CN111899534A - Traffic light intelligent control method based on road real-time capacity - Google Patents
Traffic light intelligent control method based on road real-time capacity Download PDFInfo
- Publication number
- CN111899534A CN111899534A CN202010506114.9A CN202010506114A CN111899534A CN 111899534 A CN111899534 A CN 111899534A CN 202010506114 A CN202010506114 A CN 202010506114A CN 111899534 A CN111899534 A CN 111899534A
- Authority
- CN
- China
- Prior art keywords
- vehicle
- road
- time
- green light
- phase
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/07—Controlling traffic signals
- G08G1/08—Controlling traffic signals according to detected number or speed of vehicles
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/07—Controlling traffic signals
- G08G1/081—Plural intersections under common control
- G08G1/082—Controlling the time between beginning of the same phase of a cycle at adjacent intersections
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/07—Controlling traffic signals
- G08G1/081—Plural intersections under common control
- G08G1/083—Controlling the allocation of time between phases of a cycle
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/09—Arrangements for giving variable traffic instructions
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/09—Arrangements for giving variable traffic instructions
- G08G1/095—Traffic lights
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Traffic Control Systems (AREA)
Abstract
A traffic light intelligent control method based on road real-time capacity comprises the following steps: step 1, calculating phase initial green light duration according to real-time traffic flow information, and optimizing green light duration of each phase by utilizing an algorithm; step 2, judging whether the green light time is reasonable or not, and carrying out fine adjustment to avoid traffic jam caused by overlong green light time of a certain phase; step 3, calculating road flow processing capacity according to the real-time traffic flow information, and calculating phase priority by combining the vehicle queuing length and the vehicle average waiting time; step 4, applying the phase green light duration and the phase sequence to the traffic lights; the method has the characteristic of solving the problem of road congestion by reasonably utilizing the real-time capacity of the adjacent roads.
Description
Technical Field
The invention belongs to the technical field of traffic communication, and particularly relates to an intelligent traffic light control method based on road real-time capacity.
Background
According to the statistical data of the department of commerce, the number of the Chinese automobiles in 2019 is 2.5 hundred million. Originally, people invented automobiles to bring convenience to their own travel and life, but with the rapid increase of the number of vehicles, traffic congestion is increasingly serious, and inconvenience is brought to people to travel under some conditions, and meanwhile, social problems such as frequent traffic accidents, serious air pollution, traffic congestion and the like are caused, so that the comfort level of people to travel is influenced, the environment is polluted, and the development of cities is hindered. There is therefore a need for a communication technique that alleviates the above problems. Researches find that the realization of efficient traffic signal lamp control at traffic intersections is one of important effective means for relieving urban congestion and improving urban traffic operation efficiency. In recent years, the development of technologies such as car networking technology, 5G research and automatic driving assists the development of intelligent transportation.
Intelligent traffic lights are one of the important components in achieving intelligent traffic. At present, there are three common control methods for traffic lights: timing control, inductive control and adaptive control. The first two methods cannot well adjust the traffic lights according to the real-time traffic flow, and are difficult to meet the increasing traffic demands. The self-adaptive traffic light control can timely and dynamically adjust the phase and the time of the traffic light through real-time traffic flow information, and can effectively relieve traffic jam. Therefore, the adaptive traffic light becomes a research hotspot of traffic light control.
Adaptive traffic light control is often combined with various algorithms such as fuzzy logic control, genetic algorithms, neural networks, computer vision, reinforcement learning, and the like. In fact, relieving traffic congestion at a crossing requires coordination between adjacent crossings, which can lead to rapid diffusion of congestion if coordination is not reasonable. Most of the existing adaptive traffic light control methods consider the queuing length and do not consider the influence of the residual capacity of roads on adjacent intersections.
Disclosure of Invention
In order to overcome the defects of the prior art, the invention aims to provide an intelligent traffic light control method based on road real-time capacity, which has the characteristic of solving the problem of road congestion by reasonably utilizing the real-time capacity of adjacent roads.
In order to achieve the purpose, the invention adopts the technical scheme that the traffic light intelligent control method based on the real-time capacity of the road comprises the following steps:
step 1: calculating the initial green light duration of the phase according to the real-time traffic flow information, and optimizing the green light duration of each phase by utilizing an algorithm;
step 2: judging whether the green light time is reasonable or not, and carrying out fine adjustment to avoid traffic jam caused by overlong green light time of a certain phase;
and step 3: calculating road flow processing capacity according to the real-time traffic flow information, and calculating phase priority by combining the vehicle queuing length and the vehicle average waiting time;
and 4, step 4: the phase green light duration and the phase sequence are applied to the traffic light.
The step 1 is specifically defined as follows: the vehicle arrives at the intersection through the Poisson flow with lambda as a parameter; the vehicle can know the speed v and the current position of the vehicle; for convenience of expression, setting the green light duration as a seconds after the initial green light duration takes action; the time for the first vehicle in each lane which is not queued to reach the vehicle row forming the vehicle row is t, if t > a indicates that the vehicle part without parking waiting has no profit and no penalty,
the initial green light duration of each phase is calculated in advance according to the vehicle queuing length,
OriginalGreenTimei,s(k)＝LPQI,s(k)×η+ (1)
wherein LPQi,s(k) Representing the length of the vehicle queue, η is the time required for the vehicle to pass through the intersection, is the amount of compensation for vehicle start-up delay,
and a state S: the position and the speed of the vehicle in the lane are formed, and each road is close to a 225m road part of the stop line and is divided into small grids every 7.5 m; according to the real-time vehicle queuing information, the grid marked with the vehicle in the corresponding road vehicle position matrix is 1, and the grid without the vehicle is 0; if a vehicle crosses two grids, the grid where the vehicle head is located is regarded as 1, and the other grid is regarded as 0;
behavior A: there are three control actions for traffic lights, as follows:
reward r: the reward is composed of two parts: the vehicle traffic volume in each phase green light time is used as a profit, the vehicle average waiting time is used as a penalty, and the two jointly determine the reward, which is shown as follows:
a. traffic light is about to change from red light to green light
The first vehicle in the lane without queue calculates the time t for the vehicle to reach the vehicle row according to the self information, if t < a indicates that the vehicle can pass through the intersection within the next green light time of a seconds, the traffic volume of the vehicle in the lane under the a-time strategy is as follows:
wherein NUMvehicleIndicating the number of vehicles already in line; lambda represents that the vehicle arrives at the intersection by using the Poisson flow with lambda as a parameter; a is the green time after the initial green time takes action;
b. the traffic light is changed from green light to red light
The first vehicle in the lane which is not queued calculates the time t for the first vehicle to arrive at the vehicle row according to the self information; if t < a, the waiting time of the vehicle is a-t, and the rear vehicle arrives at the intersection by using the Poisson flow with lambda as a parameter; if a vehicle arrives at time p (p is more than or equal to 0 and less than or equal to a-t) and starts to wait, an Agent is punished due to the waiting of the vehicle, the punishment degree is defined by combining the waiting duration with a constant coefficient k,
the Agent is penalized by the vehicle waiting time for that road as:
wherein NUMvehicleIndicating the number of vehicles already in line;
in summary, the reward obtained by an Agent at intersection i taking action in state s is shown as the following formula:
ri＝α*rewardi+β*[rewardiA+rewardiB+rewardiC+rewardiD](6)
wherein, rewardiReward, indicating intersection iiA、rewardiB、rewardiCAnd rewardiDRespectively representing the rewards of four intersections adjacent to the intersection i, wherein the constant coefficients of alpha and beta satisfy the formula (7) and are rewardediSpecifically, as shown in the formula (8), rewardiA、rewardiB、rewardiCAnd rewardiDSimilarly thereto:
α+β＝1 (7)
wherein rewardNS(si,ai) Reward for indicating a change from north to south to greenWE(si,ai) The reward for turning green light from east and west direction is expressed by the following specific components in formula (9) and formula (10):
further normalizing the formula such that | r is satisfiediL is less than or equal to 1, so the reward r is:
in short, the system inputs the current state and outputs various behavior Q values in the state,
Qt+1(st,at)＝Qt(st,at)+αt[rt+γmaxaQt+1(st+1,a+1)-Qt(st,at)](12)
the algorithm calculates the Q value of each action according to a formula (12), the system selects the action with the maximum Q value or randomly selects the action according to an action selection strategy, and the green light duration corresponding to the phase is optimized according to the adopted action.
The step 2 specifically comprises the following steps: if the optimized green light time is longer than the phase maximum green light time, the green light time is reduced to the phase maximum green light time, and if the optimized green light time is shorter than or equal to the phase maximum green light time, the green light time does not need to be finely adjusted.
The step 3 specifically comprises the following steps:
setting: in one period, each phase of an intersection has green light time at most once; the time required for each vehicle to pass through the intersection is 2 s;
the road length of the road (i, j) is Mij；blanki,j(k) A road length at which the road (i, j) at the start of the kth cycle has no vehicles; leave ei,j(k) Is the length of the vehicle leaving the road (i, j) in the k-th cycle; remaini,j(k) Queuing length of vehicles which can not leave the road in the road (i, j) at the end of the k-1 th period;
road residual capacity Cij(k) Number of new cars that can be accommodated by the road (i, j) in the k-th cycle, i.e. number of new cars
Cij(k)＝blanki,j(k)+leavei,j(k) (13)
RTPC (Road Traffic Processing Capability, RTPC for short, Chinese means Road Traffic Processing Capability) can be obtained by using the formula (13), i.e. the difference between the Road residual capacity and the input demand is used as one of the influence factors of the phase sequence,
RTPCi,j(k+1)＝Ci,j(k)-Li,leave(k) (14)
Li,leave(k)＝Li1,leave(k)+Li2,leave(k)+Li3,leave(k) (15)
wherein L isi,leave(k) Indicating the length of a vehicle intended to enter a road (i, j) from an intersection i in the k-th cycle, as shown in the red portion in fig. 1(b), RTPCi,j(k +1) represents the road flow handling capacity of the road (i, j) at the beginning of the k +1 th cycle;
RTPC > 0 shows that the residual capacity of the road can meet the input requirement of the road, and vehicles entering the road can be increased; RTPC <0 indicates that the road residual capacity cannot meet the road input requirement, and a vehicle from an intersection i cannot completely enter a road (i, j) and cause congestion, so that it is necessary to increase vehicles leaving the road (i, j) and decrease vehicles entering the road so as to increase the RTPC of the road (i, j), and therefore, if the RTPC of the current road is to be improved, the RTPC of an adjacent road needs to be sacrificed to realize the local balance of the RTPC, that is, the vehicles leaving the road (i, j) are increased, the road residual capacity of a downstream road is decreased, and the RTPC is reduced; decreasing vehicles entering the road (i, j), decreasing the output of vehicles on the upstream road, decreasing their RTPCs;
the phase priority is divided into three parts: the RTPC priority, the vehicle queue length priority and the vehicle average waiting time priority, and the priority of each road in each period is calculated by the following formula:
Priorityi,s(k)＝γ×PRi,s(k)+μ×PQi,s(k)+σ×AWi,s(k) (16)
γ+μ+σ＝1 (17)
wherein Priority isi,s(k) Priority, indicating the phase s of the k-th cycle crossing ii,s(k) Smaller priority higher PRi,s(k) An RTPC priority representing the phase s of the kth cycle crossing i; PQi,s(k) A vehicle queue length priority representing a phase s of a kth cycle intersection i; AWi,s(k) The average waiting time priority of the vehicle of the phase s of the k-th cycle intersection i is represented, gamma, mu and sigma are constant coefficients, and the value range is [0, 1%]And satisfies equation 17.
The Road Traffic Processing Capability English is abbreviated as RTPC, and the Chinese meaning is as follows: road traffic handling capability.
The invention has the beneficial effects that:
the influence of the real-time capacity of the road on the phase sequence of the traffic lights is considered as well as the queuing length and the average waiting time of the vehicles, the traffic volume is considered as reward and the average waiting time of the vehicles is considered as punishment when the reward of the DQN algorithm is calculated, and the reward is obtained by combining the traffic volume with the average waiting time of the vehicles. By the methods, the road throughput is improved, and the average waiting time of vehicles is reduced.
Drawings
Fig. 1 is a road remaining capacity map and a road vehicle accommodation map of the present invention.
Fig. 2 is a graph of the average throughput of the present invention.
FIG. 3 is a graph of total vehicle delay time in accordance with the present invention.
FIG. 4 is a graph of the average queue length of vehicles according to the present invention.
FIG. 5 is a graph of the effect of phase priority on the average queue length of vehicles according to the present invention.
Detailed Description
The present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.
The invention relates to a traffic light intelligent control research based on road real-time capacity, which comprises the following steps:
step one, optimizing the duration of green lights of each phase according to a DQN algorithm and real-time traffic flow information;
for ease of problem analysis, we agree on the following assumptions and definitions:
1) the vehicle arrives at the intersection through the Poisson flow with lambda as a parameter;
2) the vehicle knows the speed v and the current position of the vehicle;
3) for convenience, the green time after the action is taken for the initial green time is set to a seconds,
4) the vehicle arrives at the intersection through the poisson flow with lambda as a parameter, and the probability density function of the poisson flow with respect to time omega is as follows: the formula (18) is an existing formula, is assumed as a precondition of patent and does not belong to the specific content of the invention, and is listed here because the related calculation formula of the subsequent reward is used;
the initial green light duration of each phase is calculated in advance according to the vehicle queuing length,
OriginalGreenTimei,s(k)＝LPQI,s(k)×η+ (1)
wherein LPQi,s(k) The vehicle queuing length is shown, eta is the time required by the vehicle to pass through the intersection, and is the compensation quantity for the vehicle starting delay;
and a state S: the position and speed of the vehicle in the lane. Each 225m road part of each road, which is close to the stop line, is divided into small grids every 7.5 m; according to the real-time vehicle queuing information, the grid marked with the vehicle in the corresponding road vehicle position matrix is 1, and the grid without the vehicle is 0; if a vehicle crosses two grids, the grid where the vehicle head is located is regarded as 1, and the other grid is regarded as 0;
behavior A: there are three control actions of the traffic light for the initial green light duration as follows:
reward r: the reward is composed of two parts: the vehicle traffic volume in each phase green light time is used as a profit, the vehicle average waiting time is used as a penalty, and the two jointly determine the reward, which is shown as follows:
a. traffic light is about to change from red light to green light
The first vehicle in the lane without queue calculates the time t for the vehicle to reach the vehicle row according to the self information, if t < a indicates that the vehicle can pass through the intersection within the next green light time of a seconds, the traffic volume of the vehicle in the lane under the a-time strategy is as follows:
whereinNUMvehicleIndicating the number of vehicles already in line; lambda represents that the vehicle arrives at the intersection by using the Poisson flow with lambda as a parameter; a is the green time after the initial green time takes action;
b. the traffic light is changed from green light to red light
The first vehicle in the lane which is not queued calculates the time t for the first vehicle to arrive at the vehicle row according to the self information; if t < a, the waiting time of the vehicle is a-t, and the rear vehicle arrives at the intersection by using the Poisson flow with lambda as a parameter; if a vehicle arrives at time p (p is more than or equal to 0 and less than or equal to a-t) and starts waiting, an Agent receives punishment due to vehicle waiting, and the punishment degree is defined by combining the waiting duration with a constant coefficient k
The Agent is penalized as the vehicle waiting time of the road
In summary, the reward obtained by an Agent at intersection i taking action in state s is shown as the following formula:
ri＝α*rewardi+β*[rewardiA+rewardiB+rewardiC+rewardiD](6)
wherein, rewardiReward, indicating intersection iiA、rewardiB、rewardiCAnd rewardiDRespectively representing the rewards of four intersections adjacent to the intersection i, wherein the constant coefficients of alpha and beta satisfy the formula (7) and are rewardediSpecifically, as shown in the formula (8), rewardiA、rewardiB、rewardiCAnd rewardiDSimilarly thereto:
α+β＝1 (7)
wherein rewardNS(si,ai) Reward for indicating a change from north to south to greenWE(si,ai) The reward for turning green light from east and west direction is expressed by the following specific components in formula (9) and formula (10):
further normalizing the formula such that | r is satisfiediLess than 1, so the reward r is
In summary, the system inputs the current state, outputs the respective behavior Q values in the state,
Qt+1(st,at)＝Qt(st,at)+αt[rt+γmaxaQt+1(st+1,a+1)-Qt(st,at)](12)
calculating the Q value of each action according to a formula (12), selecting the action with the maximum Q value or randomly selecting the action by the system according to an action selection strategy, and optimizing the green light duration corresponding to the phase according to the adopted action;
step two, judging whether the green light time is reasonable or not, and carrying out fine adjustment to avoid that the waiting vehicles in other phases are too many because a certain phase has too long green light time;
if the optimized green light time is longer than the phase maximum green light time, the green light time is reduced to the phase maximum green light time, and if the optimized green light time is shorter than or equal to the phase maximum green light time, the green light time does not need to be finely adjusted;
calculating road flow processing capacity according to the real-time traffic flow information, and calculating phase priority by combining the vehicle queuing length and the vehicle average waiting time;
for ease of analysis of the problem, we agree on the following assumptions:
1) in one period, each phase of an intersection has green light time at most once;
2) the time required for each vehicle to pass through the intersection is eta;
as can be seen from FIG. 1, the road length of the road (i, j) is Mij；blanki,j(k) A road length at which the road (i, j) at the start of the kth cycle has no vehicles; leave ei,j(k) Is the length of the vehicle leaving the road (i, j) in the k-th cycle; remaini,j(k) Queuing length of vehicles which can not leave the road in the road (i, j) at the end of the k-1 th period;
road residual capacity Cij(k) Number of new cars that can be accommodated by the road (i, j) in the k-th cycle, i.e. number of new cars
Cij(k)＝blanki,j(k)+leavei,j(k) (13)
Road Traffic handling capacity (RTPC) can be obtained using equation (13), i.e., the difference between the Road remaining capacity and the input demand is used as one of the influencing factors of the phase sequence,
RTPCi,j(k+1)＝Ci,j(k)-Li,leave(k) (14)
Li,leave(k)＝Li1,leave(k)+Li2,leave(k)+Li3,leave(k) (15)
wherein L isi,leave(k) Indicating the length of a vehicle intended to enter a road (i, j) from an intersection i in the k-th cycle, as shown in the red portion in fig. 1(b), RTPCi,j(k +1) represents the road flow handling capacity of the road (i, j) at the beginning of the k +1 th cycle;
RTPC > 0 shows that the residual capacity of the road can meet the input requirement of the road, and vehicles entering the road can be increased properly; RTPC <0 indicates that the road residual capacity cannot meet the road input requirement, that is, taking fig. 1(a) as an example, a vehicle from an intersection i cannot completely enter a road (i, j) and congestion is caused, so that it is necessary to increase vehicles leaving the road (i, j) and decrease vehicles entering the road so as to increase the RTPC of the road (i, j), and therefore, if the RTPC of the current road is to be improved, the RTPC of an adjacent road needs to be sacrificed to realize the local balance of the RTPC, that is, increasing vehicles leaving the road (i, j) means that the road residual capacity of a downstream road is decreased so as to decrease the RTPC thereof; reducing vehicles entering the road (i, j) means reducing the output of upstream road vehicles lowering their RTPCs;
the phase priority is divided into three parts: the RTPC priority, the vehicle queue length priority and the vehicle average waiting time priority, and the priority of each road in each period is calculated by the following formula:
Priorityi,s(k)＝γ×PRi,s(k)+μ×PQi,s(k)+σ×AWi,s(k) (16)
γ+μ+σ＝1 (17)
wherein Priority isi,s(k) Priority, indicating the phase s of the k-th cycle crossing ii,s(k) Smaller priority higher PRi,s(k) An RTPC priority representing the phase s of the kth cycle crossing i; PQi,s(k) A vehicle queue length priority representing a phase s of a kth cycle intersection i; AWi,s(k) The average waiting time priority of the vehicle of the phase s of the k-th cycle intersection i is represented, gamma, mu and sigma are constant coefficients, and the value range is [0, 1%]And satisfies equation 17;
and step four, optimizing the initial green light time obtained by calculation according to the vehicle queuing length by using an algorithm, and comparing the initial green light time with the maximum green light time to obtain the green light time after each phase is optimized. And sequentially calculating three priorities according to the traffic flow information, and finally determining the priority sequence of each phase together. And then, corresponding to the green light time of each phase according to the phase priority order to form a traffic light control scheme in the next period, and applying the traffic light control scheme to traffic light control.
The basic idea is as follows: firstly, the vehicle queue length is obtained according to the real-time traffic flow information, and the initial green light duration is calculated, so that vehicles which are moving in a certain range need to be considered when the phase green light duration is calculated, considering that new vehicles may still arrive on roads corresponding to other phases and queue for waiting when a certain group of phases have the right of way. The vehicle distribution in a certain range of the intersection is converted into a vehicle distribution matrix, the number of queued vehicles on each road and the number of vehicles which are not queued on the first road are concerned, relevant reward calculation division is made, and the vehicle distribution state is input to output the further optimized green light duration.
Then, it is considered that the long green time of a single phase group increases the waiting time of vehicles and the queuing length of other phase groups, which causes traffic jam and reduces traffic throughput. Therefore, the maximum green light time of the phase is set to avoid that a certain phase group occupies too much right of way, so that the vehicles of other phase groups are hindered from passing, the fine adjustment of the green light time is realized, and the final phase green light time is obtained.
And finally, obtaining the traffic processing capacity of each road through related calculation according to the real-time traffic flow information. And then jointly determining the traffic light phase sequence of the next period according to the RTPC priority, the vehicle queue length priority and the vehicle average waiting time priority.
In order to verify the feasibility and the effect of the mechanism, a combined simulation platform is built by using VISSIM and Python for simulation. The VISSIM is a microscopic traffic simulation modeling tool based on time intervals and driving behaviors, can analyze traffic running conditions of an urban road network under various traffic conditions, and can effectively evaluate traffic schemes.
The control of the traffic light at a single intersection does not consider right turn and does not consider pedestrians in the experimental simulation. The lane width is set to be 3.5m, the distance between traffic intersections is set to be 500m, the saturated traffic flow of each lane is 2000veh/h, the simulation time is in 3600s, and the simulation precision is 10 time steps. Set to 0.3, α is 0.002 and γ is 0.95. The traffic flow is calculated every five time steps.
FIG. 2 is a schematic diagram of vehicle throughput for three modes of time-phased control, arterial road control (ATL) and RTPC control. It can be seen that the average throughput of RTPC is overall better than the other two methods; the time-interval control is the worst effect, and more than 200 vehicles can be left behind; ATL may cause vehicle stagnation around 400 vehicles. When the number of vehicles reaches 1600, the turndown capability of the split-time control and the ATL control gradually saturates, while the RTPC is still increasing the throughput. Therefore, compared with other two control modes, the RTPC can improve the crossing throughput.
FIG. 3 shows the total delay time of the vehicle in three modes of time-share control, arterial road control (ATL) and RTPC control. The abscissa is the number of vehicles per hour and the ordinate is the total delay time of the vehicle. It can be seen that the effects of the time-division control, the ATL control and the RTPC control are all good when the number of vehicles is less than 200, but the time-division control is delayed by the vehicles from when the number of vehicles is more than 200, and the performance is the least good; the ATL starts from about 400 to cause vehicle delay and as the number of vehicles increases, the vehicle delay time increases rapidly, and the effect of the later period is not ideal. This is because the ATL substantially sacrifices the performance of the non-main road to ensure the performance of the main road as much as possible, but if the traffic flow increases, when the flow of the main road and the non-main road is not much different, the performance is the same as the traffic light control with fixed duration; the RTPC starts vehicle delay around 780, which is the best performance of the two, and has better performance than the former two, although vehicle delay is increased with the increase of the number of vehicles.
Fig. 4 is the average vehicle queue length under the time-share control, the ATL control, and the RTPC control, the abscissa is the number of vehicles per hour, and the ordinate is the average queue length per road. It can be seen that the average queue length of the three modes is almost the same when the number of vehicles is small, but the average queue length of the time-sharing control is the fastest to grow as the number of vehicles increases, followed by the ATL method, and the RTPC is the best in performance. And the larger the number of vehicles is, the more similar the ATL and the average queuing length controlled by time-sharing, because the traffic flow is increased, when the flow of the main road and the flow of the non-main road are not greatly different, the performance of the traffic light is the same as that of the traffic light controlled by fixed time-sharing.
Fig. 5 is a comparison of a dynamically changing phase sequence with consideration of phase priority with a fixed phase sequence in an RTPC control mode. It can be seen that considering the phase priority change phase sequence at a single intersection is better than the fixed phase sequence method in the performance of the average queuing length of vehicles.
Claims (4)
1. A traffic light intelligent control method based on road real-time capacity is characterized by comprising the following steps:
step 1: calculating the initial green light duration of the phase according to the real-time traffic flow information, and optimizing the green light duration of each phase by utilizing an algorithm;
step 2: judging whether the green light time is reasonable or not, and carrying out fine adjustment to avoid traffic jam caused by overlong green light time of a certain phase;
and step 3: calculating road flow processing capacity according to the real-time traffic flow information, and calculating phase priority by combining the vehicle queuing length and the vehicle average waiting time;
and 4, step 4: the phase green light duration and the phase sequence are applied to the traffic light.
2. The intelligent traffic light control method based on real-time road capacity according to claim 1, wherein the step 1 is implemented by defining: the vehicle arrives at the intersection through the Poisson flow with lambda as a parameter; the vehicle can know the speed v and the current position of the vehicle; for convenience of expression, setting the green light duration as a seconds after the initial green light duration takes action; the time for the first vehicle in each lane to arrive at the vehicle row is t, if t > a indicates that the vehicle part without parking waiting has no profit and no penalty,
the initial green light duration of each phase is calculated in advance according to the vehicle queuing length,
OriginalGreenTimei,s(k)＝LPQI,s(k)×η+ (1)
wherein LPQi,s(k) The vehicle queuing length is shown, eta is the time required by the vehicle to pass through the intersection, and is the compensation quantity for the vehicle starting delay;
and a state S: the position and speed of the vehicles in the lane are formed, the distribution of the vehicles on the 225m road of each road close to the stop line and the corresponding vehicle speed are taken as the states, and each 7.5m is divided into a small grid; according to the real-time vehicle queuing information, the grid marked with the vehicle in the corresponding road vehicle position matrix is 1, and the grid without the vehicle is 0; if a vehicle crosses two grids, the grid where the vehicle head is located is regarded as 1, and the other grid is regarded as 0;
behavior A: there are three control actions for traffic lights, as follows:
reward r: the reward is composed of two parts: the vehicle traffic volume in each phase green light time is used as a profit, the vehicle average waiting time is used as a penalty, and the two jointly determine the reward, which is shown as follows:
a. traffic light is about to change from red light to green light
The first vehicle in the lane without queue calculates the time t for the vehicle to reach the vehicle row according to the self information, if t < a indicates that the vehicle can pass through the intersection within the next green light time of a seconds, the traffic volume of the vehicle in the lane under the a-time strategy is as follows:
wherein NUMvehicleIndicating the number of vehicles already in line; lambda represents that the vehicle arrives at the intersection by using the Poisson flow with lambda as a parameter; a is the green time after the initial green time takes action;
b. the traffic light is changed from green light to red light
The first vehicle in the lane which is not queued calculates the time t for the first vehicle to arrive at the vehicle row according to the self information; if t < a, the waiting time of the vehicle is a-t, and the rear vehicle arrives at the intersection by using the Poisson flow with lambda as a parameter; if a vehicle arrives and starts to wait at time p (p is more than or equal to 0 and less than or equal to a-t), the intelligent agent receives punishment due to the waiting of the vehicle, and defines punishment degree by combining waiting duration with a constant coefficient k,
the agent is penalized for the vehicle waiting time of the road as:
in summary, the reward obtained by the agent at intersection i taking action in state s is shown in the following formula:
ri＝α*rewardi+β*[rewardiA+rewardiB+rewardiC+rewardiD](6)
wherein, rewardiReward, indicating intersection iiA、rewardiB、rewardiCAnd rewardiDRespectively representing the rewards of four intersections adjacent to the intersection i, wherein the constant coefficients of alpha and beta satisfy the formula (7) and are rewardediSpecifically, as shown in the formula (8), rewardiA、rewardiB、rewardiCAnd rewardiDSimilarly thereto:
α+β＝1 (7)
wherein rewardNS(si，ai) Reward for indicating a change from north to south to greenWE(si，ai) The reward for turning green light from east and west direction is expressed by the following specific components in formula (9) and formula (10):
further normalizing the formula such that | r is satisfiediL is less than or equal to 1, so the reward r is:
in short, the system inputs the current state and outputs various behavior Q values in the state,
Qt+1(st，at)＝Qt(st，at)+αt[rt+γmaxaQt+1(st+1，a+1)-Qt(st，at)](12)
the algorithm calculates the Q value of each action according to a formula (12), the system selects the action with the maximum Q value or randomly selects the action according to an action selection strategy, and the green light duration corresponding to the phase is optimized according to the adopted action.
3. The intelligent traffic light control method based on real-time road capacity according to claim 1, wherein the step 2 specifically comprises: if the optimized green light time is longer than the phase maximum green light time, the green light time is reduced to the phase maximum green light time, and if the optimized green light time is shorter than or equal to the phase maximum green light time, the green light time does not need to be finely adjusted.
4. The intelligent traffic light control method based on real-time road capacity according to claim 1, wherein the step 3 is specifically implemented as follows:
setting: in one period, each phase of an intersection has green light time at most once; the time required for each vehicle to pass through the intersection is eta;
the road length of the road (i, j) is Mij；blanki，j(k) Is a road on which the road (i, j) at the beginning of the k-th cycle has no vehiclesThe length of the road; leave ei，j(k) Is the length of the vehicle leaving the road (i, j) in the k-th cycle; remaini，j(k) Queuing length of vehicles which can not leave the road in the road (i, j) at the end of the k-1 th period;
road residual capacity Cij(k) Number of new cars that can be accommodated by the road (i, j) in the k-th cycle, i.e. number of new cars
Cij(k)＝blanki，j(k)+leavei，j(k) (13)
RTPC can be obtained by using the formula (13), i.e. the difference between the road residual capacity and the input requirement is taken as one of the influence factors of the phase sequence,
RTPCi，j(k+1)＝Ci，j(k)-Li，leave(k) (14)
Li，leave(k)＝Li1，leave(k)+Li2，leave(k)+Li3，leave(k) (15)
wherein L isi，leave(k) Indicating the length of a vehicle, RTPC, intended to enter a road (i, j) from intersection i in the k-th cyclei，j(k +1) represents the road flow handling capacity of the road (i, j) at the beginning of the k +1 th cycle;
RTPC > 0 shows that the residual capacity of the road can meet the input requirement of the road, and vehicles entering the road can be increased; the RTPC <0 indicates that the residual capacity of the road cannot meet the road input requirement, and the situation that vehicles from the intersection i cannot completely enter the road (i, j) can cause congestion, so that vehicles leaving the road (i, j) need to be increased, vehicles entering the road need to be reduced, the RTPC of the road (i, j) needs to be increased, and therefore, if the RTPC of the current road needs to be improved, the RTPC of an adjacent road needs to be sacrificed, so that the local balance of the RTPC is realized, namely, vehicles leaving the road (i, j) is increased; reducing vehicles entering the road (i, j);
the phase priority is divided into three parts: the RTPC priority, the vehicle queue length priority and the vehicle average waiting time priority, and the priority of each road in each period is calculated by the following formula:
Priorityi，s(k)＝γ×PRi，s(k)+μ×PQi，s(k)+σ×AWi，s(k) (16)
γ+μ+σ＝1 (17)
wherein Priority isi，s(k) Priority, indicating the phase s of the k-th cycle crossing ii，s(k) Smaller priority higher PRi，s(k) An RTPC priority representing the phase s of the kth cycle crossing i; PQi，s(k) A vehicle queue length priority representing a phase s of a kth cycle intersection i; AWi，s(k) The average waiting time priority of the vehicle of the phase s of the k-th cycle intersection i is represented, gamma, mu and sigma are constant coefficients, and the value range is [0, 1%]And satisfies equation 17.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010506114.9A CN111899534A (en) | 2020-06-05 | 2020-06-05 | Traffic light intelligent control method based on road real-time capacity |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010506114.9A CN111899534A (en) | 2020-06-05 | 2020-06-05 | Traffic light intelligent control method based on road real-time capacity |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111899534A true CN111899534A (en) | 2020-11-06 |
Family
ID=73206622
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010506114.9A Pending CN111899534A (en) | 2020-06-05 | 2020-06-05 | Traffic light intelligent control method based on road real-time capacity |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111899534A (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112767718A (en) * | 2021-01-29 | 2021-05-07 | 浙江大华技术股份有限公司 | Signal lamp control method and device, readable storage medium and electronic equipment |
| CN112927525A (en) * | 2021-05-11 | 2021-06-08 | 华砺智行(武汉)科技有限公司 | Signal lamp phase distribution method and device under intelligent networking environment and storage medium |
| CN112991783A (en) * | 2021-02-05 | 2021-06-18 | 青岛海信网络科技股份有限公司 | Bus priority control method and device based on real-time position of internet public transport |
| CN113470387A (en) * | 2021-07-17 | 2021-10-01 | 广东汇通信息科技股份有限公司 | Intelligent traffic control method and device |
| CN113724507A (en) * | 2021-08-19 | 2021-11-30 | 复旦大学 | Traffic control and vehicle induction cooperation method and system based on deep reinforcement learning |
| CN113936475A (en) * | 2021-08-19 | 2022-01-14 | 复旦大学 | Traffic control method and system for optimizing multi-road resources |
| CN114267188A (en) * | 2021-12-17 | 2022-04-01 | 航天科技控股集团股份有限公司 | Method and system for reducing road vehicle congestion |
| CN114708744A (en) * | 2022-03-22 | 2022-07-05 | 燕山大学 | Vehicle starting optimization control method and device based on fusion traffic information |
| CN115273499A (en) * | 2022-06-30 | 2022-11-01 | 华东师范大学 | Traffic flow-based signal lamp dynamic timing method and system |
| CN115830887A (en) * | 2023-02-14 | 2023-03-21 | 武汉智安交通科技有限公司 | Self-adaptive traffic signal control method, system and readable storage medium |
| CN115953908A (en) * | 2022-12-02 | 2023-04-11 | 山东国泰建设工程有限公司 | Traffic light control method and system and traffic light |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103544840A (en) * | 2013-10-15 | 2014-01-29 | 华南理工大学 | Coordinated control diagrammatic method for urban signalized intersections |
| CN106910351A (en) * | 2017-04-19 | 2017-06-30 | 大连理工大学 | A kind of traffic signals self-adaptation control method based on deeply study |
| CN108492592A (en) * | 2018-03-01 | 2018-09-04 | 西安理工大学 | A kind of traffic light intelligent control system and its control method based on mist calculating |
-
2020
- 2020-06-05 CN CN202010506114.9A patent/CN111899534A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103544840A (en) * | 2013-10-15 | 2014-01-29 | 华南理工大学 | Coordinated control diagrammatic method for urban signalized intersections |
| CN106910351A (en) * | 2017-04-19 | 2017-06-30 | 大连理工大学 | A kind of traffic signals self-adaptation control method based on deeply study |
| CN108492592A (en) * | 2018-03-01 | 2018-09-04 | 西安理工大学 | A kind of traffic light intelligent control system and its control method based on mist calculating |
Non-Patent Citations (3)
| Title |
|---|
| 卓广平: "《严肃游戏中个性化建模及认知机制研究》", 31 August 2018 * |
| 安萌萌，樊秀梅，蔡含宇: "基于雾计算和强化学习的交通灯智能协同控制研究", 《计算机应用研究》 * |
| 安萌萌: "基于强化学习的交通灯智能调控研究", 《中国优秀博硕士学位论文全文数据库(硕士)》 * |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112767718A (en) * | 2021-01-29 | 2021-05-07 | 浙江大华技术股份有限公司 | Signal lamp control method and device, readable storage medium and electronic equipment |
| CN112991783A (en) * | 2021-02-05 | 2021-06-18 | 青岛海信网络科技股份有限公司 | Bus priority control method and device based on real-time position of internet public transport |
| CN112991783B (en) * | 2021-02-05 | 2022-04-29 | 青岛海信网络科技股份有限公司 | Bus priority control method and device based on real-time position of internet public transport |
| CN112927525A (en) * | 2021-05-11 | 2021-06-08 | 华砺智行(武汉)科技有限公司 | Signal lamp phase distribution method and device under intelligent networking environment and storage medium |
| CN113470387A (en) * | 2021-07-17 | 2021-10-01 | 广东汇通信息科技股份有限公司 | Intelligent traffic control method and device |
| CN113724507B (en) * | 2021-08-19 | 2024-01-23 | 复旦大学 | Traffic control and vehicle guidance cooperative method and system based on deep reinforcement learning |
| CN113724507A (en) * | 2021-08-19 | 2021-11-30 | 复旦大学 | Traffic control and vehicle induction cooperation method and system based on deep reinforcement learning |
| CN113936475A (en) * | 2021-08-19 | 2022-01-14 | 复旦大学 | Traffic control method and system for optimizing multi-road resources |
| CN114267188A (en) * | 2021-12-17 | 2022-04-01 | 航天科技控股集团股份有限公司 | Method and system for reducing road vehicle congestion |
| CN114708744A (en) * | 2022-03-22 | 2022-07-05 | 燕山大学 | Vehicle starting optimization control method and device based on fusion traffic information |
| CN114708744B (en) * | 2022-03-22 | 2023-06-09 | 燕山大学 | Vehicle starting optimization control method and device based on fusion traffic information |
| CN115273499B (en) * | 2022-06-30 | 2023-11-10 | 华东师范大学 | Traffic flow-based signal lamp dynamic timing method and system |
| CN115273499A (en) * | 2022-06-30 | 2022-11-01 | 华东师范大学 | Traffic flow-based signal lamp dynamic timing method and system |
| CN115953908A (en) * | 2022-12-02 | 2023-04-11 | 山东国泰建设工程有限公司 | Traffic light control method and system and traffic light |
| CN115953908B (en) * | 2022-12-02 | 2024-03-29 | 山东国泰建设工程有限公司 | Traffic light control method and system and traffic light |
| CN115830887A (en) * | 2023-02-14 | 2023-03-21 | 武汉智安交通科技有限公司 | Self-adaptive traffic signal control method, system and readable storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111899534A (en) | Traffic light intelligent control method based on road real-time capacity | |
| CN111445692B (en) | Speed collaborative optimization method for intelligent networked automobile at signal-lamp-free intersection | |
| CN109785619B (en) | Regional traffic signal coordination optimization control system and control method thereof | |
| CN108510764B (en) | Multi-intersection self-adaptive phase difference coordination control system and method based on Q learning | |
| CN111951549B (en) | Self-adaptive traffic signal lamp control method and system in networked vehicle environment | |
| CN103996289B (en) | A kind of flow-speeds match model and Travel Time Estimation Method and system | |
| CN104809895B (en) | The arterial road coordinate control model and its optimization method of Adjacent Intersections | |
| CN107730922B (en) | Unidirectional trunk line green wave coordination control self-adaptive adjustment method | |
| CN102360522B (en) | Highway optimization control method | |
| CN112037507B (en) | Supersaturated traffic state trunk line adaptive signal coordination design method and device | |
| CN104916142B (en) | A kind of crossing self-adapting traffic signal control method of giving priority to trunk roads | |
| CN112907946A (en) | Traffic control method and system for automatically driving vehicle and other vehicles to run in mixed mode | |
| CN114241751B (en) | Multi-entrance dynamic and static traffic coordination optimization method for large parking lot | |
| CN113312752B (en) | Traffic simulation method and device for main road priority control intersection | |
| CN107293133A (en) | A kind of method for controlling traffic signal lights | |
| WO2023035666A1 (en) | Urban road network traffic light control method based on expected reward estimation | |
| CN104700634A (en) | Adjacent intersection road coordinate control method based on minimum spanning tree clustering improved genetic algorithm | |
| CN115311868B (en) | Main line coordination control method and device based on bus priority | |
| CN113223324A (en) | Control method for high-speed ramp entrance confluence | |
| CN116189454A (en) | Traffic signal control method, device, electronic equipment and storage medium | |
| CN105931474A (en) | City road intersection group local overflow control method with quantum decision | |
| CN115376332A (en) | Self-adaptive traffic signal control method | |
| CN114141013A (en) | Intersection signal timing method based on low-permeability network connection track data | |
| CN115691105B (en) | Maximum-voltage signal control method considering pedestrian crossing of discrete road section of motorcade | |
| CN116013077B (en) | Dynamic control method for expressway special road under hybrid networking environment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20201106 |
|
| RJ01 | Rejection of invention patent application after publication |