modeling and analysis of the car-truck heterogeneous
TRANSCRIPT
Yang, Jin, Ran, Pu and Yang 1
Modeling and Analysis of the Car-truck Heterogeneous Traffic Flow Based on Intelligent 1 Driver Model 2 3 Da Yang, Ph.D. Candidate 4 Department of Civil and Environmental Engineering, 5 University of Wisconsin-Madison, 6 1415 Engineering Drive, Madison, WI 53706, USA 7 Phone: 1-608-334-4281 8 E-mail: [email protected] 9 10 Jing Jin, Ph.D., Postdoctoral Fellow 11 Department of Civil, Architectural, and Environmental Engineering, 12 The University of Texas-Austin, 13 1616 Guadalupe St., Ste. 4.202, Austin, TX 78701, USA 14 Phone: 1-512-232-3124 15 E-mail: [email protected] 16 17 Bin Ran, Ph.D., Professor 18 Department of Civil & Environmental Engineering, 19 University of Wisconsin-Madison, 20 1415 Engineering Drive, Madison, WI 53706, USA 21 Phone: 1-608-262-0052 Fax: 1-608-262-5199 22 Email: [email protected] 23 and 24 School of Transportation, Southeast University, 25 No.2 Si Pai Lou, Nanjing 210096, China 26 27 Yun Pu, Ph.D., Professor 28 School of Transportation and Logistics, 29 No.111 erhuanlubeiyiduan, Chengdu, 610031, China 30 Email: [email protected] 31 32 Fei Yang, Ph.D., Associate Professor 33 School of Transportation and Logistics, 34 No.111 erhuanlubeiyiduan, Chengdu, 610031, China 35 Email: [email protected] 36 37 Corresponding Author: Da Yang 38
39 Submitted for Presentation and Publication 40 To the 92nd Transportation Research Board Meeting 41 Submission Date: Sep. 15th, 2012 42
43 5449 Words + 1 Table + 6 Figures = 7199 Words 44
TRB 2013 Annual Meeting Paper revised from original submittal.
Yang, Jin, Ran, Pu and Yang 2
ABSTRACT 1
The traffic flow heterogeneity caused by the different car-following dynamics among the different types 2
of vehicles has drawn increasing attention recently. This paper explores the characteristics of the four 3
types of car-truck car-following combinations, car-following-car (CC), car-following-truck (CT), truck-4
following-car (TC) and truck-following-truck (TT), and their impacts on traffic flow stability. A 5
heterogeneous traffic flow model based on Intelligent Driver Model (IDM) is proposed and calibrated 6
using the Next Generation Simulation (NGSIM) vehicle trajectory data. Based on the calibrated model, 7
the characteristics of the car-truck heterogeneous traffic flow are evaluated using the linear stability 8
analysis, fundamental diagrams, and shock wave characteristics. The linear stability analysis identifies 9
two critical factors that can influence the stability of the car-truck heterogeneous traffic flow: the 10
stability functions and the proportions of the four types of car-truck combination. Cars and trucks can 11
both stabilize and destabilize the traffic flow depending on the combination type and the equilibrium 12
velocity. Fundamental diagrams of car-truck heterogeneous flow are found to be determined by the 13
distance headways and proportions of the four types of combination. Moreover, the fundamental 14
diagrams of different car-truck combinations converge to several clusters with the same proportion 15
difference between the CC and TT combinations. The slowing-down effect of trucks on shock wave 16
speed in the car-truck heterogeneous traffic flow is also observed in the simulation. 17
18
KEYWORDS: CAR AND TRUCK, HETEROGENEOUS TRAFFIC FLOW, INTELLIGENT 19
DRIVER MODEL, CAR-FOLLOWING COMBINATIONS 20
TRB 2013 Annual Meeting Paper revised from original submittal.
Yang, Jin, Ran, Pu and Yang 3
INTRODUCTION 1
Heterogeneity is a key characteristic of real-world traffic flow. Although traffic flow theories and 2
models are usually developed first for the homogeneous traffic flow, most of them can be easily 3
converted into their heterogeneous forms. However, the difficulty is the lack of sufficient data to 4
calibrate and explore the heterogeneous models. With the development of the data collection technology 5
in recent years, such as the video-based method and the GPS-based method, traffic information, 6
especially individual vehicle dynamics, can be obtained in greater detail. Many researchers have been 7
using these new data sources to investigate the characteristics of car-truck heterogeneous traffic flow in 8
the past few years. Early researchers focused on investigating the differences between the car and truck 9
driving behavior. Huddart and Lafont (1), McDonald et al. (2) and Sayer et al. (3) compared the 10
headway differences between these two cases: car-following-car and car-following-truck, but their 11
studies did not reach conclusive results regarding with which case had the larger headway. Peeta et al. (4, 12
5) analyzed interactions of cars and trucks in multiple lanes. Highway Capacity Manual presented that 13
trucks occupied more space, had poorer operating capabilities, and created lager gaps than cars in most of 14
situations (6). However, these studies (1-5) did not recognize that the car-following behavior also depends 15
on the following vehicle type. Ye’s study (7) first explores the impact of the following vehicle type (car 16
or truck) on traffic flow. He concluded that the four types of car-truck car-following combination should 17
be taken into account in the study of the car-truck heterogeneous traffic flow: the car-following-car (CC), 18
the car-following-truck (CT), the truck-following-car (TC), and the truck-following-truck (TT). Sarvi (8) 19
also studied the driving behavior of these three car-following combinations, car-following-car, truck-20
following-car and car-following-truck. Aghabayk et al. (9) further studied the distance headway, time 21
headway, reaction time and car-following threshold variations among the four types of combination. One 22
major limitation of those early studies on car-truck traffic flow characteristics is the lack of modeling of 23
the dynamic traffic flow characteristics. Mason and Woods (10) developed the homogeneous Optimal 24
Velocity (OV) car-following model into a heterogeneous traffic flow model to describe the interaction 25
between cars and trucks. The derived OV heterogeneous car-following model is as follows, 26
nnnnnn vxxU
dt
txd 12
2 )( (1) 27
where xn(t) and vn(t) respectively denote the location and velocity of the vehicle n at time t, xn-1(t) 28
denotes the location of the vehicle n-1 (the preceding vehicle of vehicle n) at time t, n denotes the 29
sensitivity parameter of the vehicle n, nnn xxU 1 denotes the optimal velocity function the vehicle n 30
wishes to take, and it is the function of the headway of vehicle n. 31
The heterogeneous Optimal Velocity model is used to conduct the comparison of the car-32
following dynamics between homogeneous and heterogeneous vehicles and it is found that trucks may 33
TRB 2013 Annual Meeting Paper revised from original submittal.
Yang, Jin, Ran, Pu and Yang 4
dampen congestion waves. Because of the complication in their multi-species formulations, it is difficult 1
to analyze individually how different car-truck following combinations can affect traffic flow. In this 2
study, based on Intelligent Driver Model (IDM), we developed a heterogeneous model with the different 3
sub-models for each car-truck following combination. To calibrate the new model, the model parameters 4
for each car-truck car-following combination are calibrated separately using the car-following data 5
extracted from the I-80 NGSIM (Next Generation Simulation) data for each combination. Based on the 6
calibrated model, we study the three characteristics of the car-truck heterogeneous traffic flow, the linear 7
stability, fundamental diagrams, and shock wave. 8
9
METHODOLOGY 10
An IDM-based Car-following Model for the Car-truck Heterogeneous Traffic Flow 11
Treiber et al. proposed Intelligent Driver Model in 2000 (11) for the homogeneous traffic flow. This 12
model is a widely explored car-following model (12-14), and its formulation is as follows, 13
ab
tvtvtv
V
tvsstvtvS
ltx
tvtvS
V
tva
dt
txd
nnn
nnn
n
nnnn
2
)()()(
)())(),((
)(
))(),(()(1
)(
10
2
2
2
(2) 14
where a is the maximum acceleration, V is the desired velocity, is the acceleration exponent, S() is the 15
desired minimum gap, s0 and s1 are the jam distances, is the safe time headway, b is the desired 16
deceleration, l is the leading vehicle length, and tvtvtv nnn 1 is the removal rate of the vehicle 17
n to its preceding vehicle n-1. 18
We develop the homogeneous IDM to its heterogeneous form by giving subscripts to the model 19
parameters. The proposed heterogeneous IDM formulates the four different car-truck car-following 20
combinations as follows, 21
nn
nnnn
n
nnnnnn
nn
nnn
n
nn
n
ba
tvtvtv
V
tvsstvtvS
ltx
tvtvS
V
tva
dt
txdn
2
)()()(
)())(),((
)(
))(),(()(1
)(
10
2
2
2
(3)
22
where all the parameters an, n, Vn, 0ns , 1
ns , n and bn have four alternatives. Taking an as an example, an 23
can be acc, act, atc and att. The leading vehicle length l has two alternatives, lc and lt. 24
In the homogeneous traffic flow, all vehicles at the equilibrium state have zero acceleration, the 25
same distance headway and the same velocity; while at the equilibrium state, all vehicles in the 26
TRB 2013 Annual Meeting Paper revised from original submittal.
Yang, Jin, Ran, Pu and Yang 5
heterogeneous traffic flow still have zero acceleration and the same velocity, but their distance headways 1
vary for different vehicles which can be described using the following equations: 2
vvn , 0nv , and nn hh (4) 3
where v is the equilibrium velocity which is the same for all vehicles, nh is the corresponding 4
equilibrium headway of the vehicle n, and nh varies among the vehicles. Substituting (4) into the 5
heterogeneous IDM (3) yields, 6
vvvssvs
lh
vs
v
v
IDM
010
2
0
/)0,(ˆ
0)0,(ˆ
1 (5) 7
Rewriting the above equation yields, 8
lvv
vvvssh
0
010
/1
/ (6) 9
In the car-truck heterogeneous flow, h has the four alternatives, cch ,
cth , tch and
tth which 10
correspond to the four types of combinations. It should be noticed that the vehicle length l in h only 11
depends on the leading vehicle type, namely, l = lc in cch and
tch , and l = lt in cth and
tth . 12
Then, the fundamental diagram of the heterogeneous IDM for the equilibrium state can be 13
derived based on Equation (6). Assume a given heterogeneous traffic flow on a single lane contains N 14
vehicles. The length of the entire traffic flow in the equilibrium state is: 15
N
nntotal hL
1
(7) 16
In the car-truck heterogeneous traffic flow, Ltotal has the following formula: 17
tttttctcctctcccctotal hNhNhNhNL (8) 18
where Ncc, Nct, Ntc and Ntt are the numbers of the CC, CT, TC and TT combinations in the car-truck 19
heterogeneous traffic flow. Hence, the density of traffic flow can be calculated as: 20
tttttctcctctcccc hNhNhNhN
Nk (9) 21
Rewriting equation (9) in the following form: 22
tttttctcctctcccc hPhPhPhP
k1
(10) 23
where Pcc, Pct, Ptc and Ptt are the proportions of the CC, CT, TC and TT combinations. Thus, the 24
fundamental diagram of the car-truck heterogeneous traffic flow has the following relationship: 25
TRB 2013 Annual Meeting Paper revised from original submittal.
Yang, Jin, Ran, Pu and Yang 6
tttttctcctctcccc hPhPhPhP
vq
(11) 1
where q is the flow rate of the car-truck heterogeneous traffic flow in space. 2
Equation (11) indicates that the distance headways and proportions of the four types of 3
combination determine the fundamental diagram of the car-truck heterogeneous traffic flow. 4
5
Linear Stability Criterion of the Car-truck Heterogeneous Traffic Flow 6
The linear stability analysis investigates the perturbation propagation characteristic of a vehicle platoon 7
by adding a small perturbation on the first vehicle of the platoon (13, 15). The stability criterion of IDM 8
can be derived by following the general stability criterion summarized by Wilson and Ward (13). The 9
derivation of the detailed stability criteria of IDM requires the partial differentials of the velocity, 10
headway and velocity difference between the leading and following vehicles with respect to the 11
equilibrium state of IDM. 12
vv Slh
aS
V
vaf 2
12
, vv Slh
aSf
2
2, 32
2 lhSafh (12) 13
where ),0,( hvSS . vS , vS and hS are respectively the partial differences of the desired minimum 14
gap function with respect to the vn, vn and xn at the equilibrium state ),0,( hv . Hence, according to 15
the general stability criterion introduced by Wilson and Ward (13), the stability criterion of IDM is as 16
follows, 17
02
2
v
hvvf
fff (13) 18
Substituting equation (12) into (13) yields the stability criterion of IDM, 19
02
2
2222
2
1
02
1
023
2
lh
S
v
va
lh
SS
v
v
lh
SaS
lh
S vv
(14) 20
The derivation of the linear stability criterion of the car-truck heterogeneous traffic is more 21
complicated. Ward (16) presented the general formula of the linear stability criterion of the 22
heterogeneous traffic flow as follows,
23
02/
2
,2,,,,
n njhjvnhnvnvn fffff (15) 24
TRB 2013 Annual Meeting Paper revised from original submittal.
Yang, Jin, Ran, Pu and Yang 7
where vnf , , vnf , , hnf , are the partial differences of the car-following model adopted by the vehicle n 1
with respect to v, v and x at the equilibrium state where vvn , 0 nv and nn hx . 2
However, Equation (15) cannot directly reflect the proportion information of the different 3
combinations, so we rewrite it as the following form using the induction method: 4
02/1
1
2,
22,
2,,,,
Q
i
Q
jij
NPhj
NPhivihivivii
ii ffffffNP , (16) 5
where Q is the number of the combinations in a given car-truck heterogeneous traffic flow, N is the total 6
number of vehicles in the traffic flow, Ni is the total number of the combination type i in the platoon, 7
N
NP i
i is the proportion of the combination type i in total, and 1
1
M
iiP . In the car-truck 8
heterogeneous traffic flow, Q = 4 and i has four alternatives: CC, CT, TC and TT. 9
Equation (16) can be further simplified as follows, 10
02/
12,
2,,,,
Q
i hi
vihivivii
f
ffffP (17) 11
In Equation (17), two parts determine the stability criterion of the car-truck heterogeneous traffic 12
flow: the fraction part 2,
2,,,, 2/
hi
vihivivi
f
ffff providing the stability characteristic of the combination 13
i and the proportion part iP of the combination i. We define the fraction part as the Stability Function 14
(SF) of the combination, 15
2,
2,,,,
,,,
2/),,(
hi
vihivivihivivii f
fffffffSF
(18) 16
where SFi is the stability function of the combination type i. In addition, comparing the stability criterion 17
of IDM (14) with equation (18), it can be found that SF has the same sign as (14), so SF also can be used 18
to judge the stability of the car-following combination i. The stability effects of combination type i 19
denoted as SFi in the car-truck heterogeneous traffic flow can be written as follows, 20
2
2
1
2,
1
2,
3
2
42
62
2
222
4 ii
i
ii
i
ii
ivi
ii
ii
viii
ii
i
ii
iii
lh
S
V
va
lh
SS
V
v
lh
SSa
lh
S
Sa
lhSF
i
(19)
21
Moreover, we define the left part of equation (17) as the stability function F of the car-truck 22
heterogeneous traffic flow. Thus, in the car-truck heterogeneous traffic flow, there is, 23
tttttctcctctcccc SFPSFPSFPSFPF (20) 24
TRB 2013 Annual Meeting Paper revised from original submittal.
Yang, Jin, Ran, Pu and Yang 8
When F < 0, the car-truck heterogeneous traffic flow is stable. Equation (20) indicates that the 1
proportion of the car and the truck is not the only deciding factor of the stability of heterogeneous traffic 2
flow, and the stability function of each car-truck following combination is also significant. Furthermore, 3
on a ring road, the stability function of the car-truck heterogeneous traffic flow and the proportions of 4
four types of combination have the following relationship: 5
tcct
tttcctcc
tttttctcctctccccl
PP
PPPP
SCPSCPSCPSCPF
1 (21) 6
7
Model Calibration and Experimental Design 8
The car-following model calibration is a nonlinear optimization problem. In this study, we use Genetic 9
Algorithm (GA) to solve this nonlinear optimization problem to obtain the optimal parameters. We also 10
adopt the Theil’s U function as the objective function as suggested in several existing studies (17, 18). 11
M
m
simm
M
m
realm
M
m
simm
realm
objective
yM
yM
yyM
F
1
2
1
2
1
2
11
1
(22) 12
where realmy is the real data, sim
my is the simulation result from the model, and m=1, …, M is the number 13
of the data sample. 14
To calibrate car-following models, we use the NGSIM vehicle trajectory data collected on I-80 15
San Francisco, California on April 13, 2005 (19). Since the data includes lane change and large vehicle 16
gaps, we use the following criteria to extract the car-following datasets of the four combinations 17
separately from NGSIM data. 18
a. Each car-following group contains two vehicles (car-following pair). Car-following group 19
with more than three vehicles will be divided into multiple two-vehicle groups. 20
b. Each pair of car-following vehicles is formed and decomposed based on two spacing 21
thresholds, the engaging threshold DE = 130 ft (39.62 m) and the disengaging threshold DD 22
= 150 ft (45.72 m). These values are determined based on critical density at the experimental 23
site. Using two thresholds instead of one threshold can avoid unnecessary frequent grouping 24
and ungroup of car-following vehicle pairs caused by small fluctuation of spacing value 25
around a single threshold. 26
c. If a vehicle group decomposes due to spacing increase, lane changing or reaching the end of 27
the segment, it will not be exported as a valid sample. 28
TRB 2013 Annual Meeting Paper revised from original submittal.
Yang, Jin, Ran, Pu and Yang 9
The final datasets include the 477712, 25844, 16471, and 10105 trajectory points for the CC, CT, 1
TC, and TT combinations respectively. The calibrated car-following models are also inspected using 2
several error indexes including Mean Error (ME), Mean Absolute Error (MAE) and Mean Absolute 3
Relative Error (MARE) as follows, 4
M
yy
ME
M
m
simm
realm
1
)(
(23) 5
M
yy
MAE
M
m
simm
realm
1 (24) 6
M
yyy
MARE
M
m
realm
simm
realm
1
)(
(25) 7
where m indicates the mth sample, M is the total number of samples, realmy is the mth real data sample (y 8
can be the acceleration, velocity or position), and simmy is the mth simulation data sample. 9
To investigate the stability and characteristics of the car-truck heterogeneous traffic flow, a ring-10
road numerical simulation is conducted by placing 100 vehicles on a single-lane, flat, and ring road, with 11
the first vehicle following the end vehicle. In the simulations, the initial state of the traffic flow is 12
equilibrium state, add a small perturbation on the first vehicle, and then observe the perturbation 13
development in the platoon. Many studies adopted this simulation method to explore the fundamental 14
diagram (20, 21), shock wave (21, 22), linear stability (13, 23), and other properties (24) of traffic flow. 15
16
RESULTS ANALYSIS 17
Preliminary Data Analysis 18
Some preliminary analysis is conducted to obtain the general characteristics of the different car-truck 19
car-following combination types. First, the gap versus velocity relationship for each combination type is 20
explored. The CC combination has the smallest average gap, and the TT combination has the largest 21
average gap for the same velocity. When v = 17 m/s, the CT combination has the same gap with the TC 22
combination. The CT combination has a larger gap than the TC combination before v = 17 m/s, while the 23
TC combination has a larger gap after v = 17 m/s. Second, the desired velocities of cars and trucks are 24
also studied. The statistical results show that the truck has the maximum velocity 18 m/s, and the car has 25
the maximum velocity 30 m/s. The maximum velocity of the CT combination is around 18m/s which is 26
restricted by the maximum velocity of the leading truck. The maximum velocity found for the CT 27
combination is around 23 m/s with the combined impacts of the slow-moving follower truck and fast-28
TRB 2013 Annual Meeting Paper revised from original submittal.
Yang, Jin, Ran, Pu and Yang 10
moving leader car. The CC and TT combinations respectively had the maximum velocities of 30 m/s and 1
16 m/s. Third, the significant difference in response time for the different vehicle types suggested by (25) 2
can be measured using the method introduced by Aghabayk et al. (9). The average response time of the 3
truck drivers is found to be 2.1 s, and the average response time of the car drivers is around 1.3 s. 4
Calibration and Evaluation Results 5
Table 1 lists the calibration results of the heterogeneous IDM. The calibrated desired velocities of the 6
four combinations are similar to the preliminary statistical results. The TT combination has the largest 7
jam spaces ( 0ns and 1
ns ); while the CC combination has the smallest jam spaces. The order of the safe 8
time headway n is CC < CT < TC < TT. Cars have larger deceleration bn than trucks. The acceleration 9
exponents are same for all four types of combination. The values of the error indexes of acceleration, 10
velocity, and location for each combination type can also be found in Table 1. The acceleration 11
simulation errors (MARE) for the four combinations are all less than 15%, and the velocity and location 12
simulation errors (MARE) are all less than 10%. Therefore, the calibration results can reflect the car-13
following behavior characteristics of the four combinations in the I-80 NGSIM vehicle trajectory data. 14
TABLE 1 Calibration and evaluation results of the four combinations 15
Variables CC CT TC TT
Calibration Results
an (m/s2) 1.01 1.03 0.78 0.74
Vn (m/s) 27 19.3 20.6 17.7
0ns (m) 0.85 1.35 1.11 1.53
1ns (m) 0.19 0.27 0.12 0.36
n (s) 1.2 1.4 1.8 2.0
bn (m/s2) 2.26 2.12 1.70 1.61
n 4 4 4 4
Acceleration Evaluation
ME 0.84 0.28 -0.03 0.13
MAE 1.17 1.67 1.24 1.16
MARE 0.11 0.13 0.12 0.1
Velocity Evaluation
ME -0.25 0.15 -0.23 -0.04
MAE 1.31 1.62 1.2 0.75
MARE 0.08 0.09 0.07 0.05
Location Evaluation
ME 2.44 3.00 -1.93 1.23
MAE 9.71 9.6 5.75 5.61
MARE 0.05 0.06 0.03 0.04
TRB 2013 Annual Meeting Paper revised from original submittal.
Yang
2
Line3
The 6
stabi7
four 8
Indiv7
The 11
the c12
equi13
has t14
15
16
17
18
19
20
21
22
22
23
24
23
F24
g, Jin, Ran, P
ear Stabilit
key focus of
ility of the c
types of com
vidual Stabil
stability func
car-truck het
librium velo
the following
CTCC>
most uns
combina
stable tha
CC>CT>
CC com
combina
vehicle in
CC>TC>
combina
stable tha
FIGURE 1 S
Pu and Yang
ty Analysis
f the linear st
car-truck hete
mbination.
lity of the fou
ction SFi for
terogeneous
city for each
g three cases.
>TC>TT. Th
stable combi
ation effect is
an the car as
>TC>TT. Th
mbination bec
ation is more
n the combin
>CT>TT. Th
ation is more
an the car as
Stability fun
ability analy
erogeneous t
ur combinat
each combin
traffic flow.
h combinatio
.
his case happ
ination and
s close to the
the followin
his case occur
comes the m
e stable than
nation is mor
his case occu
stable than
the leading v
nctions of the
sis is to inve
traffic flow:
tions
nation type i
. FIGURE 1
n type. The r
pens before v
the TT com
e CT combin
ng vehicle.
rs in the mid
most unstabl
the car as th
re stable than
urs when the
the TC com
vehicle.
e four comb
stigate the im
the individu
is used to ev
1 displays th
relative orde
v 9 m/s. In
mbination is t
nation. The t
dle range of
le case. The
he leading v
n the car as th
velocity is m
mbination. Th
binations wit
mpacts of the
ual stability a
valuate its ind
he relationshi
er of the SF f
n this case, t
the most stab
truck as the
velocity from
e truck as th
vehicle, and t
he following
more than 11
he truck as th
th respect to
e two critical
and the prop
dividual stab
ip between S
for each com
the CT comb
ble combina
following ve
m 9 m/s to 1
he leading v
the truck as
vehicle.
1 m/s. In this
he leading ve
o the uniform
1
factors on th
ortions of th
bility effect o
SF versus th
mbination typ
bination is th
ation. The CC
ehicle is mor
1 m/s, and th
vehicle in th
the followin
s case, the C
ehicle is mor
m velocity.
1
he
he
on
he
pe
he
C
re
he
he
ng
T
re
TRB 2013 Annual Meeting Paper revised from original submittal.
Yang
5
truck6
car-t7
26) i8
17
comb18
follo19
comb20
stabi21
stabl22
= -123
comb24
(SFc25
alter26
Vehi27
comb28
18
19
20
Impa21
As s25
weig26
car-t27
valu28
g, Jin, Ran, P
From the
ks depends o
truck heterog
in which the
In the f
binations on
owed by 10
binations are
ility function
le. The stabil
.57, which m
binations wi
t + SFtc < 0)
rnating part),
icle No.41 (t
bination part
act of the pro
shown in Eq
ghted averag
truck combin
es of the fou
Pu and Yang
e above analy
on their leadi
geneous traffi
truck or car o
following ri
traffic flow,
cars, 30 tru
e 39%, 16%
n of the car-tr
lity functions
means that th
ll suppress th
). The shock
, ascends aft
the TT comb
t). Thus, the s
oportions of
quation (21)
e of the indi
nation types.
ur combinatio
ysis, the follo
ing or follow
fic flow. This
only has one
ing-road sim
, we fill the
ucks and 30
%, 16% and 2
ruck heterog
s of the four
he CC and C
he perturbati
k wave peak
ter the Vehic
bination part
simulation re
FIGURE
f the four com
), the stabili
ividual value
. Take the ca
ons from FIG
owing conclu
wing status in
s conclusion i
e single effect
mulation, to
ring road wi
cars. Theref
29%. Set the
eneous traffi
combination
CT combinati
ion, and the
(See Figure
cle No.31 (t
t), and ascen
esults are con
2 The profi
mbinations o
ity function
es SFcc, SFct,
ase v 6 m
GURE 1. Fig
usion can be
n car-followi
is more soph
t (stabilize or
demonstrat
ith 30 vehicl
fore, the pro
e equilibrium
ic flow F = -0
ns are SFcc = 0
ions can amp
car-truck alt
e 2) fluctuate
the CC comb
nds slightly a
nsistent with
ile of the sho
on stability
of the car-t
SFtc, and SF
m/s as an ex
gure 3 (a) is t
e drawn: the
ng and the e
histicated than
r destabilize)
te the differ
es alternating
oportions of
m velocity a
0.410 indicat
0.28, SFct = 0
plify the per
ternating can
es and desce
bination part
again after th
the analytica
ock wave pea
truck heterog
Ftt according
ample and f
the scatter pl
stability effe
equilibrium v
n the previou
) on the traffi
rent effects
g between ca
the CC, CT
as 4 m/s. In
ting that the
0.43, SFtc = -
rturbation, th
n decrease the
ends at first
t), and desce
he Vehicle N
al analysis.
ak.
geneous traf
g to the prop
find the corre
ot of the stab
12
ect of cars an
velocity of th
us studies (10
fic flow.
of the fou
ars and truck
T, TC and T
this case, th
traffic flow i
-0.83, and SF
he TC and T
e perturbatio
(the car-truc
ends after th
No.71 (the CC
ffic flow is
portions of th
esponding SF
bility functio
2
nd
he
0,
ur
ks
T
he
is
Ftt
T
on
ck
he
C
a
he
SF
on
TRB 2013 Annual Meeting Paper revised from original submittal.
Yang
valu11
Equa12
FIGU13
zero 14
area 15
Whe16
the 17
How18
four 19
func20
12
13
FIG15
16
Fun16
We e27
equa28
CC, 29
critic30
fund31
show32
besid33
FIGU34
inter35
incre36
heter37
g, Jin, Ran, P
es of the car
ation (21), th
URE 3 (a) ar
stability fun
above the li
en the propor
traffic flow
wever, when
combination
ction values a
GURE 3 Sc
ndamental D
explore the f
ation (11). W
CT, TC and
cal factor to
damental diag
wn in FIGUR
de the legen
URE 4 (b) th
rsection and
ease with th
rogeneous tra
Pu and Yang
r-truck hetero
he proportion
re the stabilit
nction values
ine and the s
rtion of the C
is always u
v 1 m/s a
ns are stable.
are always les
(a)
catter plots o
Diagram A
fundamental
When examin
TT, we find
determine th
grams with
RE 4 (a). FI
nds are the p
he curves hav
increases af
he increasing
affic flow mo
ogeneous tra
ns of CC and
ty function v
s. The neutra
stable area un
CC combinat
unstable reg
as shown in
. Thus, no ne
ss than zero.
of the stabili
The case v*
Analysis
diagram cha
ing the fund
d that the diff
he shape and
similar Pd v
GURE 4 (b)
roportions o
ve an interse
fter the inter
g of Pd, wh
oves towards
affic in the sp
d TT can dete
alues, and th
al stability li
nder the line
tion is more
gardless of t
FIGURE 3
eutral stabilit
ity function
= 1 m/s. (b)
aracteristic of
amental diag
ference betw
d location of
values locate
) illustrates t
of the four co
ction, and th
rsection. Fur
hich reveals
s the high flo
pace of the C
ermine the pr
he line is the
ne splits the
e. (0, 0.22) a
than 0.5 or t
the proportio
(b), the four
ty line exists
of the car-tr
) The case v*
f the car-truc
gram curves
ween the CC a
f the overall
e close to ea
the cluster fo
ombinations
he flow decre
rthermore, th
s that the m
ow and high d
CC and CT p
roportions of
neutral stabi
quadrant in
and (0.5, 0) a
the CT comb
ons of the o
r SF are all n
s in this case
(
ruck heterog
* = 6 m/s.
ck heterogene
for all propo
and TT comb
fundamental
ach other on
for the case P
CC, CT, TC
ases with the
he flow rate
most unstabl
density area w
proportions (
f CT and TC
lity line conn
nto two areas
are the two c
bination is m
other three c
negative, wh
since the ov
(b)
geneous traf
eous traffic f
ortion combin
binations Pd
l diagram. It
n the flow-de
Pd = 0.2 and
C and TT re
e increase of
and critical
le point of
with the incr
13
(According t
C). The dots i
necting all th
s, the unstabl
critical point
ore than 0.22
combination
hich means a
verall stabilit
ffic flow. (a)
flow based o
nations of th
= Pcc - Ptt is
t is found tha
ensity plot a
d the number
espectively. I
Pcc before th
l density bot
the car-truc
easing of Pd.
3
to
in
he
le
s.
2,
s.
all
ty
on
he
a
at
as
rs
In
he
th
ck
.
TRB 2013 Annual Meeting Paper revised from original submittal.
Yang
2 3
4
FIG6
7
Sho7
The 16
diffe17
prop18
view19
boun20
traff21
truck22
flow23
shoc24
17
18
FIG20
21
g, Jin, Ran, P
GURE 4 Fu
ck Waves A
propagation
erences betw
pagations in t
wpoint. The a
ndaries betw
fic flow is th
k heterogene
w because the
ck wave trave
x
GURE 5 Pro
Pu and Yang
(a)
undamental d
diagr
Analysis
speed of the
ween cars and
the pure car t
arrow indicat
ween two diff
he same durin
eous traffic fl
e truck needs
els faster in th
(a)
opagation of
flow
)
diagrams of
ams. (b) Fun
e shock wav
d trucks. FIG
traffic flow a
tes the propa
ferent states
ng propagati
flow and the
s more time t
he pure car tr
)
f the shock w
w. (b) The ca
f the car-tru
ndamental d
ve is not unif
GURE 5 (a)
and the car-tr
agation trajec
of traffic flo
ion. Howeve
average slop
to react to th
raffic flow th
t
wave from th
ar-truck het
ck heterogen
diagram clus
form through
and (b) resp
ruck heterog
ctory of a per
ow. The slop
er, the slope
pe decreases
he speed chan
han in the car
x
he trajectory
terogeneous
(b)
neous traffic
ster at Pd =0
h the platoon
pectively dem
eneous traffi
rturbation wh
pe of the sh
changes wh
in the car-tr
nge of its pre
r-truck hetero
(b)
y viewpoint.
traffic flow.
c flow. (a) F
0.2.
n due to the r
monstrate the
ic flow from
hich gives th
hock wave in
en going thr
ruck heterog
eceding vehi
ogeneous tra
t
. (a) The pur
.
14
undamental
response tim
e shock wav
the trajector
he shock wav
n the pure ca
rough the car
eneous traffi
icle. Thus, th
affic flow.
re car traffic
4
l
me
ve
ry
ve
ar
r-
ic
he
c
TRB 2013 Annual Meeting Paper revised from original submittal.
Yang
8
prop9
traff10
and 11
wave12
heter13
traff14
9
10
FIG12
13
13
CON14
The 23
pape24
car-f25
car-f26
and 27
car-f28
heter29
char30
shoc31
28
using29
of th30
Furth31
espe32
g, Jin, Ran, P
FIGURE
pagation in bo
fic flow, and
30 trucks. G
e takes aroun
rogeneous fl
fic flow can s
GURE 6 Sim
NCLUSION
traffic flow
er, the focus
following-tru
following mo
use the NGS
following mo
rogeneous tr
acteristics of
ck wave.
We deriv
g the ring-ro
he four comb
hermore, the
ecially with th
Pu and Yang
E 6 illustrate
oth pure car
the traffic flo
iving the sam
nd 250 s to p
low, it takes
slow down th
(
mulation of t
(
N
mixed by ca
is the analys
uck (CT), tru
odels. We ex
SIM data to c
odels can ref
raffic flow.
f the car-truc
ve the linear
ad simulation
inations, are
e individual
he increasing
es the shock
and mixed c
ow in FIGUR
me perturbati
propagate to
around 400
he shock wav
a)
the propaga
(b) The car-t
ars and trucks
sis of the imp
ck-following
xtend the orig
calibrate mod
flect the car-
Based on
ck heterogene
stability crit
ns. Two criti
found to det
l stability fu
g of equilibri
k wave propa
car-truck flow
RE 6 (b) is th
ion to the tw
the last veh
s. The simu
ve.
ation speed o
truck hetero
s is the typic
pacts of the f
g-car (TC) an
ginal Intellig
del paramete
-following dy
the calibrat
eous traffic f
erion of the c
ical factors, t
termine the s
unction vari
ium velocity,
agation obse
w. The traffic
he car-truck h
wo types of tra
icle in the pu
ulations illus
of the shock
ogeneous tra
cal heterogen
four types of
nd truck-follo
gent Driver M
ers. The eval
ynamics of t
ted car-follo
flow: the line
car-truck het
the individua
stability of th
es with the
, TC become
erved in a s
c flow in FIG
heterogeneou
affic flow at
ure car traffi
strate that the
(
wave. (a) Th
affic flow.
neous traffic
f combination
owing-truck (
Model (IDM)
luation result
the four com
owing mode
ear stability, f
terogeneous t
al stability fun
he car-truck h
e increasing
es less stable
simulation of
GURE 6 (a) i
us traffic mix
the same pla
ic flow; whil
e car-truck h
b)
he pure car
flow in real
n, car-follow
(TT), using h
) to a hetero
ts show that
mbinations in
el, we explo
fundamental
traffic flow a
nction and th
heterogeneou
of equilibri
than CT. In
15
f perturbatio
is the pure ca
xed by 70 car
ace, the shoc
le in car-truc
heterogeneou
traffic flow.
traffic. In thi
wing-car (CC
heterogeneou
ogeneous form
the calibrate
n the car-truc
ore the thre
diagrams an
and validate
he proportion
us traffic flow
ium velocity
addition, car
5
on
ar
rs
ck
ck
us
.
is
C),
us
m
ed
ck
ee
nd
it
ns
w.
y;
rs
TRB 2013 Annual Meeting Paper revised from original submittal.
Yang, Jin, Ran, Pu and Yang 16
and trucks can both stabilize and destabilize the car-truck heterogeneous traffic flow and their effects 1
depend on their roles (leading vehicle or following vehicle) in car-following and the equilibrium velocity 2
of the car-truck heterogeneous traffic flow. This conclusion is more sophisticated than the previous 3
studies (10, 26) in which the truck or car only has one single effect (stabilize or destabilize) on the traffic 4
flow. We also investigate the impact of the proportions of the four combination types by simulation. The 5
results show that the proportions of the four combinations may not have significant influence on the 6
stability of traffic flow with low equilibrium velocity (e.g. on an extremely congested road); however, in 7
most cases, the neutral stability lines can be found for the car-truck heterogeneous traffic flow. 8
Fundamental diagrams of the car-truck heterogeneous traffic flow are determined by the 9
distance headways and the proportions of the four types of combination. The simulations further reveal 10
that the fundamental diagrams with the same proportion difference between the CC and TT 11
combinations cluster together. The flow rate and the critical density increase with the increasing of the 12
proportion difference between the CC and TT. Finally, the shock wave analysis indicates that trucks can 13
slow down the propagation speed of the shock wave due to their longer response times. 14
15
ACKNOWLEDGMENT 16
This work was supported by the National Natural Science Foundation of China (Grant No. 17
51278429 and Grant No.50908195) and the National High Technology Research and Development (863) 18
Program of China (Grant No. 2011AA110403). 19
20
REFERENCES 21
1. Huddart K. and R. Lafont. Close driving---Hazard or necessity? In XV European annual conference 22
on human decision making and manual control. CD-ROM. TNO, Netherlands, 1990. 23
2. McDonald M., M. Brackstone, B. Sultan and C. Roach, Close following on the motorway: initial 24
findings of an instrumented vehicle study. VISION IN VEHICLES-VII, 1999. 25
3. Sayer J. R., M. L. Mefford and R. Huang. Ann Arbor. The effect of lead-vehicle size on driver 26
following behavior. 2000. 27
4. Peeta S., P. Zhang and W. Zhou, Behavior-based analysis of freeway car-truck interactions and related 28
mitigation strategies. Transportation Research Part B: Methodological, Vol.39, No.5, 2005, pp.417-451. 29
5. Peeta S., W. Zhou and P. Zhang. Modeling and mitigation of car-truck interactions on freeways. In 30
Transportation Research Record: Journal of the Transportation Research Board, No.1899, 31
Transportation Research Board of the National Academies, Washington, D.C., 2004, pp.117-126. 32
TRB 2013 Annual Meeting Paper revised from original submittal.
Yang, Jin, Ran, Pu and Yang 17
6. TRB. Highway capacity manual. Transportation Research Board, National Research Council, 1
Washington, DC, 2000. 2
7. Ye F. and Y. Zhang. Vehicle type-specific headway analysis using freeway traffic data. In 3
Transportation Research Record: Journal of the Transportation Research Board, No.2124, 4
Transportation Research Board of the National Academies, Washington, D.C., 2009, pp.222-230. 5
8. Sarvi M., Heavy commercial vehicles-following behavior and interactions with different vehicle 6
classes. Journal of Advanced Transportation, 2011. 7
9. Aghabayk K., M. Sarvi, W. Young and Y. Wang. Investigating heavy vehicle interactions during the 8
car following process. In Transportation Research Board 91st Annual Meeting. CD-ROM. 9
Transportation Research Board of the National Academies, Washington DC, 2012. 10
10. Mason A. D. and A. W. Woods, Car-following model of multispecies systems of road traffic. 11
Physical Review E, Vol.55, No.3, 1997, pp.2203-2214. 12
11. Treiber M., A. Hennecke and D. Helbing, Congested traffic states in empirical observations and 13
microscopic simulations. Physical Review E, Vol.62, No.2, 2000, pp.1805-1824. 14
12. Treiber M., A. Kesting and D. Helbing, Delays, inaccuracies and anticipation in microscopic traffic 15
models. Physica A, Vol.360, No.1, 2006, pp.71-88. 16
13. Wilson R. and J. Ward, Car-following models: fifty years of linear stability analysis–a mathematical 17
perspective. Transportation Planning and Technology, Vol.34, No.1, 2011, pp.3-18. 18
14. Hamdar S. H. and H. S. Mahmassani. From existing accident-free car-following models to colliding 19
vehicles: exploration and assessment. In Transportation Research Record: Transportation Research 20
Record, No.2088, Transportation Research Board of the National Academies, Washington, D.C., 2008, 21
pp.45-56. 22
15. Herman R., E. W. Montroll, R. B. Potts and R. W. Rothery, Traffic Dynamics: Analysis of Stability 23
in Car Following. Operations Research, Vol.7, No.1, 1959, pp.86-106. 24
16. Ward J. A., Heterogeneity, Lane-Changing and Instability in Traffic: A Mathematical Approach. 25
Ph.D. Thesis. University of Bristol, 2009. 26
17. Brockfeld E., R. D. Kühne and P. Wagner. Calibration and validation of microscopic models of 27
traffic flow. In Transportation Research Record: Transportation Research Record, No.1934, 28
Transportation Research Board of the National Academies, Washington, D.C., 2005, pp.179-187. 29
18. Punzo V. and F. Simonelli. Analysis and Comparison of Microscopic Traffic Flow Models with Real 30
Traffic Microscopic Data. In Transportation Research Record: Transportation Research Record, 31
No.1934, Transportation Research Board of the National Academies, Washington, D.C., 2005, pp.53-63. 32
19. NGSIM data set. Federal Highway Administration. http://www.ngsim.fhwa.dot.gov. Accessed 33
December 1, 2010. 34
TRB 2013 Annual Meeting Paper revised from original submittal.
Yang, Jin, Ran, Pu and Yang 18
20. Kerner B. S. and P. Konhäuser, Structure and parameters of clusters in traffic flow. Physical Review 1
E, Vol.50, No.1, 1994, pp.54-83. 2
21. Wang J., R. Liu and F. Montgomery. Car-following model for motorway traffic. In Transportation 3
Research Record: Journal of the Transportation Research Board, No.1934, Transportation Research 4
Board of the National Academies, Washington, D.C., 2005, pp.33-42. 5
22. Nagatani T., Density waves in traffic flow. Physical Review E, Vol.61, No.4, 2000, pp.3564-3570. 6
23. Wilson R. E., An analysis of Gipps's car-following model of highway traffic. IMA journal of applied 7
mathematics, Vol.66, No.5, 2001, pp.509-537. 8
24. Hoogendoorn R. G., S. Hoogendoorn, K. A. Brookhuis and W. Daamen. Anticipation and Hysteresis: 9
Parameter Value Changes and Model Performance in Simple and Multianticipative Car-Following 10
Models. In Transportation Research Board 90th Annual Meeting. CD-ROM. Transportation Research 11
Board of the National Academies, Washington DC, 2011. 12
25. Rothery R. W.: Traffic Flow Theory. In: Traffic Flow Theory. Edited by N. H.Gartner CM, and A. K. 13
Rathi. Washington.D.C: Transportation Research Board; 1998. 14
26. Jin S., D. H. Wang, Z. Y. Huang and P. F. Tao, Visual angle model for car-following theory. Physica 15
A Statistical Mechanics and its Applications, Vol.390, 2011, pp.1931-1940. 16
TRB 2013 Annual Meeting Paper revised from original submittal.