A Novel Intelligent Traffic Light Control Scheme

1
A Novel Intelligent Traffic Light Control Scheme

• Cheng Hu, Yun Wang
• Presented by Yitian Gu

2
Objective

• Propose a novel control scheme, in which both the
  upstream and downstream information are taken
  into account and traffic light states are
  dynamically reasoned and adaptively adopt
  real-time traffic conditions
• To reduce average waiting time at a road cross

3
Results

• Proposed control scheme vs. fixed-time control

First scenario All upstream lanes have the same
arrival rate
4
Results
Proposed control scheme vs. fixed-time control
Second scenario arrival rates are all set to
0.35 at first, then increased according to a
vector for n times Vector (0.02, -0.01, -0.01,
-0.02, 0.03, -0.01, 0.02, -0.01, -0.01, -0.02,
0.03, -0.01)
5
Outline

• III. Intelligent Traffic Light Control Scheme
• The Crossroad Model
• Two-way, three lanes each direction, i.e. 12
  upstream lanes and 12 downstream lanes
• Each upstream Li with corresponding downstream
  lane Lj together forms a serial lane, denoted by
  Lij
• 12 predefined states of a traffic light, each
  state sx allows vehicles in two serial lanes Lij
  and Lmn to go through

6
Outline

• III. Intelligent Traffic Light Control Scheme
• The Crossroad Model
• The Control Scheme
• Weight of a Serial Lane
• Defines the weight function of a serial lane
  using both the upstream and downstream
  information including arrival rate, departure
  rate and spaces vacant
• Basic idea is to choose the state which contains
  as many cars as possible in upstream, and as few
  cars as possible in downstream
• Considers the Starvation problem
• Selection and Duration of Traffic Light States
• Selects the state that has the largest sum of
  weight
• Each selected state has minimum green time t0, if
  the state remains the same as the previous one, a
  time slice ?t is added.

7
Outline

• IV. Simulation and Analysis
• Effects of Arriving Rate
• First scenario
• Proposed scheme increases slower as the arrival
  rate increases
• Comparatively, average waiting time of proposed
  scheme is around 1/3 of that of the fixed scheme
• Second scenario
• Given a fixed sum of arrival rates, the greater
  difference of the arrival rates, the more
  benefits of the intelligent control scheme

8
Outline

• IV. Simulation and Analysis
• Effects of Vehicle Speed, Vehicle Length, Vehicle
  Margin and Lane Length
• Effects of Minimum Green Light Duration
• With the increase of minimum green time, the
  average waiting time increases
• Effects of Beta
• When Beta increases, the average waiting time
  decreases

9
Outline

• V. Conclusion and Future Work
• Work has done
• Describe a road cross model
• Propose an intelligent control scheme
• Do simulations to explore schemes features
• In general the average waiting time is
  significantly less than that of the fixed-time
  scheme
• Future work
• Take into account the influence of acceleration,
  pedestrian and etc.
• Extend to control a traffic network

10
Control Scheme

• Weight of a Serial Lane

n0 target vehicle number of lane Lj when there
is no congestion n(t) current vehicle number of
lane Lj fink arrival rate of upstream lane
Lk foutj departure rate of downstream lane Lj P,
Q controllable parameters
11
Control Scheme

• Weight of a Serial Lane

P and Q
k current queue length K queue capacity

1. P is inversely proportional to Q
2. If a downstream is congested, the current traffic
   status has more effects
3. If a downstream is empty, the historical status
   has more effects

12
Control Scheme

• Weight of a Serial Lane

Problem State Starvation If some serial lanes
upstream is always low arrival rate and its
downstream is always full, certain traffic light
states may not be tuned for a long time
Solution Ageing function
t0 the minimum green time ß a positive real
number
Final weight of a serial lane
13
Control Scheme

• Selection and Duration of Traffic Light States

Each state si contains two serial lanes Lmn and
Lpq with their final weights wmn and wpq, thus si
is assigned Wi wmn wpq When the duration of
state s(t) comes to an end, a new state is
launched, which holds the largest value among
those of all the serial lanes at present. i.e.
If s(t1) s(t), a time slice ?t is added to the
current states duration Otherwise, the current
state is switched to s(t1), and its duration is
set to t0
14
Reference

• Cheng Hu and Yun Wang, A Novel Intelligent
  Traffic Light Control Scheme, in 2010 9th
  International Conference on Grid and Cooperative
  Computing (GCC), 2010, pp. 372376.

15
The End