2009 International Joint Conference on Computational Sciences and Optimization

A Mixed Traffic Flow Model Based on a Modified Cellular Automaton in

Two-Lane System

Jian-Xun Ding, Hai-Jun Huang, Qiong Tian

School of Economics and Management, Beijing University of Aeronautics and Astronautics,
Beijing, 100191 China
[email protected]

Abstract plug formed fast vehicles by slow ones and ping-pang

lane-changing behavior have been found.
Based on the basic symmetric two lane cellular Researchers have mainly focused on how to reduce
automaton model (STCA), we propose a modified plug (i.e. honk effect), however, less cared about how
STCA (MSTCA) for mixed traffic flow. In MSTCA, the to generate plug and ping-pang while driving.
headway also could be used for lane changing, and Therefore, the phenomena are probably captured by
then the length for safety criterion would be extended a the unrealistic rules. In this study, rules which required
lot. Lane changing region and first cell for lane lane changing side-movement independent on the
changing on the other lane could be located. The forward movement will be discussed. Taking these
situation while there are successive vehicles satisfies issues into account, we propose a new set of lane-
both the safety criterion and the motivation for lane changing rules during forward movement into the
changing; vehicles ahead would own lane changing basic STCA.
priority. Both the two factors will influence the
frequency of lane changing; eventually affect the total 2. Two-lane traffic CA model and our
traffic flow. Simulation results show that the MSTCA work
could acquire higher value of flux and reproduce more
realistic features of the mixed traffic flow.
The update step in two-lane system is divided into
two sub-steps: according to lane-changing rules drivers
1. Introduction may change lanes at first and then the lanes are
considered as independent single-lane NS models
Cellular automaton (CA) has become an efficient (parallel system).
and excellent tool for simulating traffic flow to
reproduce real-traffic phenomena. In 1992, Nagel and 2.1. NS model for forward movement
Schreckenberg proposed the well-known Nagel ̢
Schareckenberg (NS) model [1]. Later, several A lane from NS model is assumed to be composed
extensions of the NS model were proposed, such as the of L cells; each cell can be either empty or occupied by
FI model, TT model, etc [2, 3]. Although simple, the exactly one vehicle, with velocity between 0 and vmax.
occurrence of phantom traffic jams, the realistic flow- The model consists of four rules as follows:
density relation could be reproduced. Step1: acceleration, v n o min( v n  1, v max ) ;
In multilane system, when a vehicle could not reach
its desired velocity, it will change lane for better Step2: deceleration, v n o min( v n , d n ) ;
driving condition. Rickert et al [4] presented a Step3: randomization (with a certain probability P
straightforward extension of the single-lane CA model do), v n o max ( v n  1, 0 ) ;
by introducing a set of lane-changing rules. After that
lots of modifications have been expanded, in general Step4: vehicle movementˈ xn o xn  vn .
speaking, such as considering heterogeneous vehicles Here the index xn denotes the position of a vehicle, vn
[5], various driving behaviors [6], anticipation effects its current velocity. The variable dn is the number of
[7,8], honk effect [9] et al. Such phenomena as the empty cells in front of the vehicle at cell n. The driver
accelerates if the vehicle has not reached vmax and

978-0-7695-3605-7/09 $25.00 © 2009 IEEE 124

DOI 10.1109/CSO.2009.75
brakes to avoid accidents. P is the randomization the truth; therefore the leading vehicle would own lane
parameter, and Pę [0, 1]. changing priority and the following one would stay in
the same lane, as depicted in Fig 2(b), which could
2.2. Our improved work on lane-changing depress ping-pang changing behavior to some extent.

In our model, some important modifications have 3. Simulation results and discussion
been made. Shown in Fig 1, dn,other; dn,back denote the
number of empty cells between the nth vehicle and its 3.1. Description of simulation procedure
two neighbor vehicles on the other lane, respectively.
pn,change is the lane-change probability between 0 and 1, We carry out the simulations of the models under
which could be imposed on drivers. The rules are less the periodic boundary condition. There are two types
stringent than those proposed by Chowdhury(here of vehicles, vf max =5ٛ and vs max = 3. P0=0.5 pn,change
denoted by STCA) et al. because the actual dsafe of the =0.8. A system of 2xL sites is considered (L=1000).
following vehicle used is larger than its dn,back. The length of each site is set to be 7.5 m. One iteration
time step is set to 1 second.Thus the global vehicle
density is a conserved quantity, in the form of Eq. (5).
U N / (2 u L ) (5)
Where, N denotes the number of vehicles. The ratio of
the partial densities of the fast and slow vehicles is
Fig 1. Extensions for Lane changing behavior expressed by the symbol f. Each simulation was
(Here, RLC is the abbreviation of “Region for Lane Changing”,
FCLC for”First Cell for Lane Changing”)
carried out T time steps (i.e., T=11000). After a
This MSTCA has a symmetric rule set that vehicles transient period t0 time steps (i.e., t0=10000) to allow
change lanes if the following three criteria are fulfilled. the system to reach its steady state. Each q and v
represents an average value of 20 runs, in the form of
d n ,other (t ) ! d n (t ) and min(vn (t )  1, vmax ) ! d n (t ) (1)
Esq. (6-7).
Eq.(1) means that drivers change lane if they are t T
§ N · (6)
hindered on this lane and if there is more space on the v 1 / (T  t ) v (t ) / N 0 ¦ ¨© ¦
t t 0 1 i 1
i ¸
¹
other lane, which is identical to STCA.
q Uv (7)
(2)Safety criterion
d n ,back (t ) t d safe (2)
3.2. Flow behavior
d n , back ( t )  d n ( t )  1 t d safe (3)
d safe min(vn ,back (t )  1, vmax ) (4) The fundamental diagrams for STCA/MSTCA under
Next the lane changing rules of STCA (i.e. Eq. (2)) 12 different conditions are shown in Figs. 3
U-v U-q

are revised to take the extendable safety distance into

5 0.7
(0.0,1) (0.0,1)
(0.2,1) (0.2,1)
4.5 (0.4,1) (0.4,1)

account, as shown in Eq.(3). A vehicle could change

(0.6,1) 0.6 (0.6,1)
(0.8,1) (0.8,1)
4 (1.0,1) (1.0,1)
(0.0,0) (0.0,0)
(0.2,0) (0.2,0)

lanes in the RLC, if it does not hinder the vehicle

0.5
3.5 (0.4,0) (0.4,0)
(0.6,0) (0.6,0)
(0.8,0) (0.8,0)
3
(1.0,0) (1.0,0)

behind on the other lane. That means after the lane

0.4

2.5
q
v

change, no collision would occur. The vehicle could

0.3
2

1.5

change to other lane in the RLC with a FCLC in form

0.2

0.1

as follows, xn' (t) xn(t)max min(vnback

, (t)1,vmax)dnback
, (t),0
.
0.5

0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
U U

While dn,back is large enough, this criterion will reduce

Fig 3. The fundamental diagram for STCA with
to the Eq. (3) in STCA model.
priority rule(c=0) and MSTCA(c=1), with different
(3)Priority rule for leading vehicle
pair of (f, c)
Comparing Figs 3 (a) and (b), it shows that total flux
made by MSTCA is higher than that of STCA in the
intermediate density range. The reason is that faster
vehicles have more safety space to change lane while
after slow ones. As faster vehicles need not make a
Fig 2. Priority rule for the leading vehicle change or they are most often unable to attain high
As shown in Fig 2(a), both the two vehicles will speed because of mutual hindrance, there is hardly any
change lanes with a certain pn,change. Obviously it is not

125
difference among all the curves in the small or high previous sub-section. Comparing Figs 5 (a) and (b),
density regime. the AFLC got by STCA is higher than that of STCA
with priority in nearly all density range.
3.3. Lane-Changing behavior
4. Conclusions
The average frequency of lane change (AFLC) per
vehicle per unit time has been plotted against the total In this paper, a new set of lane-changing rules is
density in Fig 4. While there is only one type of proposed to the MSTCA. We have studied the
vehicles, the AFLC exhibits a maximum at a moderate fundamental diagrams of the MSTCA in detail and
density. The reasons for the variation with density are compared it with that of STCA model. It is shown that
as follows: at very low density there is enough space in MSTCA model enhances slightly the flux of the mixed
front and a vehicle very rarely feels the need for a lane traffic system in the intermediate density range, but
change and at very high densities most of the attempts increase AFLC a lot. In other words, lane changing
of lane changing are unsuccessful either because of affects the traffic flow appropriately not crucially,
close proximity of other vehicles in front or behind the which is the similar to results from Tang [10].
empty neighboring site on the other lane.
U-l2r U-r2l
0.025
(0.0,1)
(0.2,1)
0.025
(0.0,1)
(0.2,1)
Acknowledgments
(0.4,1) (0.4,1)
0.02 0.02
(0.6,1) (0.6,1)

The project is supported by the National Natural

(0.8,1) (0.8,1)
(1.0,1) (1.0,1)
0.015 (0.0,0) 0.015 (0.0,0)
(0.2,0)
(0.4,0)
(0.2,0)
(0.4,0) Science Foundation of China (Grant No 70701002,
l2r

r2l

70521001), the National Basic Research Program of

(0.6,0) (0.6,0)
0.01 (0.8,0) 0.01 (0.8,0)
(1.0,0) (1.0,0)

0.005 0.005
China (Grant No 2006CB705503).

0
0 0.5
U
1
0
0 0.5
U
1
References
Fig 4. The AFLC, (a) from left lane to right one, (b)
[1] K. Nagel, M. Schreckenberg, “A Cellular automaton
right to left. The pair symbols and values have the
model for freeway traffic”, Journal of Physique, I (France), 2,
same meaning as in Fig 3 1992, pp. 221-229.
In the mixed flow, two peaks appear over a range of [2] M. Fukui, Y. Ishibashi, “Traffic flow in 1D cellular
the total density. Faster vehicles contributed to the left automaton model including cars moving with high speed”, J
peak and slower vehicles to the right peak. Fewer Phys Soc Jp, 65 (6) ,1996, pp. 1868–1870.
faster ones would have more probability be hindered [3] M. Takayasu, “H.1/f noise in a traffic model”, Fractal, 1,
by slow ones, stronger incentive and enough space at 1993, pp. 860-866.
lower density range. Therefore, the smaller f(nonzero), [5] M. Rickert, K. Nagel, and M. Schreckenberg, “Two lane
the higher the left peak. Comparing Figs 4 (a) and (b), traffic simulations using cellular automata”, Physica A, 231,
1996, pp. 534-550.
we find that the AFLC obtained by MSTCA is higher
[6] D. Chowdhury, D.E. Wolf, and M. Schreckenberg,
than that of STCA in nearly all density range. “Particle hopping models for two-lane traffic with two kinds
of vehicles: effects of lane changing rules”, Physica A, 235,
3.4. Vehicle priority rule on lane changing 1997, pp. 417-439.
[7] N. Moussaa, A.K. Daoudia, “Numerical study of two
7
x 10
-3
U-l2r

(0.0,1)
7
x 10
-3
U-r2l

(0.0,1)
classes of cellular automaton models for traffic flow on a
two-lane roadway”, Eur Phys J B, 31, 2003, pp. 413-420.
(0.2,1) (0.2,1)
(0.4,1) (0.4,1)
6 (0.6,1) 6 (0.6,1)
(0.8,1) (0.8,1)

5
(1.0,1)
(0.0,0)
(0.2,0) 5
(1.0,1)
(0.0,0)
(0.2,0)
[8] W. Knospe, L.Santen, and A. Schadschneider, “Disorder
effects in cellular automata for two-lane traffic”, Physica A,
(0.4,0) (0.4,0)
(0.6,0) (0.6,0)
(0.8,0) (0.8,0)
(1.0,0) (1.0,0)
4 4

265, 1999, pp. 614-633.

l2r

r2l

[9] B. Jia, R. Jiang, and Q.S. Wu, “Honk effect in the two-
3 3

2 2

lane cellular automaton model for traffic flow”, Physica A,

1 1
348, 2005, pp. 544-552.
0
0 0.2 0.4
U
0.6 0.8 1
0
0 0.2 0.4
U
0.6 0.8 1 [10] T.Q. Tang , H.J. Huang, S.C. Wong, and R. Jiang,
“Lane changing analysis for two-lane traffic flow”, Acta
Fig 5. Comparing AFLC between STCA(c=1) and Mech Sin, 23, 2007, pp. 49–54.
STCA with priority(c=0). The symbols (r2l,l2r,(f,c))
and values have the same meaning as in Fig 3-4
Two peaks appear over a range of the total density,
as expected. The reasons are the same with the

126