CN113470361A

CN113470361A - Traffic signal timing evaluation method and device and storage medium

Info

Publication number: CN113470361A
Application number: CN202110844577.0A
Authority: CN
Inventors: 丁乃侃; 逯兆友; 曾子祺; 卢林盛; 田壮
Original assignee: Wuhan Institute of Technology
Current assignee: Wuhan Institute of Technology
Priority date: 2021-07-26
Filing date: 2021-07-26
Publication date: 2021-10-01

Abstract

The invention provides a traffic signal timing evaluation method, a device and a storage medium, wherein the method comprises the following steps: importing intersection detection information, and carrying out fuzzy control analysis on the intersection detection information to obtain a signal period; and evaluating signal timing of the intersection detection information and the signal period to obtain an evaluation result. The invention combines the detection information and the signal period for evaluation, improves the accuracy of the evaluation result, avoids traffic confusion and achieves the effect of intelligent traffic.

Description

Traffic signal timing evaluation method and device and storage medium

Technical Field

The invention mainly relates to the technical field of traffic signal processing, in particular to a traffic signal timing evaluation method, a traffic signal timing evaluation device and a storage medium.

Background

Most of the existing traffic signal researches are evaluated based on delay, and the indexes for evaluating the intersection optimization effect are single, so that the evaluation result is easy to deviate, traffic is disordered, and the intelligent traffic effect cannot be achieved.

Disclosure of Invention

The invention aims to solve the technical problem of the prior art and provides a traffic signal timing evaluation method, a traffic signal timing evaluation device and a storage medium.

The technical scheme for solving the technical problems is as follows: a traffic signal timing evaluation method comprises the following steps:

importing intersection detection information, and carrying out fuzzy control analysis on the intersection detection information to obtain a signal period;

and evaluating the signal timing of the intersection detection information and the signal period to obtain an evaluation result.

Another technical solution of the present invention for solving the above technical problems is as follows: a traffic signal timing evaluation device includes:

the fuzzy control analysis module is used for importing intersection detection information and carrying out fuzzy control analysis on the intersection detection information to obtain a signal period;

and the evaluation result obtaining module is used for evaluating signal timing of the intersection detection information and the signal period to obtain an evaluation result.

Another technical solution of the present invention for solving the above technical problems is as follows: a traffic signal timing evaluation device comprises a memory, a processor and a computer program which is stored in the memory and can run on the processor, and when the processor executes the computer program, the traffic signal timing evaluation method is realized.

Another technical solution of the present invention for solving the above technical problems is as follows: a computer-readable storage medium, which stores a computer program that, when executed by a processor, implements a traffic signal timing evaluation method as described above.

The invention has the beneficial effects that: the signal period is obtained through fuzzy control analysis of the intersection detection information, the evaluation result is obtained through signal timing evaluation of the intersection detection information and the signal period, the detection information and the signal period are combined for evaluation, the accuracy of the evaluation result is improved, traffic chaos is avoided, and therefore the intelligent traffic effect is achieved.

Drawings

Fig. 1 is a schematic flow chart of a traffic signal timing evaluation method according to an embodiment of the present invention;

fig. 2 is a block diagram of a traffic signal timing evaluation apparatus according to an embodiment of the present invention.

Detailed Description

The principles and features of this invention are described below in conjunction with the following drawings, which are set forth by way of illustration only and are not intended to limit the scope of the invention.

Fig. 1 is a schematic flow chart of a traffic signal timing evaluation method according to an embodiment of the present invention.

As shown in fig. 1, a traffic signal timing evaluation method includes the following steps:

It should be understood that the intersection detection information may be obtained from a detector.

It should be understood that a fuzzy control system (fcs) is a control system that uses the basic ideas and theories of fuzzy mathematics, which is a control mode that can simulate the fuzzy logic thinking of the human brain. The fuzzy control system has good adaptability, so that the fuzzy control system is selected as a controller for controlling the cycle length in consideration of the running stability of the intersection. The fuzzy control system is divided into four components: (1) fuzzification, (2) fuzzy rule, (3) fuzzy reasoning and (4) clarification. The function of the fuzzy control part is to determine the signal period by inputting two variables that influence the formulation of the signal period.

In the embodiment, the signal cycle is obtained through fuzzy control analysis of the intersection detection information, the evaluation result is obtained through signal timing evaluation of the intersection detection information and the signal cycle, and the evaluation is performed by combining multiple information of the detection information and the signal cycle, so that the accuracy of the evaluation result is improved, the traffic disorder is avoided, and the intelligent traffic effect is achieved.

Optionally, as an embodiment of the present invention, the intersection detection information includes the number of vehicles at an intersection in a current cycle and the number of vehicles at an intersection in a previous cycle, and the process of performing fuzzy control on the intersection detection information to obtain a signal cycle includes:

calculating the saturation flow rate of the number of vehicles at the intersection in the current period to obtain the saturation flow rate;

calculating the flow change ratio of the number of vehicles at the intersection in the current period and the number of vehicles at the intersection in the previous period to obtain the flow change ratio;

determining a fuzzy rule by the saturation flow rate and the flow variation ratio;

carrying out fuzzy reasoning on the saturation flow rate and the flow change ratio according to the fuzzy rule to obtain a signal period grade;

and performing defuzzification processing on the signal period grade to obtain a signal period.

It should be understood that the number of vehicles at the intersection of the current period and the number of vehicles at the intersection of the previous period are both obtained from the detector, and the periods are divided according to preset time, and if the previous period is set to be 5 minutes, the start of the current period is 5 minutes later.

It should be understood that the present invention takes as input variables a real-time flow to saturation flow Ratio (RS), i.e. the saturation flow rate, and a real-time flow variation ratio (RR), i.e. the flow variation ratio. In other words, the ratio of real-time flow to saturated flow is used as the factor in determining the signal period length. The value range of the RS (namely the saturation flow rate) is [0,1], the closer to 1, the higher the saturation of the intersection is, and according to the traffic control principle, the higher the saturation is, the shorter the signal timing period is. The RS (i.e. the saturation flow rate) is divided into 5 levels: PB (Positive big), PS (Positive Small), ZO (zero), NS (negative Small), NB (negative big), wherein the membership function is 'trimf'. The value range of RR (namely the flow change ratio) is [ -1,1], and the closer the absolute value is to 1, the faster the traffic flow change at the intersection is, so that a smaller period is needed to adapt to the rapid change. Likewise, the RR (i.e., the flow rate variation ratio) is divided into 5 levels PB, PS, ZO, NS, nb. There are many methods for selecting the membership function, such as expert scoring and fuzzy statistics. This is not the focus of the discussion of the present invention.

Specifically, as shown in table one, table one is a specific fuzzy rule.

Table one:

the determination of fuzzy rules is the key to fuzzy logic algorithms. The larger the RS (i.e., the saturation flow rate), the shorter the signal period; the larger the absolute value of RR (i.e., the flow rate variation ratio), the shorter the signal period. Therefore, considering both RS (i.e. the saturation flow rate) and RR (i.e. the flow variation ratio) together, the resulting fuzzy rule should be symmetrical. The fuzzy rules of the present invention employ the 'if … and … the …' statements. For example, if RS is PB and RR is PB, then the signal period needs to be small, therefore, if RS is PB and RR is PB the Cycle is L.However, if RS (i.e. the saturation flow rate) is NS, then a larger period is needed, and RR (i.e. the flow rate variation ratio) is NB, then a smaller period is needed, and the two interact, then a medium period is selected. If one of the two represents a medium Cycle and the other represents a long Cycle or a short Cycle, the calculation is performed according to the long Cycle or the short Cycle.

It will be appreciated that the input variables (i.e. the saturation flow rate and the flow variation ratio) are input to a fuzzy controller, the level of the output variable is derived by fuzzy rules, and the specific values are derived by defuzzification. The invention takes the signal period as an output variable. Dividing the signal period into three levels: l (long), m (medium), s (short), the value range of the output variable can be calculated by the history data. The cycle of each time interval is obtained by using a Webster timing method, the maximum value and the minimum value are used as value range endpoints, and the value range of the obtained cycle is [65,120 ]. Accordingly, the membership function takes 'trimf'. In conjunction with fuzzy rules, one can see the approximate correspondence of input variables to output variables.

In the embodiment, the saturated flow rate of the number of vehicles at the intersection in the current period is calculated to obtain the saturated flow rate, the flow change ratio of the number of vehicles at the intersection in the current period to the number of vehicles at the intersection in the previous period is calculated to obtain the flow change ratio, the fuzzy rule is determined through the saturated flow rate and the flow change ratio, the signal period grade is obtained through fuzzy reasoning on the saturated flow rate and the flow change ratio according to the fuzzy rule, the signal period is obtained through defuzzification processing on the signal period grade, the signal control operation stability of the intersection is ensured, the stability of the signal period is also improved, therefore, the purpose that the large traffic flow is adapted to the smaller period is achieved, the traffic disorder is avoided, and the intelligent traffic effect is achieved.

Optionally, as an embodiment of the present invention, the calculating a saturation flow rate of the number of vehicles at the intersection of the current cycle, and obtaining the saturation flow rate includes:

calculating a saturation flow rate of the number of vehicles at the intersection of the current cycle by a first equation, wherein the first equation is as follows:

wherein m is_diThe number of vehicles at the intersection in the ith period is S, the saturation flow of the intersection is S, the detection time is delta t, and the saturation flow rate is RS.

Understandably, m_diThe number of vehicles at the whole intersection of the ith cycle (namely the number of vehicles at the intersection of the ith cycle) m measured by the detector_dIs the number of vehicles at the entire intersection measured by the detector within Δ t. Δ t is the detection time. S is the saturation flow of the intersection,the value of which is equal to the sum of the individual inlet saturation flows.

In the embodiment, the saturation flow rate of the number of vehicles at the intersection in the current period is calculated in the first mode to obtain the saturation flow rate, basic data are provided for subsequent processing, intersection signal control operation is stable, and the stability of the signal period is improved, so that the large traffic flow is adapted to a smaller period, the disorder of traffic is avoided, and the intelligent traffic effect is achieved.

Optionally, as an embodiment of the present invention, the calculating a flow rate change ratio between the number of vehicles at the intersection in the current cycle and the number of vehicles at the intersection in the previous cycle, and obtaining the flow rate change ratio includes:

calculating a flow change ratio of the number of vehicles at the intersection in the current period to the number of vehicles at the intersection in the previous period by using a second formula to obtain the flow change ratio, wherein the second formula is as follows:

wherein m is_diNumber of vehicles at intersection of i-th cycle, m_di-1The number of vehicles at the intersection in the i-1 th cycle is shown, and RR is the flow change ratio.

Understandably, m_di-1The number of vehicles measured by the detector in the previous cycle of the ith cycle (namely the number of vehicles at the intersection of the ith-1 cycle). i starts at 1 and the invention starts at 6a.m.

In the embodiment, the flow change ratio is obtained by calculating the flow change ratio of the number of vehicles at the intersection in the current period and the number of vehicles at the intersection in the previous period through the second formula, so that basic data are provided for subsequent processing, the intersection signal control operation is stable, and the stability of the signal period is improved, so that the large traffic flow is adapted to a smaller period, the traffic disorder is avoided, and the intelligent traffic effect is achieved.

Optionally, as an embodiment of the present invention, the intersection detection information further includes a trunk number set, a branch number set, a plurality of incoming vehicle flows, a plurality of incoming vehicle saturated flows, a plurality of red light periods, a plurality of single-lane traffic flows, and a plurality of maximum queuing lengths, and the evaluating the signal timing of the intersection detection information and the signal period to obtain the evaluation result includes:

s1: calculating the trunk road number set, the branch road number set, the signal period, the multiple incoming vehicle flows and the signal lamp time of the multiple saturated vehicle flows to obtain the green lamp time of the trunk road, the green lamp time of the branch road and the yellow lamp time of the intersection;

s2: calculating total delay of the intersection according to the third formula of the multiple incoming vehicle flow rates, the multiple saturated vehicle flow rates and the multiple red light time periods of the entrances to obtain a total delay value of the intersection, wherein the third formula is as follows:

wherein D (t) is the total delay value of the intersection when the branch green light duration is t, t_iIs the ith inlet red light duration, m_iThe flow of the incoming vehicle at the ith inlet, f_iThe saturation flow of the ith inlet vehicle is shown, A is a main road number set, and B is a branch road number set;

s3: calculating the total number of conflicts at the intersection by a fourth formula for the single-lane traffic volumes and the maximum queuing lengths, wherein the fourth formula is as follows:

E(t)＝∑_i∈AV_i ^0.65exp(-2.046+0.0122Q_i)+∑_i∈BV_i ^0.65exp(-2.046+0.0122Q_i)，

wherein E (t) is the total number of conflicts at the intersection when the branch green light time is t, V_iFor the ith import single lane traffic, Q_iMaximum queuing length for the ith inlet;

s4: adjusting the time of the green light time of the trunk road and the green light time of the branch road by using a Q-learning reinforcement learning algorithm to obtain the adjusted green light time of the trunk road and the adjusted green light time of the branch road, and obtaining a signal timing scheme according to the adjusted green light time of the trunk road, the adjusted green light time of the branch road, the signal period and the yellow light time of the intersection;

s5: importing next-time intersection detection information, wherein the next-time intersection detection information comprises a plurality of next-time incoming vehicle flow rates, a plurality of next-time incoming vehicle saturated flow rates, a plurality of next-time red light incoming time lengths, a plurality of next-time single-lane traffic volumes and a plurality of next-time maximum queuing lengths, and executing steps S2-S3 to obtain a next-time intersection total delay value and a next-time intersection total conflict number;

s6: and calculating an evaluation matrix for the total delay value of the intersection, the total number of conflicts of the intersection, the total delay value of the intersection at the next moment and the total number of conflicts of the intersection at the next moment to obtain an evaluation matrix, and taking the evaluation matrix as an evaluation result.

It should be understood that Reinforcement Learning (RL), one of the paradigms and methodologies of machine Learning, is used to describe and solve the problem of agent (agent) in interacting with the environment by Learning strategies to achieve maximum return or achieve specific goals. The Q-learning algorithm includes the following components: 1) the method comprises the steps of Agent,2) Action,3) Environment,4) State and 5) reward.

The key to reinforcement learning is the recursion based on Bellman Equation (Bellman Equation) as shown below:

from the above equation, it can be seen that RL is actually a markov process, i.e. the next state and the current state are related. And obtaining the maximum Q value through continuous trial and error learning. Most RL algorithms are based on equation 1, including Q-learning. The updated formula of Q-learning is shown as follows:

N_ewQ(s_t,a_t)＝Q(s_t,a_t)+α[R(s_t,a_t)+γmax Q(s_t+1,a_t+1)-Q(s_t,a_t)]，

q-learning differs from Bellman equalization in that Bellman equalization uses the expected value of the next state to estimate the current Q value, while Q-learning uses the maximum Q value of the next state to estimate the current Q value.

The RL is an online learning technology different from supervised learning and unsupervised learning, the learning is regarded as a process of 'heuristic evaluation', firstly, a learning system is called agent to sense the environment state, a certain action is taken to act on the environment, the environment changes the state after receiving the action, and simultaneously a return reward or punishment is given back to the reinforcement learning system, the reinforcement system selects the next action according to a reinforcement signal and the current state of the environment, and the selection principle is to increase the probability of the re-excitation. When agent interacts with the environment, the following steps are carried out:

1) agent perceives the environmental state s (t) at time t;

2) for the current status and instant report r (t), agent selects an execution action a (t);

3) when the action selected by agent acts on the environment, the environment changes, the environment state is transferred to the next new state s (t +1), and the instant report r (t) is given;

4) real-time reporting r (t) is fed back to agent;

5) and turning to the step 2, and if the new state is the end state, stopping the circulation.

Wherein, the real-time report r (t) is determined by the environmental status s (t) and the output a (t) of agent.

It should be understood that the initial state herein is defined as: after the initial period (namely the signal period) is solved by fuzzy control, the initial green time of each phase is calculated according to the proportional relation between the flow and the saturated flow of each intersection, and finally the determined signal timing scheme is obtained.

Specifically, the Q-learning algorithm aims to adjust the green time of each phase (i.e., the trunk green time and the branch green time), so as to improve the overall operation efficiency and safety of the intersection. And the Agent selects action according to the running condition of the intersection in the environment to adjust the green light duration of each phase (namely the green light time of the trunk road and the green light time of the branch road). Specifically, the timing scheme is re-programmed by increasing or decreasing the leg green time period (i.e., the leg green time) and correspondingly decreasing or increasing the trunk green time period (i.e., the trunk green time). And then, feeding back the running condition of the intersection in the environment to the agent through a report value, and then carrying out further decision by the agent. In the present invention, if a ∈ { -2, -1,0,3}, that is to say, each time an action that an agent can select in the present invention is: 1) reducing the branch green duration by 2s or 1s, 2) reaching an optimum without performing any action,3) increasing by 3 s. The subtraction of 3s is not chosen here to avoid repetitive steps, and if too much increase is found after 3s has been added, fine tuning can be performed by subtracting 2s or 1 s.

Like state, action size is also [0, b ]. b is the number of actions that can be selected when the action for each selection of agent is 3. The value of b is determined by the signal period, and it is generally considered that the traffic flow of the branch is not larger than the traffic flow of the trunk, so the green duration of the branch (i.e., the branch green time) is smaller than or equal to the green duration of the trunk (i.e., the trunk green time). Thus, we can initially establish an action matrix, which is expressed as follows:

the two-dimensional action matrix is b rows and b columns in size, where each element represents an action. However, many actions are not necessary, such as actions matrix, where the elements directly opposite the diagonal are not fetched. The actions in { -2, -1,0,3} may be performed except for the case of state 0 and state b.

A7X 7 action matrix is selected for the following description. The reason why such a large matrix is chosen is that 7 x 7 is enough in the calculation process of the intersection studied by the patent, in other words, the maximum value of b can be taken to be 6. After the action is executed, if the green lamp time length is found to be excessively increased, fine adjustment is needed, so that-1 or-2 can be selected. When the state is 6, the description is optimal, and no action is performed.

Specifically, the criteria for judging the operation efficiency of the intersection mainly include delay, queuing length, parking number and the like. The application considers the delay as an evaluation index to reflect the running efficiency of the intersection. In order to reduce algorithm complexity and accurately estimate intersection delay, the following formula is adopted for calculation:

wherein D (t) is the total delay of the intersection when the branch green light duration is t. t is t_iThe red light duration for the ith inlet.

And introducing the number of conflicts into intersection signal control as a consideration factor for evaluating the operation safety of the intersection. 6 models are established according to factors such as volume per lane cycle, shock wave area, maximum queue length, back-moving shock wave speed and plane ratio and the like. The intersection studied by the invention meets the model3, and the calculation formula is as follows:

E(t)＝

∑_i∈AV_i ^0.65exp(-2.046+0.0122Q_i)+∑_i∈BV_i ^0.65exp(-2.046+0.0122Q_i)

wherein E (t) represents the total number of conflicts at the intersection when the branch green light time is t. V_iRepresents the volume per lane per cycle (i.e., the single lane traffic) at the i-th import. Q_iDenotes the maximum queue length (i.e. the maximum queue length) of the ith entry.

Currently, indices, d (t) and e (t), representing the operating efficiency and the operating safety, respectively, are obtained by calculation. However, how to combine the two together is also a critical issue. The application provides that the change rate of the delay and the number of conflicts after the agent selects the action each time is considered in the calculation of the return function. For example, the percentage of efficiency improvement compared to before after the increase in the branch green time (i.e., the branch green time) plus the percentage of safety improvement is greater than 0, indicating that this action is optional. If all the reported values are less than or equal to 0 in a state, the state is an optimal state. Of course, their sum is inverted, as it is better if either the delay or the collision number is smaller. Therefore, the fourth equation is proposed as a reward function.

Thus, a complete Q matrix is built up. In the code operation, the return value of the element which cannot be taken is represented by a large negative number, and the invention takes-10000.

Specifically, in order to obtain a more accurate optimization result, calibration of some parameters in Q-learning is very important, such as the number of iterations, discount rate and learning rate.

The intersection self-adaptive signal system needs a quick and real-time timing scheme, so that the iteration times are selected for 100 times in order to improve the code running efficiency. In addition to this, 100 times are sufficient for the model proposed by the present invention.

The Learning rate (Learning rate) is an important super-parameter in supervised Learning and deep Learning, and determines whether and when the objective function can converge to a local minimum. An appropriate learning rate enables the objective function to converge to a local minimum in an appropriate time. For q-learning, the learning rate decreases exponentially as the number of training rounds increases, as shown in the following equation.

α＝0.95^{iteration_num}α₀，

Wherein iteration _ num is the iteration number.

The agent's value at the current state is the value at which all possible discounts to future are present. This allows an agent to focus on possible future turns, but needs to focus more on current turns, since the longer the time, the less accurate the prediction of the future. The discount coefficient of the present application is taken to be 0.8.

In the above embodiment, the evaluation result is obtained by evaluating the signal timing of the intersection detection information and the signal period, so that the calculation complexity is reduced, the intersection delay can be accurately estimated, the overall operation efficiency and safety of the intersection are improved, the accuracy of the evaluation result is improved, the traffic disorder is avoided, and the intelligent traffic effect is achieved.

Optionally, as an embodiment of the present invention, the signal period includes a period duration, and the process of step S1 includes:

calculating the trunk road number set, the branch road number set, the signal period, the inlet incoming traffic flow and the signal lamp time of the inlet vehicle saturated flow through a fifth formula to obtain the green light time of the trunk road, the branch road green light time and the intersection yellow light time, wherein the fifth formula is as follows:

wherein the content of the first and second substances,

Y＝YA+YB，

wherein GA is the green light time of the trunk road, GB is the green light time of the branch road, H is the yellow light time of the intersection, Y is the traffic flow ratio of the intersection, C is the period duration, YA is the sum of the traffic flow of the trunk road and the saturated flow ratio, YB is the sum of the traffic flow of the branch road and the saturated flow ratio, m is_jThe flow of incoming vehicles at the jth inlet, f_jAnd the saturated flow of the jth inlet vehicle, wherein A is a main road number set, and B is a branch road number set.

In the above embodiment, the trunk road number set, the branch road number set, the signal period, the signal lamp time of the incoming vehicle flows and the saturated vehicle flows of the imported vehicles are calculated by the fifth formula to obtain the green light time of the trunk road, the green light time of the branch road and the yellow light time of the intersection, so that the calculation complexity is reduced, the intersection delay can be accurately estimated, the overall operation efficiency and safety of the intersection are improved, the accuracy of the evaluation result is improved, the traffic disorder is avoided, and the intelligent traffic effect is achieved.

Optionally, as an embodiment of the present invention, in step S6, the calculating of the evaluation matrix for the total intersection delay value, the total intersection collision count, the total intersection delay value at the next time, and the total intersection collision count at the next time, and the obtaining of the evaluation matrix includes:

calculating an evaluation matrix for the total delay value of the intersection, the total number of conflicts of the intersection, the total delay value of the intersection at the next moment and the total number of conflicts of the intersection at the next moment through a sixth formula to obtain the evaluation matrix, wherein the sixth formula is as follows:

wherein R(s)_t+1,a_t+1) For the evaluation matrix, D (t) is the total intersection delay value when the branch green light time is t, E (t) is the total intersection conflict number when the branch green light time is t, E (t +1) is the total intersection delay value when the branch green light time is t +1, D (t +1) is the total intersection delay value when the branch green light time is t +1, and rho is the eccentricity parameter.

It should be understood that if agent focuses more on safety aspects, then ρ <0.5, and conversely, if efficiency of operation is more focused, then ρ >0.5, the present invention considers efficiency and safety equally important, so ρ 0.5 is appropriate.

In the above embodiment, the evaluation matrix of the total delay value of the intersection, the total number of conflicts of the intersection, the total delay value of the intersection at the next moment and the total number of conflicts of the intersection at the next moment is calculated by the sixth formula to obtain the evaluation matrix, so that the calculation complexity is reduced, the intersection delay can be accurately estimated, the overall operation efficiency and safety of the intersection are improved, the accuracy of the evaluation result is improved, the traffic disorder is avoided, and the intelligent traffic effect is achieved.

Alternatively, as an embodiment of the present invention, when calculating, it may happen that the branch green time is too short. Therefore, a minimum green time is required to ensure the traffic of the vehicles and pedestrians in the branch road, as shown in the following formula:

it can be seen that the minimum green light duration G_minThe time (road width L) of crossing the street with the pedestrian_PPedestrian crossing speed v_P) And green light loss time. When the calculated green light duration is less than G_minTime, make the duration of green light equal to G_min。

Optionally, as an embodiment of the invention, the invention is directed to a main branch intersection, and the intersection of the karyork city and the vegetable street city in Harbin city is selected as an object, and the geometric structure and the traffic running condition of the intersection are investigated. The east-west inlet of the intersection is a branch road, and the south-north inlet is a main road. The east import is two-way lanes, and the south import is four-way lanes. The branch road is located the residential block, and the traffic fluctuation is obvious. The west inlet is a single-way passage and can only go out but not go in, so the description thereof is not repeated. The specific parameters are shown in table two, which is the specific parameters of traffic operation.

Table two:

Approach	East	South	North
				Road grade	Branch	Arterial	Arterial
Design speed	40km/h	60km/h	60km/h
				Road width	6.75m	16m	16m
Number of lanes	1	4	4
				Lane width	3.75m	3.75m	3.75m
Approach length	111m	96m	80m

one of key links in the VISSIM simulation process during parameter calibration. Correct parameter calibration is important in order to obtain a true and accurate simulation result.

The designed speed of the main trunk is 60km/h, and the designed speed of the branch is 40 km/h. In actual observation, the vehicle speed is mostly distributed between 25km/h and 32 km/h. The invention takes the 85 th vehicle speed as the lower limit of the expected vehicle speed, and takes the maximum design vehicle speed as the upper limit of the expected vehicle speed, namely the expected vehicle speed is [43,60].

Calibration of the following model is also important. Some default values in VISSIM are not in accordance with the actual situation at the intersection and therefore need to be adjusted according to the investigation situation. The following model selects Wiedemann 74 which is more suitable for urban roads. The weather conditions are good on the observation day, and 3 vehicles in front of the observation are suitable to be selected. At the intersection, the speed of the car is low in the following process. After the following state is counted, the average value of the locomotive headway is 1.4 s. The chinese driving rule is to drive right, so a left-side overtaking is provided in the following model.

The branch (east import) has only one import lane, so the position of the path decision point has little influence on the traffic flow. However, considering that the branch is located in a residential area, the local residents are familiar with the route, and the traveling purpose is clear, the route decision point is arranged close to the vehicle input point. In contrast, the vehicles on the main road are mostly based on the road indication signs when selecting the driving lanes. The path decision points are thus set at the location of the road sign (the north and south entrances are all about 50m from the stop line).

In the observation process, the position of the conflict point at the intersection is found to be obvious. The conflict point is mainly focused on the left turning direction of the main road traffic flow. In actual observation, the main road is often decelerated when the main road is driven to the left turning direction to the position just before the intersection. To simulate this situation, a deceleration zone is provided at the location of the collision point.

The simulation of the invention needs to compare three schemes, namely fixed timing, simple fuzzy control and the fuzzy control & q-learning provided by the invention. The simulation steps are as follows:

step 1: vehicle input is simulated through Python, and the traffic flow conforms to Poisson distribution.

Step 2: and respectively calculating the signal period of each scheme according to the vehicle input.

Step 3: and calculating the green light time length of each phase of each scheme, and determining a complete signal timing scheme.

Step 4: the aforementioned data was input into the VISSIM for simulation.

The simulated traffic volume of the invention is from traffic survey data of 12 hours of continuous working days. The traffic data is shown in table three, which is a specific parameter of traffic operation.

Table three:

wherein, Note is in "7^a"means 07:00-08:00 am; "341^b"means 341passenger car units (pcu) (i.e., number of standard vehicle equivalents).

Alternatively, as an embodiment of the present invention, the present invention is intended to evaluate the method of the present invention in terms of efficiency, safety, and stability. Basic data such as queuing length, delay and parking number are obtained through the continuous 12-hour simulation of the VISSIM on the target intersection, and then TTC, DRAC, mTTC and r are calculated in the SSAM and Python_DRACAnd an index such as the number of collisions. Then, the standard deviation of average delay (SAD), the standard deviation of average queuing length (SAQL) and the mean value (m) of the standard deviation of average delay are obtained according to the obtained data statistical analysis_SAD) And standard deviation of average queue length (m)_SAQL) Etc. to discuss the stability of the method of the invention.

Alternatively, as an embodiment of the present invention, the present invention can summarize 3 bright spots compared to previous studies: 1) and determining the cycle time by considering the dynamic change of the flow of the branch road and the trunk road and the saturation of the flow of the intersection. The invention proposes that when the ratio of RR to RS is constant, the resulting cycle duration is fixed, while most fuzzy control studies, similar to inductive control, increase the duration of its green light when the main road is dominant, thereby increasing the cycle duration. Meanwhile, the fuzzy control algorithm has strong adaptability, and stable intersection signal control operation can be ensured. Compared with the prior art, the cycle length obtained by calculation is more stable. During peak hours, the present invention uses smaller cycles to accommodate large traffic flows. 2) According to the green light time distribution method, the green light time is distributed by adopting a q-learning algorithm, and traffic safety and operation efficiency are considered simultaneously by the modeling objective function. Many studies based on reinforcement learning are only aimed at improving intersection efficiency. Wherein the operating efficiency is optimized by reducing intersection delays. And the traffic safety is improved by reducing the number of conflicts and the conflict rate. 3) And analyzing the efficiency, safety and stability of the intersection through various indexes. In the current research, indexes for evaluating the optimization effect of the intersection are single. Most of researches are evaluated based on delay, and other researches are based on a plurality of evaluation indexes.

Alternatively, as an embodiment of the present invention, when considering the operation safety of the intersection, it is not sufficient to consider only the number of collisions. Here we consider two indicators DRAC and mTTC, measured in time as the amount of collision. DRAC is defined as: if the speed of the two vehicles with the close following distance is larger than that of the front vehicle, the deceleration rate to the altitude crash is required by the rear vehicle in order not to collide with the front vehicle. The calculation formula is as follows:

TTC is one of the most common time collision models, defined as: at the current moment, the speed of the rear vehicle is greater than that of the front vehicle, if the two vehicles keep the original speed and the original running track unchanged (assuming that a driver does not take evasive measures), the collision will occur at a certain moment, and the time period from the beginning of the collision to the occurrence of the collision is collision time. It is clear from the definition that larger values of TTC are more advantageous for security, while smaller values of DRAC are more advantageous for security. The prior art proposes the concept of mTTC, that is, mTTC is 1/ttc, which is the same as DRAC, and is more safe as smaller, and the calculation formula is as follows:

in addition to DRAC and mTTC, previous studies (Ding et al.2019) also suggested a rate of change r with mTTC_mTTCAnd rate of change r of DRAC_DRACTo reflect traffic safety conditions. r is_mTTCThe calculation formula of (a) is as follows:

r_DRACthe formula of (1) is as follows:

the safety evaluation parameters obtained by simulation in the invention when the peak hour is as shown in table four, and the table four is the safety evaluation parameters.

Table four:

FCQ, the DRAC was reduced by 23.53% and 45.83% compared to FLC and FT, respectively. FCQ the mTTC is reduced by 6.33% and 6.92% over FLC and FT, respectively. Thus, a 3 point conclusion can be reached: 1.r_DRACAnd r_mTTCThe performance of the system is improved, the numerical value is negative because only the peak period is calculated, the traffic flow is intensive in the peak period, and the operation safety is gradually reduced. 2. In conjunction with the peak hour collision rate, the collision rate of FCQ (0.14) decreased by 17.65% and 39.13% with FLC (0.17) and FT (0.23), respectively. Trend with DRAC, mTTC, r_DRAC，r_mTTCAre the same, and thus can be considered asThe modeling method proposed by the present invention is feasible. 3. When the traffic flow advances, the FCQ track is more sparse, the headway is more long, and the safety is improved.

Alternatively, as an embodiment of the invention, the invention proposes to use the mean (m) of the mean delay standard deviation_SAD) And standard deviation of average queue length (m)_SAQL) To show the stability of the respective solutions as the traffic volume increases. m is_SADThe calculation formula of (a) is as follows:

wherein N is_DGTVRepresenting the number of traffic class groups, d_iFor the ith delay value in each group,

as a mean value of the delay, N_jThe number of data in the jth group.

m_SAQLThe calculation formula of (a) is as follows:

wherein q is_iFor the value of the ith queue length in each group,

is the average value of the delays.

Alternatively, as an embodiment of the invention, the invention proposes to use the mean (m) of the mean delay averages_MAD) And average of average queue length averages (m)_MAQL) To show the stability of each solution as the traffic standard deviation increases. m is_MADThe calculation formula of (a) is as follows:

wherein the content of the first and second substances,N_STVis the number of traffic standard deviation groups.

m_MAQLThe calculation formula of (a) is as follows:

optionally, as an embodiment of the present invention, the present invention employs multiple indexes for analysis. Through discussion and analysis, the method provided by the invention is found to be satisfactory in efficiency, safety and stability. All the evaluation indexes are normalized to obtain a radar map, and the smaller each index is, the better each index is. It should be noted that r_DRACAnd r_mTTCThe calculation results are all negative values, so the trend is consistent with other indexes.

The calculation formula is as follows:

wherein the content of the first and second substances,

the value of the evaluation index of the ith method representing the mth evaluation index,

indicates the value of the m index of the i method. m may represent Delay, Queue Length, Number of profiles, DRAC, r_DRAC,mTTC,r_mTTC,m_SAD,m_SAQL,m_MAD,m_MAQLAnd the like. n represents the number of methods.

From the calculation results, FCQ and FLC have obvious optimization effects on efficiency and stability compared with FT, which shows that the fuzzy control algorithm plays a certain role in the optimization of stability and efficiency. The FLC performance is not satisfactory in terms of safety, and the optimization effect of FCQ is still obvious, which shows that the q-learning model based on delay and conflict number adopted by the invention is feasible in terms of efficiency and safety optimization.

Alternatively, as another embodiment of the present invention, as shown in fig. 2, a traffic signal timing evaluation apparatus includes:

Alternatively, another embodiment of the present invention provides a traffic signal timing evaluation apparatus, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and when the processor executes the computer program, the traffic signal timing evaluation method as described above is implemented. The device may be a computer or the like.

Alternatively, another embodiment of the present invention provides a computer-readable storage medium storing a computer program which, when executed by a processor, implements the traffic signal timing evaluation method as described above.

It is noted that, herein, relational terms such as first and second, and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus.

It is clear to those skilled in the art that, for convenience and brevity of description, the specific working processes of the above-described apparatuses and units may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again.

In the several embodiments provided in the present application, it should be understood that the disclosed apparatus and method may be implemented in other ways. For example, the above-described apparatus embodiments are merely illustrative, and for example, a division of a unit is merely a logical division, and an actual implementation may have another division, for example, a plurality of units or components may be combined or integrated into another system, or some features may be omitted, or not executed.

Units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment of the present invention.

In addition, functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.

The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention essentially or partially contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product stored in a storage medium and including instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.

The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.

Claims

1. A traffic signal timing evaluation method is characterized by comprising the following steps:

2. The traffic signal timing evaluation method according to claim 1, wherein the intersection detection information includes the number of vehicles at an intersection in a current cycle and the number of vehicles at an intersection in a previous cycle; the process of carrying out fuzzy control analysis on the intersection detection information to obtain a signal period comprises the following steps:

3. The traffic signal timing evaluation method according to claim 2, wherein the calculating of the saturation flow rate of the number of vehicles at the intersection of the current cycle, and the obtaining of the saturation flow rate includes:

4. The traffic signal timing evaluation method according to claim 2, wherein the step of calculating the flow rate change ratio between the number of vehicles at the intersection in the current cycle and the number of vehicles at the intersection in the previous cycle to obtain the flow rate change ratio comprises:

5. The traffic signal timing evaluation method according to claim 4, wherein the intersection detection information further includes a trunk number set, a branch number set, a plurality of incoming vehicle flows, a plurality of incoming vehicle saturated flows, a plurality of incoming red light durations, a plurality of single-lane traffic volumes, and a plurality of maximum queuing lengths; the process of evaluating the signal timing of the intersection detection information and the signal period to obtain an evaluation result comprises the following steps:

6. The traffic signal timing evaluation method according to claim 5, wherein the signal period includes a period duration, and the process of step S1 includes:

wherein the content of the first and second substances,

Y＝YA+YB，

wherein GA is the green time of the main road, GB is the green time of the branch road, H is the yellow time of the intersection, Y is the traffic flow ratio of the intersection, C is the period duration, YA is the sum of the traffic flow of the main road and the saturation flow ratio, YB is the branchSum of ratio of traffic flow to saturation flow, m_jThe flow of incoming vehicles at the jth inlet, f_jAnd the saturated flow of the jth inlet vehicle, wherein A is a main road number set, and B is a branch road number set.

7. The method for evaluating traffic signal timing according to claim 5, wherein in step S6, the calculation of the evaluation matrix for the total intersection delay value, the total intersection collision count, the total intersection delay value at the next time, and the total intersection collision count at the next time includes:

8. A traffic signal timing evaluation device, comprising:

9. A traffic signal timing evaluation apparatus comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that the processor, when executing the computer program, implements a traffic signal timing evaluation method as claimed in any one of claims 1 to 7.

10. A computer-readable storage medium, in which a computer program is stored which, when being executed by a processor, carries out a method of evaluating a traffic signal timing according to any one of claims 1 to 7.