Professional Documents
Culture Documents
10 1142@S0217984919500738
10 1142@S0217984919500738
Traffic flow dynamics and energy consumption differs under dissimilar weather condi-
tions, while seldom investigations have been conducted with a cellular automata model.
In this paper, the friction coefficient between ground and tire is considered as the quanti-
tative label of weather, a dynamic safe gap based on friction coefficient to avoid rear-end
crash is introduced. We developed a safer one-dimensional model to examine the kinetic
energy consumption under different weathers. Numerical results show that previous mod-
els overestimated the kinetic energy consumption in medium density flow (density < 0.5).
In medium flow, speed limit will not reduce energy consumption on rainy and snowy
days in most cases, but is necessary for prevention of accidents. Inversely, the effect of
speed control on energy consumption is obvious under extreme weather. Our work can
promote a better understanding of traffic dynamics, reduce energy dissipation and be
applied to real traffic management.
Keywords: Cellular automaton; traffic flow dynamics; energy dissipation; safety gap.
1. Introduction
A better understanding of complex traffic problems can provide workable schemes
to alleviate the situation we are facing, such as car crashes frequently occurring,
energy deficiency and air pollution. Models based on statistics, kinematics, car-
following models and cellular automata (CA for short) models in Refs. 1–6 have
already gained considerable achievements. CA models can reproduce many complex
real traffic phenomena, including traffic jams, stop-to-go, queuing in a signalized
intersection, and so on.
∗ Corresponding author.
1950073-1
L. Huang, D.-Y. Guan & X.-H. Qiang
the capacity and made it. In Ref. 4, the anticipant speed of leading vehicle and a
fixed safety gap were taken into consideration, microscopic features of traffic flow are
more precisely described. Later, CA models were widely used for modeling actual
traffic problems. Recently, Qian et al.5 established a mixed traffic flow CA model
within a signalized intersection, driving characteristics and other parameters were
taken into consideration, the evolution and propagation characteristics of gathering
dissipating of the traffic wave was reproduced. Malecki performed a simulation
about on-street parking problem with CA model based on field research data.6 Jin
Mod. Phys. Lett. B Downloaded from www.worldscientific.com
et al.7 used the empirical data to calibrate and validate the parameters of CA model.
CA model is also capable of modeling synchronized flow of three-phase theory,8 such
as KKW model9 and modified KKW model.10 Their models work very well, while
there is no consideration of actual road conditions.
Energy dissipation of traffic flow was investigated in a car-following model for
the first time in Ref. 11, further researches on this problem were based on dif-
ferent car-following models.12–14 Zhang et al.15 proposed a similar definition of
kinetic energy loss as in Refs. 13 and 14, they identified the energy loss in both
deterministic (no random perturbation in traffic flow) and non-deterministic con-
dition in NaSch model. Tian et al.16 studied the kinetic energy dissipation in
mixed traffic flow (vehicles with different length and different maximum speed).
Wen et al.17 investigated the energy loss in other CA models. In recent years,
researches mainly focus on the energy loss in complex road sections and bot-
tlenecks.18–20 Their contributions are of great inspiration to understanding the
energy consumption in road systems. Nevertheless, few studies on energy loss
consider safe driving and weather conditions. Our work will enrich the research
on the energy dissipation of road system, and facilitate further understanding of
traffic flow.
The outline of our study can be summarized as three points: (1) considers the
friction coefficient of different pavement as a label of different weathers; (2) intro-
duces the safety gap between leading vehicle and the follower, proposing a safer
one-dimensional CA model; (3) simulations are carried out and kinetic energy loss
under different conditions are discussed in detail.
This paper is organized as follows. The following part is devoted to the
derivation of dynamic safety gap and the description of our safer CA model,
the definition of kinetic energy loss is given. Then, series of simulation scenar-
ios and experimental studies are discussed. Finally, the paper concludes with a
summary.
1950073-2
Modeling the kinetic energy dissipation of road system
by UNIVERSITY OF NEW ENGLAND on 02/14/19. Re-use and distribution is strictly not permitted, except for Open Access articles.
2. Proposed Models
Mod. Phys. Lett. B Downloaded from www.worldscientific.com
vn2 v2
gapn = vn τn + − n+1 + d0 , (2)
2an 2an+1
even the reaction time of different drivers varies, most CA models ignore the reaction
time for simplicity. In this paper, reaction time of drivers was also ignored, i.e.
τn = 0. We assume that all vehicles have the same deceleration limitation bmax . A
crash would not happen unless d0 is a negative value, so d0 ≥ 0 must always meet
in the model. With all the above hypothesis, a simplified form of Eq. (2) is acquired
as
vn2 − vn+1
2
gapn = . (3)
2bmax
In the process of braking, all tires are forced to slow down or stop by the road
braking force FX as shown in Fig. 2(a).21 Here, we ignore the lateral motion of
1950073-3
L. Huang, D.-Y. Guan & X.-H. Qiang
by UNIVERSITY OF NEW ENGLAND on 02/14/19. Re-use and distribution is strictly not permitted, except for Open Access articles.
Mod. Phys. Lett. B Downloaded from www.worldscientific.com
(a) (b)
Fig. 2. (a) Mechanic analysis of tires when braking and (b) the change of road braking force.
1950073-4
Modeling the kinetic energy dissipation of road system
each cell equals to the mean length of a car. Cells have a binary state, empty or
taken by a car. All vehicles update in a single lane following the four steps of NaSch
model are as follows:
(i) Acceleration:
vn (t + 1/3) = min[vn (t) + 1, vmax ] . (8)
(ii) Deterministic brake to avoid collision:
vn (t + 2/3) = min[vn (t + 1/3), dn ] . (9)
Mod. Phys. Lett. B Downloaded from www.worldscientific.com
1950073-5
L. Huang, D.-Y. Guan & X.-H. Qiang
(ii) Acceleration:
vn (t + 2/3) = min[vn (t + 1/3) + 1, vmax ] . (15)
(iii) Deceleration:
by UNIVERSITY OF NEW ENGLAND on 02/14/19. Re-use and distribution is strictly not permitted, except for Open Access articles.
During the braking, the energy will be lost. According to Refs. 13 and 14, the
dissipated energy of vehicle labeled n from instant t to t + 1 is
[vn2 (t) − vn2 (t + 1)]/2, if vn (t) > vn (t + 1) ,
en (t) = (18)
0, otherwise.
There is an assumption that the mass of all vehicles equals 1. The mean kinetic
energy dissipation of the system in time interval T is
t0 +T XN
1 1 X
∆E = en (t) , (19)
T N t=t +1 n=1
0
where N is the number of vehicles in the system and t0 is the burning time. More-
over, ∆Ed denotes the energy loss caused by deterministic braking, ∆Er for the
energy dissipation resulting from randomization turbulence. Thus, another form of
mean energy dissipation is
∆E = ∆Ed + ∆Er . (20)
1950073-6
Modeling the kinetic energy dissipation of road system
by UNIVERSITY OF NEW ENGLAND on 02/14/19. Re-use and distribution is strictly not permitted, except for Open Access articles.
(a) (b)
Fig. 3. (Color online) (a) Fundamental diagram and (b) density-velocity curve.
Mod. Phys. Lett. B Downloaded from www.worldscientific.com
1950073-7
L. Huang, D.-Y. Guan & X.-H. Qiang
by UNIVERSITY OF NEW ENGLAND on 02/14/19. Re-use and distribution is strictly not permitted, except for Open Access articles.
∆Er of NaSch model keeps the lowest, for deterministic brake step is followed
by the random slowdown step in NaSch model while noise first (random slowdown
is at first step) in NF and SNF model. In free flow, ∆Er of SNF model is equal
to ∆Er of NF model. No congestion when k < k1 , where k1 is the critical density,
approximately equals to 0.11 in the simulation. Since our density step is 0.02, there
is a decline both in flux and ∆Er when k = 0.12. The average random energy loss of
SNF model keeps highest against other models. Although our model can maintain
a higher average speed, the introduction of the dynamic safety gap and the speed
of LV, the speed difference of the adjacent time steps increases accordingly. Thus,
the random energy consumption of SNF model is higher.
From Fig. 4(b), we can see that ∆Ed of SNF model is much lower than that of
NF model when k ∈ [0.12, 0.5], also lower than NaSch model when k ∈ [0.02, 0.46].
This finding suggests SNF model can reduce the chance of emergency braking and
cut down the consumption of kinetic energy, keeping a stable traffic flow meanwhile.
The total energy loss curve of SNF model shown in Fig. 4(c) is approximately con-
vex when k ∈ [0.12, 0.5], local minimum and global minimum of the total energy
loss ∆E can be found, the corresponding vehicle density is meaningful for traffic
management. ∆E of SNF model is lower than other models when k ∈ [0.12, 0.44], in-
dicating that previous models overestimated the energy consumption under medium
traffic density. This observation also infers that our revised model is more capable
of reflecting the real-life traffic flow dynamics.
1950073-8
Modeling the kinetic energy dissipation of road system
by UNIVERSITY OF NEW ENGLAND on 02/14/19. Re-use and distribution is strictly not permitted, except for Open Access articles.
limit is vmax = 6 (129.6 km/h) on dry pavement, vmax = 5 (108 km/h) in rainy
days, vmax = 3 (64.8 km/h) in a snowy day and vmax = 2 (43.2 km/h) on the icy
roads. We compared the effects of different speed limit schemes on kinetic energy
loss under different weather conditions. Still, more attention is being paid to energy
consumption under low- and medium-traffic flow (k ≤ 0.5).
Notable differences of ∆Er on fine days and rainy days with speed limit are
found only in free flow as shown in Fig. 5(a), there is little difference between the
random energy loss on dry roads and ∆Er on rainy days without speed limit. But
as a whole, ∆Er with speed limit keeps lower than that without speed control.
In rainy days, limiting the maximum speed does little to reduce random energy
consumption, but in many cases (k ∈ [0.3, 0.5] and k = 0.2, 0.24, 0.26) it inversely
increases deterministic energy consumption as shown in Fig. 5(b). Furthermore,
the total energy loss rate ∆E also presents this numerical relationship and trend as
presented in Fig. 5(c). From Fig. 5(b), we can see that ∆Ed of all cases are zero in
free flow as there are sufficient gaps for vehicles to maintain the high speed. After
the critical point, ∆Ed is on the rise until k = k2 . Then, it keeps declining until
it is zero again. According to Eq. (21), critical density k1 will be no big difference
when vmax = 6 (k1 ≈ 0.12) and vmax = 5 (k1 ≈ 0.14), therefore, we will only say k1
when we discuss energy loss on rainy days. For total energy loss ∆E in Fig. 5(c), in
the range of k1 < k < k2 , speed control does not affect the overall changing trend
of two cases, while ∆E with speed limit is higher in most situations.
When the road is snow-covered, for 0.2 < k < 0.5 (the boundaries are the
critical points of NF model when vmax = 3), the most obvious feature is that
the ∆Ed and total energy loss ∆E increase monotonously as well as ∆Er has
monotonous decreasing tendency, all shown in Figs. 6(a), 6(b) and 6(c). The above
changes are so regular that there are no wide oscillations as in other cases. It can
be inferred that the dynamic safety distance scarcely has influence on the speed of
the system at this time. In fact, the same simulation results were obtained when
we set vmax = 3 in NF model. For ∆Er on snowy day with speed control scheme
1950073-9
L. Huang, D.-Y. Guan & X.-H. Qiang
by UNIVERSITY OF NEW ENGLAND on 02/14/19. Re-use and distribution is strictly not permitted, except for Open Access articles.
and a fine day, (b) caused by deterministic brake in a snowy day and a fine day and (c) total
energy loss rate in a snowy day and a fine day.
when k < 0.5, it is rather obvious that its value is lower than other conditions as in
Fig. 6(a). The deterministic energy loss plots in Fig. 6(b) show that speed control
scheme will definitely aggrandize energy consumption when k is in the range of
0.2 to 0.5. We can see that total energy loss on snowy day without speed limit is
higher than that with speed control at critical point k1 = 0.2 as shown in Fig. 6(c).
But soon (k = 0.24), the situation is reversed. In the following two density steps,
∆E without speed limit is back to the higher again. But from now on, ∆E with
speed control always maintains the higher until k = 0.5. Over the second critical
point k2 = 0.5, the energy consumption is the same in all cases for the failure of
dynamic safety gap, because the average space between vehicles is no more than
1 cell. Although the energy consumption with the speed limit may be higher than
that in the case of non-speed limit with a high probability, it is necessary to adopt
speed control measures in snowy days for security.
Comparing Fig. 7 with Fig. 6, respectively, it is found that the changing regu-
larity of energy consumption with speed control on icy pavement are similar with
snowy days, the first critical point for the phase transition is different for the differ-
ence of vmax , k1 ≈ 0.3 when vmax = 2. The significant difference of random energy
consumption is found only when k < k1 as presented in Fig. 7(a). Different from
∆Ed on rainy days and snowy days, ∆Ed on an icy road with speed limit is always
lower than that without speed limit management when k < 0.5. It also creates
a situation that the total energy consumption without speed limit is higher than
the situation before density reaches the second critical point as shown in Fig. 7(c).
When compared to normal weather (µ = 0.7), ∆E without speed control on an
icy road (µ = 0.1) is much higher. This comparison shows that driving in extreme
weather will consume more energy if speed limit scheme were castoff. This also
proves that speed limit control in extreme weather is a ponderable and effective
traffic management measure in terms of safety and energy consumption.
1950073-10
Modeling the kinetic energy dissipation of road system
by UNIVERSITY OF NEW ENGLAND on 02/14/19. Re-use and distribution is strictly not permitted, except for Open Access articles.
a dry road, (b) caused by deterministic brake on icy roads and a dry road and (c) total energy
loss rate on icy roads and a dry road.
From the above, we see speed control will not reduce total energy loss on rainy
days and snow-covered roads in most cases. But speed limit scheme can reduce
energy consumption in extreme road conditions (coefficient of friction is very small
in our simulation). In general, although speed limit does not always reduce energy
consumption, for the sake of safety, reasonable speed limit measures are relatively
simple and effective traffic management measures. In addition, we find that when
the speed limit is set as 3 or 2, SNF model will degenerate into NF model, because
of the failure of dynamic safety gap. This also shows that the low-speed motion of
vehicles can be well described by the previous models, but the high-speed motion
still needs to be further studied.
4. Conclusion
The aim of the study is to demonstrate that a reasonable modification of existing
models could lead to a further understanding of traffic flow dynamics and energy
loss of the road system, the results of this study are satisfactory. In this paper, the
friction coefficient of pavement is used as a quantitative index of different weather
conditions, dynamic safety distance can be calculated based on different friction
coefficient and the speed difference between LV and FV. A safer and realistic CA
model was proposed and applied to depict the kinetic energy consumption under
different weather conditions. We found that our new model can maintain a high flux
in medium density (k < 0.5), and most preceding models overestimated the kinetic
energy loss. Cut down trips and traffic control measures are suggested when bad
weather comes, for energy consumption increases according to our findings. Surpris-
ingly, speed limits on a rainy or snowy day will increase the energy consumption in
most cases, but in case of collisions, speed limits scheme must be adopted.
It must be stressed that the model has only been tested in a simulation envi-
ronment. Kinetic energy loss during vehicle deceleration is considered and defined
1950073-11
L. Huang, D.-Y. Guan & X.-H. Qiang
only, which does not involve energy consumption in the acceleration process. In
addition, the kinetic energy loss defined in this paper is the kinetic energy dissipa-
tion of traffic flow rather than fuel consumption. We hope to further explore the
real energy consumption (or fuel consumption) in the acceleration and deceleration
by UNIVERSITY OF NEW ENGLAND on 02/14/19. Re-use and distribution is strictly not permitted, except for Open Access articles.
process with actual experiments in the future. Although there are deficiencies, it
has guiding significance for traffic management and promotes our understanding of
traffic flow dynamics and energy consumption of road system.
Acknowledgments
This research was funded by Key R&D Project of Shandong Province (Grant
No. 2017GGX50104) and Scientific Research Foundation of Shandong University of
Science Technology for Recruited Talents (Grant No. 2015RCJJ032).
Mod. Phys. Lett. B Downloaded from www.worldscientific.com
References
1. K. Nagel and M. Scherckenberg, J. Phys. I. France 2 (1992) 2221.
2. X. B. Li, Q. S. Wu and R. Jiang, Phys. Rev. E 64 (2001) 066128.
3. Y. Xu, L. Y. Dong, L. Li and S. Q. Dai, Phys. Rev. E 71 (2005) 026123.
4. W. Knospe et al., J. Phys. A 33 (2000) 48.
5. Y. S. Qian et al., Nonlinear Dynam. 89(2) (2017) 1099.
6. K. Malecki, J. Comput. Sci. 28 (2018) 32.
7. J. W. Zeng et al., Mod. Phys. Lett. B 32(25) (2018) 1850301.
8. B. S. Kerner, Physica A 333 (2004) 379.
9. B. S. Kerner, S. L. Klenov and D. E. Wolf, J. Phys. A 35(47) (2002) 9971.
10. Y. S. Qian, X. Xiao and J. W. Zeng, Physica A 479 (2017) 509.
11. Y. Nakayama, Y. Sugiyama and K. Hasebe, Phys. Rev. E 65 (2002) 016112.
12. T. Wang, Z. Y. Gao and X. M. Zhao, Acta Phys. Sin. 55 (2006) 634.
13. W. Shi and Y. Xue, Physica A 381 (2007) 399.
14. B. A. Toledo et al., Phys. Rev. E 75 (2007) 026108.
15. W. Zhang, W. Zhang and Q. Xian, Physica A 387 (2008) 4657.
16. H. H. Tian et al., Acta Phys. Sin. 58 (2009) 4506.
17. J. Wen et al., Acta Phys. Sin. 59 (2010) 7693.
18. W. X. Zhu, Int. J. Mod. Phys. C 24 (2013) 1350046.
19. W. X. Zhu and L. D. Zhang, Int. J. Mod. Phys. C 25 (2014) 1450018.
20. Y. S. Qian et al., Mod. Phys. Lett. B 29 (2015) 1450264.
21. H. Q. Zhang, Ph.D. thesis, Southeast University, Nanjing (2016).
22. H. C. Dan, L. H. He and B. Xu, Int. J. Pavement Eng. 18 (2017) 485.
1950073-12