I S S N 1 0 0 7 - 0 2 1 4 1 2 / 1 4 pp 2 1 4 - 2 2 2
Volume 12, Number 2, April 2007
Introduction
In the past several years, traffic micro-simulations
based on one-dimensional cellular automata (CA) have
been widely used because of their unique advantages
to describe the complexity of city traffic[1,2]. This simple algorithm, which can be easily adapted for parallel
computing, has easily-modified parameters which can
be used to model different traffic conditions. The algorithm is suitable to large scale computer simulations
which yield statistical results that can be easily compared with macro data for real conditions.
In 1992, Nagel and Schreckenberg[3] proposed a
one-dimensional cellular automata model to simulate
traffic flow on a freeway, providing the basic principles for more complex surroundings such as city traffic
flow. Improvements have been made to the N-S
Received: 2006-10-01; revised: 2006-12-05
(Nagel-Schreckenberg) model to adapt it to more realistic circumstances, including the additional slow-tostart rule and the extension from single lane to multilane models[4-7]. Other scholars have used the CA
model to simulate pedestrian flow with a variety of results. Blue et al.[8] proposed a pedestrian movement
model for large scale open areas. Muramastu et al.[9]
developed a pedestrian movement model based on stochastic processes. Blue and Adler[10] then developed a
four-directional pedestrian crosswalk model and a year
later proposed a bi-directional pedestrian crosswalk
model[11].
Most of these literatures have focused on the interactions between vehicles and other vehicles, or between
pedestrians and other pedestrians, while the interactions between vehicles and pedestrians have been
rarely discussed. However, in many cities in China,
large numbers of pedestrians compete with vehicles to
cross crosswalks. When the traffic light is not reasonably configured, a serious traffic chaos occurs leading to
a notable decline in transit efficiency and traffic
safety[12].
This paper describes a mixed traffic flow model
based on cellular automata to analyze the characteristics of mixed traffic flow at crosswalks mainly
composed of vehicle and pedestrian flows. The microsimulation model consists of three sub-models. The
vehicle sub-model simulates the vehicle flow based on
the N-S model; the pedestrian sub-model simulates the
pedestrian flow at the crosswalk; and the interaction
model describes the interactions between the vehicles
and the pedestrians at the crosswalk. The simulations
of the interactions between the vehicles and pedestrians are used to analyze the influence of these interactions on the traffic capacity and efficiency. The simulation results can be used to develop better traffic management strategies for mixed traffic conditions to increase safety, enhance efficiency, and avoid congestion.
1.1
Vehicle sub-model
215
216
(1) Eliminate conflicts: Two walkers that are laterally adjacent may not sidestep into each other. An
empty cell between two walkers is available to one of
them with a 50/50 random assignment.
(2) For each pedestrian, the same or adjacent lane
which best advances forward movement up to the
maximum speed is selected. This gap is calculated by a
sub-procedure.
(3) To evade a pedestrian coming in the opposite direction, each pedestrian may choose to step laterally.
(4) If movement to overtake the walker ahead of you
with a lower speed is interrupted by a walker in the
opposite direction, then the pedestrian will choose to
step behind the walker going in the same direction.
(5) The probability to change lane is defined as:
Each of two adjacent walkways has a 50% probability
that they will change lanes. If none of the directions
are precluded, then the probability is 80/10/10
Interaction sub-model
The interaction sub-model is the most important submodel for describing the interactions between vehicles
and pedestrians at the crosswalks. Generally, in urban
areas, pedestrian behavior is always more aggressive
than the vehicles, which means that vehicles always
decelerate voluntarily before the stop lines to avoid
conflicts with pedestrians while pedestrians always
cross the street as soon as possible when they feel the
situation is safe. To describe these characteristics, the
model mainly focuses on the pedestrian behavior, including the possibility that pedestrians will not abide
by the traffic laws and the group effect of the pedestrians. The basic rules are shown in Fig. 2.
217
(1)
218
linearly, which implies that the riskers have little influence on the vehicle flow. However, Fig. 4b for a larger
pedestrian volume ( pped = 0.8 ) shows that saturation
2.2
Therefore, the riskers strongly affect the traffic capacity and efficiency in crowded areas, so policies are
needed to restrict the behaviors of those riskers. For
example, a pedestrian bridge may be required.
The average vehicle and pedestrian waiting time for
these conditions are shown in Figs. 4c and 4d. In Fig. 4c,
when pped is 0.5, the average vehicle waiting time is
less than 200 s. However, when pped is 0.8, the waiting
time increases to more than 1000 s, indicating heavy traffic congestion. This also demonstrates the significance
of restricting the behavior of riskers in crowded areas.
The results in Fig. 4d show that the average pedestrian waiting time curves are very similar for different
pveh . This again shows that the pedestrians have
higher priorities than vehicles, so their behavior is not
219
nomenon is important for traffic regulators when designing traffic control strategies. For instance, for traffic conditions similar to the model with pveh = 0.6,
2.3
The influence of the traffic light split on the traffic capacity and how efficiency was investigated for pedestrians generated with probability pped = 0.8 or 0.5,
riskers generated with probability prisk = 0.3, and the
traffic light period set to 120 s.
Figure 5a shows the vehicle flow rate for various red
light phases. When the red light phase is relatively
220
light at the crosswalk. When there are not many vehicles on the road ( pveh = 0.1 and 0.3 ), the normalized
Figures 5c and 5d show the effect of the red light timing on the average vehicle and pedestrian waiting time.
The average vehicle waiting time increases if the light is
red too long; whereas the average pedestrian waiting
time decreases linearly as Tred increases, which is the
reason for adding a traffic light at the crosswalk: to reduce the average pedestrian waiting time at the crosswalks and facilitate pedestrians crossing the streets.
2.4
Spatiotemporal maps
221
Conclusions
222
References
model considering the velocity effect of a car on the successive car. Phys. Rev. E, 2001, 64(6): 066128.