Real-Time Traffic Light Scheduling Algorithm Based on Genetic Algorithm and Machine Learning

Abstract

Traffic signals are essential to provide safe driving that allows all traffic flows to share road intersection. However, they decrease the traffic flow fluency because of the queuing delay at each road intersection. In order to improve the traffic efficiency all over the road network, Intelligent Traffic Light Scheduling (ITLS) algorithm has been proposed. In this work, we introduce an ITLS algorithm based on Genetic Algorithm (GA) merging with Machine Learning (ML) algorithm. This algorithm schedules the time phases of each traffic light according to each real-time traffic flow that intends to cross the road intersection, whilst considering next time phases of traffic flow at each intersection by ML. In order to get each next time phases of traffic flow, we use Linear Regression (LR) algorithm as ML algorithm. The introduced algorithm aims to increase traffic fluency by decreasing the total waiting delay of all traveling vehicles at each road intersection in the road network. We compare the performance of our algorithm with the unimproved one for different simulated data. Results shows that, our algorithm increases the traffic fluency and decreases the waiting delay by 21.5 % compared with the unimproved one.

Appendix: ITLSGMLR Scenario

A Scene Simplification

In order to expedite scheduling algorithm modeling, we have to simplify the problem as follows:

• (1) Only red light and green light are taken into consideration, regardless of yellow illumination;

• (2) Driving directions have three conditions which are going straight, turning left or right, without turning around;

• (3) At each intersection, the probability of going straight, turning left or right respectively are α = 0.8, β = 0.1, γ = 0.1, where α + β + γ = 1;

• (4) On Traffic signal location network, we just care about 4-leg intersection or 3-leg intersection;

• (5) T is unit time and the initial time of red and green light is T;

• (6) Traffic light cycle time must be set in multiples of T;

• (7) At each light cycle scheduling, light must switch to another color without exceeding maximum allowable time (4T) for that phase;

• (8) The distance between any two traffic lights is equal, meanwhile, any vehicle just cost T time from one traffic light to another adjacent traffic light.

B Data structure schema

It illustrates the location of traffic light in the road network which can be expressed as table as above. All type of table field are Integer.

Table 4. Traffic_light_table

This table is a dynamic time dependent table. It was shown as above at the moment of t_x. All type of table field are Integer. ‘1’ as green light on, ‘0’ as red light on, ‘-1’ as no light in this direction.

Table 5. Traffic_light_status_table(t _x )

Table 6. Traffic_flow_table

It says traffic flow between two traffic signals during the period in the simulated scenario. The type of table field, TrafficLightID and FromID are Integer. trafficFlow’s type is Array (Table 7).

Table 7. Vehicle_through_rate_table

It illustrates how many number of vehicles can pass the green light during unit time T.

C Penalty

Penalty denotes total waiting unit time of all traveling vehicles at each road intersection in the road network.

At moment t₀, from Fig. 9, we can get traffic_flow_table (tl₄, tl₁, t₀) = 23. This variable states that 23 vehicles will arrive at tl₄ from tl₁ at moment t₀. According to α = 0.8, β = 0.1, γ = 0.1(APPENDIX A.3), the number of cars going straightly, turning left and turning right are 18, 2 and 3 respectively. Then, the traffic_light_status_table(t₀)(APPENDIX B.Table 8) presents that the straightLight of road tl₁-tl₄ is green, with the red leftLight and rightLight. Since straightThroughRate is 16, only 16 cars which go straightly can pass tl₄, 2 cars stay. Cars turning left or turning right stay. At last,

Fig. 9.

Part of road network

Table 8. Traffic_light_status_table(t₀)

$$ {\text{Penalty }}\left( {{\text{tl}}_{ 4} ,{\text{ tl}}_{ 1} ,{\text{ t}}_{0} } \right) = 2 + 2 + 3 = 7 $$

Now, we get penalty of intersection tl₄ at time t₀. Using this method, we can get each road penalty at time t₀. At moment t₁, traffic_flow_table (tl₄, tl₁, t₁) = 15. According to α = 0.8, β = 0.1, γ = 0.1(APPENDIX A.3), the number of cars going straightly, turning left and turning right are 12, 2 and 1 respectively. The traffic_light_status_table(t₁)(APPENDIX B.Table 9) illustrates that the straightLight and rightLight color of road tl₁-tl₄ are green, while the leftLight is red. So the 12 going straight cars with 2 cars which was stayed during last unit time(t₀), as well as the straightThroughRate is 16, 14 cars can all pass the intersection tl₄ straightly. The number of turning left and turning right cars are both 4(already add the number of stayed cars during last unit time(t₀)). While rightThroughRate is 2, so 2 turning right cars have to stay. And, the 4 turning left cars also have to stay because of the red leftLight. Eventually, the penalty at t₁ of the road tl₁-tl₄ is:

Table 9. Traffic_light_status_table(t₁)

Penalty (tl₄, tl₁, t₁) = 2 + 4 = 6

The total penalty:

Penalty (tl₄, tl₁, t₀ + t₁) = 7 + 6 = 13

Now, we get penalty of road tl₁-tl₄ during t₀ and t₁.

The key point is: the less the penalty, the less the waiting delay.