorithm and the Markov decision process, the ANFIS model parameters are optimized by the improved firefly algorithm, and the adaptive intelligent control of traffic signal timing is realized. The learning ability of neural network and the reasoning ability of fuzzy logic are integrated, and the ability of the model to deal with nonlinear and uncertain problems is enhanced. The optimized traffic flow, queue length and comparison with different algorithms correspond to Figure 5 、 Figure 6 and Figure 7 .

[0169] The application also provides a traffic trunk coordination signal timing optimization system, which specifically comprises:

[0170] A coordination control module is configured to obtain traffic trunk coordination signal timing parameters and traffic flow parameters, establish a traffic signal control model based on Markov decision process (MDP) according to the traffic trunk coordination signal timing parameters and the traffic flow parameters, and obtain a value function according to the traffic signal control model.

[0171] A model construction module is configured to introduce the value function into a brightness function and a position updating function of the firefly algorithm, introduce an adaptive light intensity absorption coefficient and a dynamic random disturbance factor, and construct an improved firefly algorithm; and train an adaptive neuro-fuzzy inference system based on the improved firefly algorithm to obtain a signal timing optimization model.

[0172] A decision output module is configured to input real-time traffic trunk data into the signal timing optimization model and output an adjustment amount of the signal timing parameters.

[0173] The modules in the traffic trunk coordination signal timing optimization system can be realized by software, hardware and combinations thereof. The modules can be embedded in or independent of the processor in the computer device in hardware form, or stored in the memory in the computer device in software form, so that the processor can call and execute the operations of the modules.

[0174] The present application also provides a computer device, comprising a memory, a processor and a computer program stored in the memory, and the processor executes the computer program to implement the steps in the method embodiment of coordinating signal timing optimization of traffic trunks. The specific implementation method can be referred to the method embodiment, which will not be repeated here.

[0175] Further, the present application also provides a non-transitory computer readable storage medium comprising instructions, and the storage medium stores a computer program. For example, the storage medium comprises instructions, and the above-mentioned instructions can be executed by the processor of the computer device to complete the above-mentioned method. For example, the non-transitory computer readable storage medium can be a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk and an optical data storage device, etc. When the computer program is executed by the processor, the steps in the method embodiment of coordinating signal timing optimization of traffic trunks can be implemented. The specific implementation method can be referred to the method embodiment, which will not be repeated here.

[0176] Those skilled in the art should understand that the embodiments of the present application can provide a method, a system or a computer program product. Therefore, the present application can take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer usable program code.

[0177] The present application is described with reference to flowcharts and / or block diagrams of the method, device (system) and computer program product according to the embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams can be implemented by computer program instructions, and the combination of the flows and / or blocks in the flowcharts and / or block diagrams. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device produce a device that implements the functions specified in the flowcharts and / or block diagrams. Figure 1 one or more flows and / or blocks Figure 1 an apparatus that performs the functions specified in one or more flows and / or blocks.

[0178] These computer program instructions can also be stored in a computer readable memory that can guide the computer or other programmable data processing device to work in a specific way, so that the instructions stored in the computer readable memory produce a product comprising instruction apparatus, which implements the functions specified in the flowcharts and / or block diagrams. Figure 1 one or more flows and / or blocks Figure 1 an apparatus that performs the functions specified in one or more flows and / or blocks.

[0179] These computer program instructions can also be loaded into a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 These computer program instructions can also be loaded into a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 These computer program instructions can also be loaded into a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks.

[0180] It should be noted that the above detailed description and specific examples, which have been presented for clarity of understanding, are not intended to limit the application, but only to describe exemplary embodiments of the application. Therefore, although the present application has been described in detail with reference to particular implementations, it is not intended to limit or restrict the application in any way. For instance, those skilled in the art will recognize that modifications and / or substitutions for certain techniques, designs, methods, and processes can be made without deviating from the spirit or scope of the present application. Accordingly, modifications and / or improvements can be made in the arrangement, operation, and details of the methods and examples disclosed herein without departing from the scope of the application. Moreover, although applied to the preferred embodiments, the application can be applied to other types of devices and systems. Therefore, the scope of the application should be determined not with reference to the above description, but should instead be determined with reference to the appended claims, along with the full scope of equivalents to which such claims are entitled. The claims should not be considered limited to the features or embodiments described herein, but should be considered to include any and all equivalents within the scope of the claims.

Claims

1. A method for optimizing signal timing of traffic arterial coordination, characterized in that, The method comprises the following steps: Traffic trunk coordination signal timing parameters and traffic flow parameters are acquired, a traffic signal control model based on a Markov decision process (MDP) is established according to the traffic trunk coordination signal timing parameters and the traffic flow parameters, and a value function is obtained according to the traffic signal control model; The value function is introduced into a brightness function and a position updating function of the glowworm swarm optimization algorithm, and an adaptive light intensity absorption coefficient and a dynamic random disturbance factor are introduced to form an improved glowworm swarm optimization algorithm; the adaptive neuro-fuzzy inference system is trained based on the improved glowworm swarm optimization algorithm to obtain a signal timing optimization model; the value function is introduced into the brightness function and the position updating function of the glowworm swarm optimization algorithm, and the adaptive light intensity absorption coefficient and the dynamic random disturbance factor are introduced to form the improved glowworm swarm optimization algorithm, specifically by the following steps: The value function is introduced into the brightness function, and specifically: ； wherein, is the objective function value for the th firefly, , is the value function for the state corresponding to the th firefly, is the minimum value to avoid division by zero; The attraction degree function in the glowworm swarm optimization algorithm is specifically: ； wherein, is the value function for the state corresponding to the i-th firefly; is the value function for the state corresponding to the i-th firefly; is the initial attractiveness of the firefly; denotes the Euclidean distance between the i-th firefly and the j-th firefly; denotes the Euclidean distance between the i-th firefly and the j-th firefly; denotes the Euclidean distance between the i-th firefly and the j-th firefly; The position updating function is specifically: ； wherein, is the first generation firefly in the first dimension; is a random number following a standard normal distribution; is a dynamic random disturbance factor; Setting parameter adaptive adjustment, wherein the adaptive light intensity absorption coefficient The adjustment formula is as follows, random disturbance factor The adjustment formula is: ； wherein is the attenuation coefficient; At position update, consider current state and action on value function, move in direction that maximizes value function ; The value gradient of the value function is obtained by calculating the partial derivative of the value function with respect to the action and the partial derivative of the action with respect to the decision variable, and specifically: ； Guidance is performed based on the value function, and a specific formula is as follows: ； wherein, is the learning rate; The adaptive neuro-fuzzy inference system is trained based on the improved glowworm swarm optimization algorithm to obtain the signal timing optimization model, specifically comprising the following steps: Collecting coordinated intersection traffic data, including queue length , vehicle speed , saturation , and corresponding optimal green time adjustment amounts ; The glowworm swarm population is randomly initialized to obtain glowworm swarm population parameters; The glowworm swarm population parameters are input, and a fitness function is calculated, the fitness function being specifically: ； wherein, is the number of samples; is the actual output; is the output predicted by the adaptive neuro-fuzzy inference system (ANFIS) model; The position of the glowworm, that is, the parameters of the ANFIS model, are updated according to the fitness function value; The model is trained until a convergence condition is met to obtain the signal timing optimization model; Real-time traffic trunk data is input into the signal timing optimization model, and an adjustment amount of the signal timing parameter is output.

2. The method of claim 1, wherein, The traffic signal control model based on the Markov decision process (MDP) is established according to the traffic trunk coordination signal timing parameters and the traffic flow parameters, specifically comprising the following steps: Obtaining a state space vector of a traffic artery The state space vector includes traffic states of each intersection, i.e. queue length and delay time, and a state vector formula Specifically: ； wherein, is time; is the number of intersections on the arterial; denotes the queue length; Obtaining an action space vector for a traffic artery The action space vector comprises signal timing strategies, i.e. green times, red times and phase differences, action vectors The formula is specifically: ； wherein, is the phase difference for the th intersection; is the effective green time for the th intersection; A state transition probability function is established to represent the probability of the system transitioning to the next state given the current state and action , specifically: ； In the state Take action The obtained immediate reward is a reward function, which is specifically: ； wherein, denotes the delay time of the th intersection at time ; denotes the queue length; denotes the capacity; From the state Starting, the expected value of future cumulative rewards is the value function, which is specifically: ； wherein, is a discount factor; is an expected value; The value function The calculation of the value function employs the Bellman equation, in particular: ； The value iteration algorithm is adopted to solve the value function, and specifically: 。 3. The method of claim 2, wherein, The target function is specifically: ； The constraint condition is specifically to constrain the signal cycle, the green ratio, the phase difference, the minimum green time and the saturation, and specifically: ； ； ； ； ； wherein, is the green split for the th intersection; is the phase difference for the th intersection; is the effective green time for the th intersection; is the red time for the th intersection; denotes the saturation of the th intersection at time , , denotes the average vehicle length; denotes the average vehicle speed.

4. The method of claim 1, wherein, The fuzzy rule of the adaptive neuro-fuzzy inference system is established, and the fuzzy rule is specifically: when the vehicle queue length is long and the traffic flow is large, the green time is extended; when the vehicle queue length is short and the traffic flow is small, the green time is shortened.

5. A system for arterial coordination signal timing optimization, comprising: The method comprises the following steps: The coordination control module is configured to acquire traffic trunk coordination signal timing parameters and traffic flow parameters, establish a traffic signal control model based on a Markov decision process (MDP) according to the traffic trunk coordination signal timing parameters and the traffic flow parameters, and obtain a value function according to the traffic signal control model. The model construction module is used for introducing the value function into a brightness function and a position updating function of the glowworm swarm optimization algorithm, introducing an adaptive light intensity absorption coefficient and a dynamic random disturbance factor, and constructing an improved glowworm swarm optimization algorithm; the adaptive neuro-fuzzy inference system is trained based on the improved glowworm swarm optimization algorithm to obtain a signal timing optimization model; the value function is introduced into the brightness function and the position updating function of the glowworm swarm optimization algorithm, the adaptive light intensity absorption coefficient and the dynamic random disturbance factor are introduced, and the improved glowworm swarm optimization algorithm is constructed, and the improved glowworm swarm optimization algorithm is specifically constructed through the following steps: The value function is introduced into the brightness function, and the value function is specifically as follows: ； wherein, is the objective function value for the th firefly, , is the value function for the state corresponding to the th firefly, is the minimum value to avoid division by zero; The attraction degree function in the glowworm swarm optimization algorithm is specifically as follows: ； in, For the first The value function of the state corresponding to each firefly; The initial attraction level of fireflies; Indicates the first The firefly and the first Euclidean distance between fireflies; The position updating function is specifically as follows: ； wherein, is the first generation firefly in the first dimension; is a random number subject to a standard normal distribution; is a dynamic random perturbation factor; Setting parameter adaptive adjustment, wherein the adaptive light intensity absorption coefficient The adjustment formula is as follows, random disturbance factor The adjustment formula is: ； wherein is the attenuation coefficient; At position update, consider current state and action on value function, move in direction that maximizes value function ; The value gradient of the value function is obtained by calculating a partial derivative of the value function to an action and a partial derivative of the action to a decision variable, and the value gradient is specifically as follows: ； The value function is guided based on the value function, and a specific formula is as follows: ； wherein, is the learning rate; The adaptive neuro-fuzzy inference system is trained based on the improved glowworm swarm optimization algorithm to obtain the signal timing optimization model, and the signal timing optimization model specifically includes the following steps: Collecting coordinated intersection traffic data, including queue length , vehicle speed , saturation , and corresponding optimal green time adjustment amounts ; A glowworm swarm population is randomly initialized to obtain glowworm swarm population parameters; The glowworm swarm population parameters are input, and a fitness function is calculated, and the fitness function is specifically as follows: ； wherein, is the number of samples; is the actual output; is the output predicted by the adaptive neuro-fuzzy inference system (ANFIS) model; The position of the glowworm, that is, the parameters of the ANFIS model, is updated according to the fitness function value; The model is trained until a convergence condition is met to obtain the signal timing optimization model The decision output module is used for inputting real-time traffic trunk data into the signal timing optimization model and outputting an adjustment amount of a signal timing parameter.

6. A computer device comprising a memory, a processor, and a computer program stored on the memory, wherein the computer program comprises instructions that, when executed by the processor, cause the processor to perform the method of any one of claims 1-5. The processor executes the computer program to implement the steps of the method in any one of claims 1 to 4.

7. A computer readable storage medium having stored thereon a computer program, characterized in that, The computer program is loaded by the processor to implement the steps of the method in any one of claims 1 to 4.

Images