Page 1

Collision Avoidance Analysis for Lane Changing and

Merging
1
H. Jula, E. Kosmatopoulos, P. Ioannou
USC Center for Advanced Transportation Technologies
Department of Electrical Engineering-Systems
University of Southern California,
3740 McClintock Ave., EEB-200
Los Angeles, CA 90089-2562
{jula | kosmatop | ioannou}@usc.edu
Abstract
One of the riskiest maneuvers that a driver has to perform in a conventional highway system is to
merge into the traffic and/or to perform a lane changing maneuver. Lane changing/merging
collisions are responsible for one-tenth of all crash-caused traffic delays often resulting in
congestion. Traffic delays and congestion, in general, increases travel time and has a negative
economic impact.
In this paper, we analyze the kinematics of the vehicles involved in a lane changing/merging
maneuver, and study the conditions under which lane changing/merging crashes can be avoided.
That is, given a particular lane change/merge scenario, we calculate the minimum longitudinal
spacing which the vehicles involved should initially have so that no collision, of any type, takes
place during the maneuver.
Simulations of a number of examples of lane changing maneuvers are used to demonstrate the
results. The results of this paper could be used to assess the safety of lane changing maneuvers
and provide warnings or take evasive actions to avoid collision when combined with appropriate
hardware on board of vehicles.
Keywords: Lane Changing, Lane Merging, Crash Avoidance, Minimum Safety Spacing (MSS).
1. Introduction
The inter-vehicle spacing or headway affects both safety and highway capacity. For collision free
1. This work is supported by the California Department of Transportation through PATH of the University
of California. The contents of this paper reect the views of the authors who are responsible for the facts
and the accuracy of the data presented herein. The contents do not necessarily reect the ofcial views or
policies of the State of California. This paper does not constitute, a standard, specication, or regulation.
Page 2
vehicle following, the spacing should be large enough in order to guarantee no collisions during
all possible vehicle maneuvers. Lane changing/merging accidents consist of various types of
vehicle collisions, such as rear-end collisions, single vehicle road departure accidents, side-wipe,
and angle collisions. Lane changing/merging collisions constituted about 4.0% of all police-
reported collisions in 1991, and accounted for about 0.5% of all fatalities [1].
Although the lane change crash problem is small relative to other types of crashes and does not
account for a high percentage of traffic fatalities, this crash type is responsible for one-tenth of
all crash-caused traffic delays often resulting into congestion. Traffic delays and congestion in
general increases travel time and has a negative economic impact [2]. In practice, the possibility
of merging collisions can be reduced by adjusting relative velocities and increasing the
longitudinal inter-vehicles spacing. Since roadway capacity is proportional to vehicle speed and
inversely proportional to longitudinal inter-vehicle spacing, a large reduction in speed or a large
increase in spacing leads to a low capacity highway system. For a high capacity highway system,
the headway setting should be as small as possible. Since safety cannot be easily traded off, the
choice of Minimum Safety Spacing (MSS) between vehicles for a collision free environment is
important both from safety and capacity point of view.
Godbole, et. al. [3], and Shiller and Sundar [4] analyzed the obstacle avoidance problem. They
considered the problem of collision between a moving vehicle and an existing static obstacle,
such as a disabled vehicle or a large object. For avoiding the obstacle, two approaches were
considered: stopping in the same lane, or performing a lane changing maneuver. Shiller and
Sunder [4] developed what they called clearance and stopping curves that specify the collision
avoidance maneuver. These curves divide the phase-plane (initial longitudinal velocity of vehicle
under consideration versus its distance to the obstacle plane) into three regions. In region I the
vehicle performs normal (non-emergency) full stop or normal lane changing maneuver. In region
II the vehicle performs normal lane changing maneuver only, and in region III the vehicle
performs full stop maneuver in the same lane. When the vehicle is in Region I, lane changing has
priority over full stop if the lane changing is feasible. The vehicle is in Region III when a
collision is imminent. In that case it is shown that an emergency stop in the same lane will
reduce the severity of collision and is therefore preferable compared to lane changing. Godbole,
et. al. [3] formulated the lane changing maneuver for obstacle avoidance problem as an optimal
control problem. For normal lane changing the collision avoidance problem is formulated as an
optimization problem where the time for completing the lane changing maneuver is minimized
by selecting the appropriate lateral and longitudinal control inputs. In an emergency lane
changing situation the lateral and longitudinal control inputs are calculated to minimize the
longitudinal distance between the vehicle and the obstacle. Both aforementioned works, [3] and
[4], do not take into account the vehicles in neighboring lanes, which could be obstacles to lane
changing.
Bascunna [5] determined the conditions for safe and unsafe lane changing. The conditions were
obtained for two vehicles involved in the lane changing maneuver. Initially the vehicles are
travelling in two neighboring lanes and vehicle 1 performs a merging maneuver to the other lane.
He considered four cases. Case I: The initial longitudinal velocity of vehicle 1 is less than that of
vehicle 2, and vehicle 1 intends to complete the lane changing with constant longitudinal
velocity, and then follows vehicle 2 . Case II: The initial longitudinal velocity of vehicle 1 is less
Page 3
than that of vehicle 2, and vehicle 1 intends to complete the lane changing by applying constant
longitudinal acceleration and leads vehicle 2 . Case III: The initial longitudinal velocity of
vehicle 1 is greater than that of vehicle 2, and vehicle 1 intends to complete the lane changing
with constant longitudinal velocity and leads vehicle 2. Case IV: The initial longitudinal velocity
of vehicle 1 is greater than vehicle 2, and vehicle 1 intends to complete the lane changing with
constant longitudinal deceleration and then follows vehicle 2 .
In this paper, we examine the problem of safe lane changing and merging maneuvers in highway
systems. By analyzing the kinematics of the vehicles involved in a lane changing or merging
scenario, we present a general algorithm to calculate whether a particular lane changing/merging
maneuver is safe, i.e., free of collisions. We present a general algorithm for calculating the
Minimum Longitudinal Safety Spacing (MSS) for all vehicles involved. That is, given a
particular lane change/merging scenario, we calculate the minimum longitudinal spacings that
the vehicles, which affected by the lane changing manuever, should initially follow for no
collision to occur.
We examine special cases of lane changing/merging scenarios. For the cases where the merging
vehicle moves with either constant longitudinal velocity [5] or acceleration, we explicitly
calculate the MSSs and we show that the regions in the initial longitudinal spacing/ relative
longitudinal velocity plane can be divided into safe and unsafe regions; once the vehicles start
the lane changing/merging maneuver within the safe regions then collision-free maneuver is
guaranteed. We finally analyze the switching longitudinal acceleration case [6], i.e., the case
where the merging vehicle initially accelerates/decelerates with constant longitudinal
acceleration in order to create enough spacing for the lane changing/merging maneuver and then
it switches to another constant longitudinal acceleration/deceleration in order to adjust its
velocity with the velocity in the destination lane. For this case we show that the results obtained
for the constant acceleration case can be used in order to decide whether a particular lane
changing/merging scenario is collision-free or not. In particular, we use the method of
isoclines[7] to illustrate graphically the trajectory of the required lane changing maneuver that
could move the vehicle from the unsafe region in to the safe one.
2. Minimum Longitudinal Safety Spacing
Let us consider a lane changing situation where vehicle M in Figure 1 will move from its current
position between vehicles L
o
and F
o
to a new position between vehicles L
d
and F
d
in the
neighboring lane. We refer to vehicles L
d
, F
d
, L
o
, F
o
and M as the leading vehicle in the
destination lane, following vehicle in the destination lane, leading vehicle in the originating lane,
following vehicle in the originating lane, and the vehicle which must perform the lane-changing
(which will be called thereafter the merging vehicle), respectively.
Without loss of generality, we assume that the merging vehicle, M, starts the lane-changing
maneuver at t=0. This maneuver consists of two parts. Initially M adjusts its longitudinal velocity
and spacing for a time-interval t
adj
, and then applies lateral acceleration to merge to the
destination lane. In other words, t
adj
is the required time for the merging vehicle to adjust its
Page 4
longitudinal position and velocity before it starts merging to the destination lane.
Fig 1: Pre-lane changing conguration showing position of merging vehicle M.
To measure the lateral and longitudinal positions of the vehicles involved in the maneuver, an
arbitrary origin, which is denoted by "O" in Figure 1, is selected. The axis y is directed to ward
the destination lane and the axis x is aligned with the lateral side of the merging vehicle which is
closer to the destination lane. Figure 1 shows the top view of the pre-lane changing configuration
of the vehicles involved in the lane changing maneuver where the destination lane is assumed to
be on the left of the originating lane; axes x and y is chosen correspondingly. The origin O and
the axes x, y in Figure 1 are assumed to be fixed at the beginning of the lane changing maneuver,
t=0, till the end of maneuver.
Hereafter, the longitudinal acceleration/deceleration, the longitudinal velocity, the longitudinal
position, and the lateral position of vehicle i will be denoted by a
i
(t), v
i
(t), x
i
(t), and y
i
(t),
respectively, where i{L
d
, F
d
, L
o
, F
o
, M}. More precisely, x
i
(t) and y
i
(t) are, respectively, the
longitudinal and lateral distances between the front-left corner of the vehicle i (e.g. denoted by
"P" for merging vehicle in Fig. 1) and the origin O.
With the exception of the merging vehicle, the lateral acceleration of all other vehicles is
assumed to be zero. In our analysis, we assume a simple but realistic model for the lateral
acceleration a
lat
(t) of the merging vehicle that is used to complete the lane change maneuver. It
can be modeled as a Sine function of time [8], and is given as follows:
(1)
In equation 1, H is the total lateral displacement for the merging vehicle, t
adj
is the time elapse
before lateral acceleration applies, and t
lat
is the total time, after t
adj
, needed to complete the lane
change. It should be noted that the lateral acceleration a
lat
(t), according to equation(1), is
positive within the first half of the lateral displacement, i.e. t<(t
lat
/2)+t
adj
, and negative in the
second half. Given a
lat
(t), the lateral velocity v
lat
(t), and lateral position y
lat
(t), of the front-left
corner of the merging vehicle M (denoted by "p" in Figure 1) can be computed, easily.
A "lane change crash" occurs when the merging vehicle M attempts to change its lane and strikes
or is struck by a vehicle in the adjacent lane. The model (1) is considered as an accurate model
F
d
L
d
F
o
L
o
x
y
o
M
P
a
lat
t ( )
2H
t
lat
2
-----------
2
t
lat
------- t t
adj
( )
. ,
| `
sin t
adj
t t
lat
t
adj
+
0 Otherwise

'

=
Page 5
for many simple lane change/merge maneuvers during which more than two-thirds of lane
change/merge crashes occur [2].
The objective of this section is to use the simple lane change model described above and the
longitudinal acceleration profiles of the five vehicles in Figure 1 to find the initial minimum
longitudinal spacing between M and each of the other vehicles such that during a specified time-
interval [0,T], no collision, of any type, occurs. The length of the time interval T denotes the time
under consideration. In the following subsections we develop analytical expressions for the
minimum longitudinal spacing between the merging vehicle M and all other vehicles involved, in
order to guarantee a lane change without collision.
2.1. Minimum Longitudinal Safety Spacing between M and L
d
Lets consider the vehicles M, merging vehicle, and L
d
, leading vehicle in destination lane,
during a lane change/merge maneuver as shown in figure 2. The merging vehicle, M, starts the
lane-changing maneuver at t=0 by adjusting its longitudinal position and velocity, and then
applying lateral acceleration at t=t
adj
according to (1). The type of collision between M and L
d
could be of angle, side-wipe, or rear-end collision.
Let S denote the initial lateral distance between the upper side of the merging vehicle and the
lower side of the vehicle L
d
. Since the leading vehicle L
d
remains in the destination lane, an
angle and/or a side-wipe collision may occur as M passes the line LS in figure 2; LS is the
tangent to the lower side of the leading vehicle L
d
. The front-left corner of M is the first point of
the merging vehicle which passes the line LS at the point C.
It should be noted that, since the lateral acceleration of the leading vehicle L
d
is zero, the lateral
position of L
d
, y
Ld
is constant.
Fig 2: The marginal collision point between the merging vehicle M and the leading vehicle L
d
.
Let t
C
+t
adj
be the time-instant at which the front-left corner of the merging vehicle is at the point
C in figure 2. The type of collision which may take place at or after this time-instant is angle,
side-wipe or rear-end collision. An angle or a side-wipe collision may occur during or after the
front-left corner of M passes the point C, i.e., at or after the time-instant t
C
+t
adj
. A rear-end
collision may occur after the vehicle M accomplishes the lane changing maneuver, i.e., after the
merging vehicle has completely merged. The time-instant t
C
+t
adj
can be found by solving the
L
d
M
C
h
H
S
W
L
d
LS
Page 6
following equation for t=t
C
+t
adj
,
(2)
where w
Ld
is the width of the leading vehicle L
d
. By taking all types of collisions mentioned
above into account, the condition for no collision between M and L
d
is given by:
(3)
where l
Ld
is the length of the leading vehicle L
d
, w
M
is the width of the merging vehicle M, and
(t) is the angle between the tangent of the lane changing trajectory at the point y
lat
(t) and the
horizontal axis. The last term in inequality (3), , is to prevent any angle collision
between any point on the front bumper of the merging vehicle M and the rear-right corner of
leading vehicle L
d
in the time interval [t
C
+t
adj
, t
lat
+t
adj
]. From the definition of , it follows:
(4)
Equation (4) indicates that the value of tan((t)) and consequently Sin((t)) can be evaluated at
each time instant based on the lateral and longitudinal velocity of the merging vehicle. The
maximum value of (t) and consequently the maximum value of Sin((t)) in inequality(3) will be
at the time instant t=t
C
+t
adj
. (Note that t
C
+t
adj
>(t
lat
/2)+t
adj
, i.e. a
lat
(t
c
+t
adj
)<0). Lets define
, then inequality (3) can be
simplified as follows:
(5)
Let Sr(t) be the longitudinal spacing between point P of vehicle M and the rear end (bumper) of
vehicle L
d
(Note that y
Ld
(t)=0). That is:
(6)
As long as the longitudinal spacing is greater than zero, i.e. Sr(t)>0 for tt
C
+t
adj
, no collision will
occur during the lane-changing maneuver. Based on (5), we can rewrite equation (6) as:
(7)
y
lat
t ( ) S y
Ld
w
Ld
= =
x
M
t ( ) x
Ld
t ( ) l
Ld
w
M
Sin t ( ) ( ) < t t
C
t
adj
+ T , [ ]
w
M
Sin t ( ) ( )
t ( ) ( ) tan
y
lat
t ( )
x
M
t ( )
------------------
y
lat
t ( ) t
x
M
t ( ) t
---------------------------
v
lat
t ( )
v
M
t ( )
--------------- = = =
l
L1
l
Ld
w
M
Max Sin t ( ) ( ) l
Ld
w
M
Sin t
C
t
adj
+ ( ) ( ) + + =
t
x
M
t ( ) x
Ld
t ( ) l
L1
< t t
C
t
adj
+ T , [ ]
Sr t ( ) x
Ld
t ( ) l
L1
x
M
t ( ) = t t
C
t
adj
+ T , [ ]
Sr t ( ) Sr 0 ( ) a
Ld
( ) a
M
( ) ( ) d v
Ld
0 ( ) v
M
0 ( ) ( )t + d
0

0
t

+
. ,
| `
0 > =
t t
C
t
adj
+ T , [ ]
Page 7
where . Our objective is to find the initial minimum value of
Sr(0) which guarantees no collision between the leading vehicle L
d
and the merging vehicle M.
The minimum value of Sr(0) is the minimum initial longitudinal relative spacing between L
d
and
M, for collision free vehicle merging and is denoted by MSS(L
d
,M). It is calculated using (7) as
follows,
(8)
From equation (8) it is clear that the minimum initial longitudinal relative spacing between L
d
and M, MSS(Ld,M), depends on the relative longitudinal acceleration, the relative initial
longitudinal velocity between the two vehicles, as well as the time interval [t
C
+t
adj
,T]. This time
interval depends on the lateral distance S, the lateral time t
lat
, and the adjustment time t
adj
.
2.2. Minimum Longitudinal Safety Spacing between M and F
d
:
Now, let us consider the merging vehicle, M, and the following vehicle in the destination lane,
F
d
, during a lane changing/merging maneuver as shown in figure 3. The merging vehicle, M,
starts its merging maneuver at t=0, and its lateral movement at t=t
adj
. The type of possible
collision between M and F
d
could be of angle, side-wipe, and rear-end collision.
Since the following vehicle F
d
is in the destination lane and has zero lateral motion, an angle
and/or a side-wipe collision may occur during or after the vehicle M passes the line LS in figure
3, where the line LS is the tangent to the lower side of the following vehicle F
d
. We define the
point C as the intersection between the rear-left corner of the vehicle M and the line LS;
obviously the point C is the marginal point that a collision between the two vehicles could occur.
Fig 3: The marginal collision point between the merging vehicle M and the following vehicle F
d
.
In this case we need to find the coordinates of the point C as well as the time-instant at which the
rear-left corner of the merging vehicle passes this point. In order to calculate the motion of the
rear-left corner of the merging vehicle, we apply first order approximation (tangent to the
vehicles path) to calculate the lateral position of the other corners of the merging vehicle as
Sr 0 ( ) x
Ld
0 ( ) l
Ld
x
M
0 ( ) =
MSS Ld M , ( )
t
Max a
M
( ) a
Ld
( ) ( ) d v
M
0 ( ) v
Ld
0 ( ) ( )t + d
0

0
t

. ,
| `
=
t t
C
t
adj
+ T , [ ]
F
d
M
h
H
S
W
F
d
C
LS
Page 8
follows:
(9)
where (t) is the angle between the tangent of the path at point y
lat
(t) and the horizontal axis, and
l
M
, w
M
are the length and width of the merging vehicle, respectively.
Let t
C
+t
adj
be the time-instant at which the rear-left corner of the merging vehicle is at point C in
Figure 3. Using equation (9), the time-instant t
C
+t
adj
can be found by solving the following
equation:
(10)
Where w
Fd
is the width of the following vehicle F
d
. Using equation (4), equation (10) can be
rewritten as,
(11)
Considering all possible types of collision, the condition for collision avoidance between M and
F
d
would be:
(12)
It should be noted that the maximum value of Cos((t)), in the time internal [t
C
+t
adj
,T], will be at
or after the time instant t=t
lat
+t
adj
, where the value of (t) is minimum and is equal to zero.
Therefore, equation (12) can be simplified as follows:
(13)
The above approximation results is a conservative condition for no collision condition during the
time interval t
C
+t
adj
t t
lat
+t
adj
, i.e. before M completes its lane changing maneuver. The
longitudinal spacing between the rear of the vehicle M and the front of the vehicle F
d
is given
by:
(14)
As long as the longitudinal spacing in (14) is positive, i.e. Sr(t)>0 for tt
C
+t
adj
, no collision
y
rear left
t ( ) y
lat
t ( ) l
M
Sin t ( ) ( )
y
front right
t ( ) y
lat
t ( ) w
M
Cos t ( ) ( )
y
rear right
t ( ) y
lat
t ( ) l
M
Sin t ( ) ( ) w
M
Cos t ( ) ( ) + ( )
y
lat
t ( ) l
M
Sin t ( ) ( ) S y
Fd
w
Fd
=
y
lat
t ( ) l
M
v
lat
t ( )
v
lat
2
t ( ) v
M
2
t ( ) +
--------------------------------------- S
x
Fd
t ( ) x
M
t ( ) l
M
Cos t ( ) ( ) < t t
C
t
adj
+ T , [ ]
x
Fd
t ( ) x
M
t ( ) l
M
< t t
C
t
adj
+ T , [ ]
Sr t ( ) x
M
t ( ) l
M
x
Fd
t ( ) = t t
C
t
adj
+ T , [ ]
Page 9
occurs. Based on (13), we can rewrite equation (14) as follows:
(15)
where . is the initial longitudinal relative spacing between F
d
and
M. For collision free vehicle merging, the minimum value of Sr(0) is denoted by MSS(M,F
d
) and
is calculated using (15) as follows:
(16)
From equation (16) it is clear that the minimum initial longitudinal relative spacing between F
d
and M, MSS(M,Fd), depends on the relative longitudinal acceleration, the relative initial
longitudinal velocity between the two vehicles, as well as the time interval [t
C
+t
adj
,T].
2.3. Minimum Longitudinal Safety Spacing between M and L
o
:
Consider now the case of the merging vehicle, M, and the leading vehicle in the originating lane,
L
o
, during the lane changing maneuver shown in figure 4. The vehicle M starts its merging
maneuver at t=0, and its lateral movement at t=t
adj
. The type of collision between M and L
o
could be of angle, or rear-end collision, but not side-wipe collision.
Since the leading vehicle L
o
remains in the originating lane, an angle collision may occur before
the front-right corner of M passes the line LS at point C in figure 4.
Fig 4: The marginal collision point between the merging vehicle M and the leading vehicle L
o
.
Since the lateral acceleration of the leading vehicle L
o
is zero, the lateral position of L
o
, y
Lo
, is
constant.
Sr t ( ) Sr 0 ( ) a
M
( ) a
Fd
( ) ( ) d v
M
0 ( ) v
Fd
0 ( ) ( )t + d
0

0
t

+
. ,
| `
0 > =
t t
C
t
adj
+ T , [ ]
Sr 0 ( ) x
M
0 ( ) l
M
x
Fd
0 ( ) =
MSS M Fd , ( )
t
Max a
Fd
( ) a
M
( ) ( ) d v
Fd
0 ( ) v
M
0 ( ) ( )t + d
0

0
t

. ,
| `
=
t t
C
t
adj
+ T , [ ]
Lo
M
h
H
W
L
o
C LS
Page 10
Let t
C
+t
adj
be the time-instant at which the front-right corner of the merging vehicle is at the
point C as shown in figure 4. Considering the first order approximation in (9), the time-instant
t
C
+t
adj
can be found by solving the following equation.
(17)
where w
M
is the width of the merging vehicle M. Using (4), equation (17) can be rewritten as:
(18)
Considering all types of collision, the condition for collision avoidance between M and L
o
is:
(19)
where l
Lo
is the length of the leading vehicle L
o
, and (t) is the angle between the tangent of the
path at point y
lat
(t) and the horizontal axis.
The last term in equation (19), , is to prevent any angle collision between any point
on the rear bumper of the leading vehicle L
o
and the front-right corner of the merging vehicle M
in the time interval [t
adj
, t
C
+t
adj
]. It should be noted that the maximum value of (t) and
consequently the maximum value of Sin((t)) in equation (19) will be at the time instant
t=t
C
+t
adj
. Lets define l
L1
=l
Lo
+Max(Sin((t))), then equation (19) can be rewritten as follows:
(20)
According to (20), the longitudinal spacing between the front of the vehicle M and the rear of
vehicle L
o
would be:
(21)
As long as the spacing Sr(t) is greater than zero, i.e. Sr(t)>0 for tt
C
+t
adj
, no collision occurs.
According to (20), we can rewrite equation (21) as:
(22)
The minimum value of Sr(0) is the minimum initial longitudinal relative spacing between L
o
and
M, for collision free vehicle merging and is denoted by MSS(Lo,M).
y
lat
t ( ) w
M
Cos t ( ) ( ) S y =
Lo

y
lat
t ( ) w
M
v
M
t ( )
v
lat
2
t ( ) v
M
2
t ( ) +
--------------------------------------- S
x
M
t ( ) x
Lo
t ( ) l
Lo
w
M
Sin t ( ) ( ) < t 0 t ,
C
t
adj
+ [ ]
w
M
Sin t ( ) ( )
x
M
t ( ) x
Lo
t ( ) l
L1
< t 0 t ,
C
t
adj
+ [ ]
Sr t ( ) x
Lo
t ( ) l
L1
x
M
t ( ) = t 0 t ,
C
t
adj
+ [ ]
Sr t ( ) Sr 0 ( ) a
Lo
( ) a
M
( ) ( ) d v
Lo
0 ( ) v
M
0 ( ) ( )t + d
0

0
t

+
. ,
| `
0 > =
t 0 t
C
, t
adj
+ [ ]
Page 11
(23)
Equation (23) indicates that MSS(Lo,M) depends on the relative longitudinal acceleration, the
relative initial velocity between the two vehicles, as well as the time interval [0,t
C
+t
adj
].
2.4. Minimum Longitudinal Safety Spacing between M and F
o
:
In this subsection, we consider the merging vehicle M, and the following vehicle F
o
in the
originating lane, during a lane change/merge maneuver as shown in figure 5. The merging
vehicle, M, starts its merging maneuver at t=0, and its lateral movement at t=t
adj
. The types of
possible collision between M and F
o
could be of angle, and rear-end collision, but not side-wipe
collision.
Since the following vehicle F
o
is in the originating lane and has zero lateral motion, an angle
and/or a rear-end collision may occur during or before the vehicle M passes the line LS in figure
5, where LS is the tangent to the upper side of the following vehicle F
o
.
We define the point C as the intersection between the rear-right corner of the vehicle M and the
line LS; obviously the point C is the marginal point that a collision between two vehicles may
occur.
Fig 5: The last possible collision point between the merging vehicle Mand the following vehicle Fo.
Since the following vehicle in originating lane, F
o
, has no lateral acceleration, the lateral position
of y
Fo
is constant. Let t
C
+t
adj
be the time-instant at which the rear-right corner of the merging
vehicle is at the point C in figure 5. Using the first order approximation in (9), the time-instant
t
C
+t
adj
can be found by solving the following equation.
(24)
MSS Lo M , ( )
t
Max a
M
( ) a
Lo
( ) ( ) d v
M
0 ( ) v
Lo
0 ( ) ( )t + d
0

0
t

. ,
| `
0 ,

' '

=
t 0 t ,
C
t
adj
+ [ ]
F
o
M
h
H
W
F
o
C
LS
y
lat
t ( ) l
M
Sin t ( ) ( ) w
M
Cos t ( ) ( ) + ( ) S y =
Fo

Page 12
Using (4), equation (24) can be rewritten as follows:
(25)
Considering all possible types of collision, the condition for no collision between M and F
o
is
given by:
(26)
It should be noted that the maximum value of Cos((t)), in the time internal [0,t
C
+t
adj
], will be
before the time instant t=t
adj
, where the value of (t) is minimum and is equal to zero. Therefore,
equation (26) can be simplified as follows:
(27)
Equation(27) results in a conservative condition for no collision during the time interval
[t
adj
,t
C
+t
adj
], i.e. after M starts its lane changing maneuver. According to (27), the longitudinal
spacing between the rear of the vehicle M and the front of the vehicle F
o
is given by:
(28)
As long as the longitudinal spacing in (28) is greater than zero, i.e. Sr(t)>0 for tt
C
+t
adj
, no
collision occurs. Based on (27), we can rewrite equation (28) as follows:
(29)
The minimum value of Sr(0) is the minimum initial longitudinal relative spacing between M and
F
o
, for collision free vehicle merging and is denoted by MSS(M,Fo). It is calculated using (29)
as follows:
(30)
From equation (30) it is clear that MSS(M,Fo) depends on the relative longitudinal acceleration,
the relative initial longitudinal velocity between two vehicles, as well as the time interval t
C
+t
adj
.
y
lat
t ( ) l
M
v
lat
t ( )
v
lat
2
t ( ) v
M
2
t ( ) +
--------------------------------------- w
M
+
v
M
t ( )
v
lat
2
t ( ) v
M
2
t ( ) +
---------------------------------------
. ,

| `
S
x
Fo
t ( ) x
M
t ( ) l
M
Cos t ( ) ( ) < t 0 t ,
C
t
adj
+ [ ]
x
Fo
t ( ) x
M
t ( ) l
M
< t 0 t ,
C
t
adj
+ [ ]
Sr t ( ) x
M
t ( ) l
M
x
Fo
t ( ) = t 0 t ,
C
t
adj
+ [ ]
Sr t ( ) Sr 0 ( ) a
M
( ) a
Fo
( ) ( ) d v
M
0 ( ) v
Fo
0 ( ) ( )t + d
0

0
t

+
. ,
| `
0 > =
t 0 t ,
C
t
adj
+ [ ]
MSS M Fo , ( )
t
Max a
Fo
( ) a
M
( ) ( ) d v
Fo
0 ( ) v
M
0 ( ) ( )t + d
0

0
t

. ,
| `
0 ,

' '

=
t 0 t ,
C
t
adj
+ [ ]
Page 13
This time interval depends on the lateral position y
Fo
, lateral time t
lat
, and adjustment t
adj
.
In all the expressions for MSSs, the value of MSS depends on the value of t
adj
, the time elapsed
before lateral acceleration is applied. During this time the merging vehicle M adjusts its
longitudinal position and velocity before it starts merging.
3. Special Cases and Simulations
In this section we consider specific profiles for the longitudinal acceleration of the vehicles
involved in lane changing maneuvers in order to derive closed form expressions for MSS. We
assume constant longitudinal velocity for all vehicles in Figure 1 except for M whose
longitudinal velocity may change when the lane changing/merging scenario starts. In the
following subsections we calculate the minimum longitudinal safety spacing MSS as a function
of the relative longitudinal velocity between the merging vehicle M and each of the other
vehicles in figure 1 by considering two cases:
Case I: The merging vehicle M performs the merging scenario with constant longitudinal
velocity. Obviously, the longitudinal velocity of the merging vehicle will remain the same as the
longitudinal velocity of the vehicles in the originating lane, i.e. the velocity before starting its
maneuver. Case II: The merging vehicle M applies a constant longitudinal acceleration/
deceleration in order to reach the longitudinal velocity of the vehicles in the destination lane after
the specific time t
long
+t
adj
.
In each case, and based on the relative longitudinal velocity and position between M and the
other vehicles, the merging vehicle M can determine whether the merging scenario is safe or
unsafe before the merging maneuver starts.
3.1. Constant Longitudinal Velocity
This is the case where all five vehicles are moving with constant longitudinal velocity, i.e. the
longitudinal acceleration for all vehicles is zero, a
i
(t)=0 for t[0, T] and i{L
d
, F
d
, L
o
, F
o
, M}.
The merging vehicle M keeps its longitudinal velocity constant throughout the merging
maneuver.
Minimum Longitudinal Safety Spacing between M and L
d
According to (7), the condition for collision avoidance between M and L
d
with constant
longitudinal velocity is:
(31)
Since the relative longitudinal velocity, v
M
- v
Ld
, is constant, the minimum initial longitudinal
safety spacing, MSS(Ld,M) is:
Sr t ( ) Sr 0 ( ) v
Ld
v
M
( )t + ( ) 0 > = t t
C
t
adj
T , + [ ]
Page 14
(32)
Minimum Longitudinal Safety Spacing between M and F
d
Equation (15) provides the condition for collision avoidance between M and F
d
. In the case of
constant longitudinal velocity, this condition is as follows:
(33)
And, the minimum initial longitudinal safety spacing, MSS(M,Fd) will be:
(34)
Minimum Longitudinal Safety Spacing between M and L
o
According to (22), the condition for collision avoidance between M and L
o
with constant
longitudinal velocity is as follows:
(35)
The minimum initial longitudinal safety spacing, MSS(Lo,M) will be:
(36)
Minimum Longitudinal Safety Spacing between M and F
o
According to equation (29), the condition for collision avoidance between M and F
o
with
constant longitudinal velocity is as follows:
(37)
The minimum initial longitudinal safety spacing, MSS(M,Fo) will be:
(38)
Simulation results:
Simulations are used to demonstrate equations (32), (34), (36), and (38). In these simulations, we
MSS Ld M , ( )
v
M
v
Ld
( ) T v
M
v
Ld
0
v
M
v
Ld
( ) t
C
t
adj
+ ( ) Otherwise

'

=
Sr t ( ) Sr 0 ( ) v
M
v
Fd
( )t + ( ) 0 > = t t
C
t
adj
T , + [ ]
MSS M Fd , ( )
v
Fd
v
M
( ) T v
Fd
v
M
0
v
Fd
v
M
( ) t
C
t
adj
+ ( ) Otherwise

'

=
Sr t ( ) Sr 0 ( ) v
Lo
v
M
( )t + ( ) 0 > = t 0 t ,
C
t
adj
+ [ ]
MSS Lo M , ( )
v
M
v
Lo
( ) t
C
t
adj
+ ( ) v
M
v
Lo
0
0 Otherwise

'

=
Sr t ( ) Sr 0 ( ) v
M
v
Fo
( )t + ( ) 0 > = t 0 t ,
C
t
adj
+ [ ]
MSS M Fo , ( )
v
Fo
v
M
( ) t
C
t
adj
+ ( ) v
Fo
v
M
0
0 Otherwise

'

=
Page 15
set the time T=50 sec, the adjustment time t
adj
=0 sec, the lateral time, in equation (1), t
lat
,=5 sec,
and the lateral displacement H=12 feet. Figures 6, 7, 8, and 9 show the initial relative
longitudinal spacings versus the relative longitudinal velocity between M and the other four
vehicles involved in the lane-changing maneuver. The solid lines (which will be called thereafter
safety margins) in these figures represent the margins between safe and unsafe lane changing
regions.
For positive relative velocity, v
M
-v
Ld
, in figure 6, the safety margin is a line with slope equal to
T=50. For negative relative velocity, it is approximately a line with slope equal to 2.8 which is
the value of t
C
+t
adj
in equation (2). Thus for constant longitudinal velocity, the safety margin
consists of two lines passing through the origin with different tangents. This conclusion complies
with equation (31).
A similar situation appears in figures 7 and 8 for the spacings between M and F
d
,and M and L
o
,
i.e. the safety margins consist of two lines with different tangents which are given by equations
(33) and (35). In the case of the spacing between M and F
o
in figure 9, one of the components of
the safety margin is not a straight line; the reason for this is due to the approximation we made in
equation (25). The smoother the lane changing trajectory is, the more accurate the approximation
is in (25).
.
Fig 6: The collision region between M
and L
d
, Case I: Constant velocity.
Fig 7: The collision region between M
and F
d
; Case I: Constant velocity.
Page 16
3.2. Switching Longitudinal Acceleration
In this subsection we will examine the acceleration profile proposed in [6] for lane changing.
According to [6] all vehicles in both lanes, except M, are moving with constant longitudinal
velocity, i.e. a
i
(t)=0 for i{L
d
, F
d
, L
o
, F
o
], while the longitudinal acceleration profile of the
merging vehicle, M, is the one plotted in Figure 10. More precisely, the merging vehicle initially
accelerates/decelerates with constant longitudinal acceleration a
adj
, in order to create enough
spacing with the rest four vehicles in Figure 1. At the time-instant t
adj
, the merging vehicle starts
merging and it switches its longitudinal acceleration to a
M
; the merging vehicle continues to
accelerate with acceleration a
M
, until its longitudinal velocity becomes equal to the velocity of
the vehicles in the destination lane at the time-instant t
adj
+t
long
. After this time-instant the
merging vehicles longitudinal acceleration becomes zero.
In the following discussions and in order to get some insight about the problem, we first analyze
the case where t
adj
=0 and then the more general case t
adj
>0.
Fig 10: The switching longitudinal accelerating prole of merging vehicle, M.
Minimum Longitudinal Safety Spacing between M and L
d
I) t
adj
=0
In this case, the velocity of all vehicles except M is constant while the velocity of the vehicle M
become equal to the velocity of L
d
at the time instant t
long
and remain constant thereafter.
Fig 8: The collision region between M
and L
o
; Case I: Constant velocity.
Fig 9: The collision region between M
and F
o
; Case I: Constant velocity.
time
a
c
c
e
l
e
r
a
t
i
o
n
t
adj
t
adj
+t
long
a
M
a
adj
Page 17
Therefore, the value of the longitudinal acceleration of the merging vehicle, a
M
, is as follows:
(39)
According to equation (7) the condition for collision avoidance between M and L
d
is:
(40)
Considering different values of initial relative longitudinal velocity, v
M
(0) - v
Ld
, the minimum
initial longitudinal safety spacing, MSS(Ld,M) can be obtained analytically as follows:
(41)
Simulations are used to demonstrate the result in equation (41). In these simulations, we set
parameters for lane changing profile as those in section 3.1; and t
long
=10 sec. The safe and
unsafe region is shown in figure 11.
Fig 11: The collision region between M and L
d
; Case II: switching longitudinal acceleration.
Figure 11 shows the initial relative longitudinal distance versus the relative longitudinal velocity
between M and L
d
. The solid curve is the safety margin between safe and unsafe regions. For
positive relative velocities, v
M
(0)-v
Ld
, the safety margin corresponds to a line with slope equal to
t
long
/2=5 while for negative relative velocities, it is a line with tangent equal to 2.8 which is the
value of t
C
+t
adj
in equation (2).
a
M
v
Ld
v
M
0 ( )
t
long
----------------------------- t t
long

0 Otherwise

'

=
Sr t ( ) Sr 0 ( ) v
Ld
v
M
0 ( ) ( ) t
t
2
2t
long
--------------
. ,

| `
+
. ,

| `
0 > = t t
C
t
long
, [ ]
MSS Ld M , ( )
v
M
0 ( ) v
Ld
( ) t
long
2 v
M
0 ( ) v
Ld
0
v
M
0 ( ) v
Ld
( ) t
C
Otherwise

'

=
Page 18
Comparison between figures 6 and 11 indicates that the safe region has been expanded.
Therefore the switching longitudinal acceleration policy with t
adj
=0 is more reliable than the
constant longitudinal velocity policy for the case of vehicles M and L
d
.
II) t
adj
>0:
For the case of t
adj
=0 even if initially the two vehicles relative spacing and velocity is in the
unsafe region, it may happen that no collision occurs due to the switching acceleration policy.
The initial relative spacing and velocity between the vehicles M and L
d
, defines a starting point
in figure 11. If this point happens to be in the safe area, there is no need to apply any adjustment
acceleration, a
adj
to the merging vehicle, since the lane changing will be safe. If the initial
relative spacing and velocity of the two vehicles belong to the unsafe region, then we need to
apply the switching acceleration policy, in order to achieve appropriate relative spacing and
velocity before the merging vehicle starts merging. In other words, we want to move from the
unsafe region into the safe region and then start merging. Lets now define the following state
space variables:
(42)
By differentiating the above variables with respect to time, it is easy to see that the following
equations are valid,
(43)
Using the technique of isoclines [7], and solving the differential equation in (43), we obtain the
isoclines as follows:
(44)
The constant c is the integration constant which depends on the initial values, x
1
(0) and x
2
(0).
Figure 12 shows the isoclines that correspond to different values of a
adj
for the simulation in
figure 11. The initial state (initial relative spacing and velocity) has been chosen to be in the
unsafe region. Applying negative a
adj
, it is possible to move into the safe region in order to start
the merging maneuver. The larger the absolute value of a
adj
is, the faster we move into the safe
region. The minimum value of t
adj
for each a
adj
is determined by the point of intersection
between the corresponding isocline curve of a
adj
and the safety margin in figure 12.
x
1
x
Ld
x
M
l
L1
=
x
2
v
M
v
Ld
=
x
1
x
Ld
x
M
v
Ld
v
M
x
2
= = =
x
2
v
M
a
adj
= =
x
1
x
2
2
2a
adj
------------ c + =
Page 19
Fig 12: Applying a
adj
to move from unsafe area into safe area.
It should be noted that, a
adj
is limited by the acceleration/braking capabilities of the vehicle.
Moreover, in order to maintain passenger comfort, the must be less than a value say a
comf
which defines the maximum acceleration that maintains passenger comfort [6]. In addition, large
results in larger "shock wave" propagating down the originating lane as well as larger
minimum initial longitudinal relative spacing Sr(0), which in turn reduces the throughput of the
highway.
In the above discussion we assumed that we can apply a
adj
for a period of t
adj
seconds. This may
not be always feasible because the vehicles velocity may exceed its limits or, even worse, we
may have the situation where the velocity of the vehicle becomes zero. In this case the
acceleration profile of figure 10 cannot be applied; however, one can use the modified
acceleration profile shown in figure 13.
Fig 13: The modied switching longitudinal accelerating prole of merging vehicle, M.
Constant velocity in the interval [t
a
,t
adj
] will help the merging vehicle to create enough relative
spacing in order to enter the safe region. In this case, and for the time-interval [t
a
,t
adj
], the state
space equations can be rewritten as:
(45)
a
adj
a
adj
time
a
c
c
e
l
e
r
a
t
i
o
n
t
adj
t
adj
+t
long
t
a
a
M
a
adj
x
1
x
2
=
x
2
0 =
Page 20
which results in,
(46)
If we intend to increase the relative spacing, x
1
, we have to make sure that x
2
(t
a
) has a negative
value (negative relative acceleration). In other words, we have to make sure that, at t=t
a
, we are
in the left half plane in figure 11 and then we can set a
adj
equal to zero in order to move into the
safe region. Figure 14 shows the trajectory of moving from the unsafe into the safe region.
Fig 14: Applying modied a
adj
to move from unsafe area into safe area.
Minimum Longitudinal Safety Spacing between M and F
d
I) t
adj
=0:
Here, we assume that the longitudinal velocity of all vehicles except M is constant while the
longitudinal velocity of M will be increasing/decreasing until it becomes equal to the velocity of
F
d
at t=t
long
. Therefore the value of a
M
will be:
(47)
According to equation (15) the condition for collision avoidance between M and F
d
is:
(48)
And the minimum initial longitudinal safety spacing, MSS(M,Fd) is:
x
2
c x
2
t
a
( ) = =
x
1
x
2
t
a
( ) ( ) t t
a
( ) x
1
t
a
( ) + = t t
a
t
adj
, [ ]
a
M
v
Fd
v
M
0 ( )
t
long
----------------------------- t t
long

0 Otherwise

'

=
Sr t ( ) Sr 0 ( ) v
M
0 ( ) v
Fd
( ) t
t
2
2t
long
--------------
. ,

| `
+
. ,

| `
0 > = t t
C
t
long
, [ ]
Page 21
(49)
We use simulations to demonstrate the above results. The simulations parameters are set as those
in previous subsection (safety spacing between M and L
d
); The safe and unsafe regions are
shown in figure 15. In figure 15 for initial positive relative velocities, v
Fd
-v
M
(0), the safety
margin is a line with slope equal to t
long
/2=5. For negative relative velocities, it is a line with
tangent equal to 2.95 which is the value of t
C
+t
adj
in equation (10).
Fig 15: The collision region between F
d
and M; Case II: switching longitudinal acceleration.
Comparison between figures 7 and 15 indicates that the safe region has been expanded here, too.
II) t
adj
>0:
Similar to the case of vehicles F
d
and M, we define the state variables as follows:
(50)
and we differentiate them with respect to time, to obtain,
(51)
Therefore the isoclines are as follows:
(52)
Here c is a constant that depends on the initial values, x
1
(0) and x
2
(0). Figure 16 shows the
MSS M Fd , ( )
v
Fd
v
M
0 ( ) ( ) t
long
2 v
Fd
v
M
0 ( ) 0
v
Fd
v
M
0 ( ) ( ) t
C
Otherwise

'

=
x
1
x
M
x
Fd
l
M
=
x
2
v
Fd
v
M
=
x
1
x
M
x
Fd
v
M
v
Fd
x
2
= = =
x
2
v
M
a
adj
= =
x
1
x
2
2
2a
adj
------------ c + =
Page 22
isoclines that correspond to different values of a
adj
for the simulation in figure 15. The initial
state is chosen to be in the unsafe region. By applying positive a
adj
, it is possible to move into the
safe region in order to start the merging maneuver. The larger the value of a
adj
is, the faster we
move into the safe region. The minimum value of t
adj
for each a
adj
is determined by the point of
intersection between the isocline that corresponds to the a
adj
and the safety margin.
The longitudinal velocity of merging vehicle, v
M
, cannot exceed an upper bound, which is
determined by the capabilities of the vehicle and passenger comfort. Similar to the case of
vehicles L
d
and M, we may use the modified profile in figure 13, in the case where the
acceleration profile of figure 10 requires velocities that exceed the aforementioned upper bound.
Here again, we need to make sure that we are in the left half plane, x
2
<0, and then we may set
a
adj
=0. Figure 17 shows the trajectory of moving from the unsafe region into the safe region.
Minimum Longitudinal Safety Spacing between M and L
o
:
I) t
adj
=0:
The velocity of all vehicles except M is assumed to be constant while the velocity of M becomes
equal to the velocity of the destination lane after the time-instant t
long
and remains constant
thereafter, i.e. we assume v
M
(t)=v
Ld
=v
Fd
for t in the time interval [t
long
,T]. The value of a
M
is:
(53)
Fig 16: Applying a
adj
to move from
unsafe area into safe area.
Fig 17: Applying modied a
adj
to move
from unsafe area into safe area.
a
M
v
Lo
v
M
0 ( )
t
long
---------------------------- - t t
long

0 Otherwise

'

=
Page 23
Without loss of generality we assume that t
long
>t
C
. According to equation (22) the condition for
collision avoidance between M and L
o
is:
(54)
It should be noted that the last term in equation (54) is independent of v
M
, but depends on the
difference between the velocity of the vehicles in the originating and destination lanes. In order
to find MSS(L
0
,M) analytically, we define the following variables:
(55)
The above transformation is applicable provided that v
Ld
-v
M
(0) is not zero. If v
Ld
-v
M
(0) is zero,
the value of a
M
is zero, too. In this case, we can apply the results for the case of constant
velocity, (subsection 3.1). Let us also define the following function,
(56)
Using the above definitions we can analytically solve equation (69) as shown in Table 1.
Table 1: Analytical values of MSS(L
o
,M)
IF & IF & IF THEN
<0 v
Ld
<v
M
(0), v
Lo
>v
M
(0) MSS(Lo,M)= S(0)
" v
Ld
>v
M
(0), v
Lo
<v
M
(0) MSS(Lo,M)=S(t
C
)
>0 t
max
>t
C
v
Ld
<v
M
(0), v
Lo
<v
M
(0) MSS(Lo,M)=S(t
C
)
" " v
Ld
>v
M
(0), v
Lo
>v
M
(0) MSS(Lo,M)=S(0)
>0 t
max
<t
C,
t
max
>t
C
/2 v
Ld
<v
M
(0), v
Lo
<v
M
(0) MSS(Lo,M)=S(t
max
)
" " v
Ld
>v
M
(0), v
Lo
>v
M
(0) MSS(Lo,M)=S(0)
>0 t
max
<t
C
/2 v
Ld
<v
M
(0), v
Lo
<v
M
(0) MSS(Lo,M)=S(t
max
)
" " v
Ld
>v
M
(0), v
Lo
>v
M
(0) MSS(Lo,M)=S(t
C
)
Sr t ( ) Sr 0 ( ) v
Lo
v
M
0 ( ) ( ) t
t
2
2t
long
--------------
. ,

| `
v
Lo
v
Ld
( )
t
2
2t
long
-------------- + +
. ,

| `
0 > =
t 0 t
C
, [ ]

v
Lo
v
M
0 ( )
v
Ld
v
M
0 ( )
----------------------------- =
t
max
t
long
=
S t ( ) Max v
M
0 ( ) v
Lo
( ) t
t
2
2t
long
--------------
. ,

| `
v
Ld
v
Lo
( )
t
2
2t
long
-------------- +
. ,

| `
0 ,

' '

=
Page 24
We performed simulations in order to demonstrate our theoretical results. In the simulations, we
set T=50 sec, t
adj
=0 sec, t
lat
=5 sec, H=12 feet, and t
long
=10 sec. The relative speed between the
vehicles L
d
and L
o
was set equal to 20 and -20 miles/hour. The safe and unsafe regions are
shown in figure 18. It should be noted that the less the relative velocity between the vehicles L
d
and L
o
is, the larger the safe region becomes. Comparison between figures 8 and 18 indicates
that the slope of the safety margins remain almost the same, while there is a "horizontal shift" in
the safety margins.
Fig 18: The collision region between M and Ld; Case II: switching longitudinal acceleration.
II) t
adj
>0:
Similar to the previous cases we define the variables,
(57)
After obtaining the state space equations, the isoclines can be found as follows:
(58)
where constant c is a constant that depends on the initial values, x
1
(0) and x
2
(0). Figure 19 shows
the isoclines corresponding to various values of a
adj
. The initial point has been chosen to be in
the unsafe region. By applying a
adj
, it is possible to move into the safe region in order to start the
merging maneuver. It should be noted that only two values of a
adj
, i.e. -5, -7, are acceptable here.
The other values result in Sr(t)<0, for some t in the interval [0,t
adj
], which cause collision
between M and L
d
(the shaded area in figure 19 corresponds to the negative values of Sr(t) which
are infeasible). The minimum value of t
adj
for each a
adj
is chosen to be the time-instant the
x
1
x
Lo
x
M
l
L1
=
x
2
v
M
v
Lo
=
x
1
x
2
2
2a
adj
------------ c + =
Page 25
corresponding isocline curve intersects with the safety region.
Unfortunately in this case we can not easily apply the acceleration profile of figure 13. This is
largely due to the existence of the infeasible region, and the small area bounded between the
MSSs curve and the infeasible area. It should be noted that in the previous cases we use this area,
where the relative longitudinal velocity between the merging vehicle M and the vehicles in the
destination lane is negative, to move into before adjusting the longitudinal acceleration zero. In
the case of positive relative velocity between L
d
and L
o
, i.e. v
Ld
-v
Lo
>0, it is unrealistic to exploit
the modified switching acceleration profile, see figure 19.
Fig 19: Applying a
adj
to move from the unsafe region into the safe region.
Minimum Longitudinal Safety Spacing between M and F
o
:
I) t
adj
=0:
The velocity of all vehicles except M is assumed to be constant while the longitudinal velocity of
M becomes equal to the longitudinal velocity of the vehicles in the destination line at the time-
instant t
long
and remains constant thereafter. Therefore the value of a
M
is as same as that in
equation (53). According to equation (29) the condition for collision avoidance between M and
F
o
is:
(59)
It should be noted that the last term in equation (59) is dependent on the difference between the
velocity of originating and destination lane. In order to find the minimum initial longitudinal
safety spacing, MSS(M,F
0
), we define the following variables:
Sr t ( ) Sr 0 ( ) v
M
0 ( ) v
Fo
( ) t
t
2
2t
long
--------------
. ,

| `
v
Ld
v
Fo
( )
t
2
2t
long
-------------- + +
. ,

| `
0 > =
t 0 t
C
, [ ]
Page 26
(60)
It the relative velocity between two vehicles M and L
d
, v
M
(0)-v
Ld
, happens to be zero in equation
(60) we can apply the results of subsection 3.1. Let us define the variable S(t) as follows:
(61)
Using the above definitions we can analytically find MSS(M,F
0
) as shown in Table 2.
For simulation we set T=50 sec, t
adj
=0 sec, t
lat
=5 sec, H=12 feet, and t
long
=10 sec. The relative
velocity between the vehicles L
d
and F
o
was set equal to 20 and -20 miles/hour. The safe and
unsafe regions are shown in figure 20. Comparison between figures 9 and 20 indicates that the
tangent of the safety margins remain almost the same, while again we observe a "horizontal
shift" on the safety margins.
Table 2: Analytical values of MSS(M,F
o
)
IF & IF & IF THEN
<0 v
Ld
<v
M
(0), v
Fo
>v
M
(0) MSS(M,Fo)=S(t
C
)
" v
Ld
>v
M
(0), v
Fo
<v
M
(0) MSS(M,Fo)=S(0)
>0 t
max
>t
C
v
Ld
<v
M
(0), v
Fo
<v
M
(0) MSS(M,Fo)=S(0)
" " v
Ld
>v
M
(0), v
Fo
>v
M
(0) MSS(M,Fo)=S(t
C
)
>0 t
max
<t
C,
t
max
>t
C
/2 v
Ld
<v
M
(0), v
Fo
<v
M
(0) MSS(M,Fo)=S(0)
" " v
Ld
>v
M
(0), v
Fo
>v
M
(0) MSS(M,Fo)=S(t
max
)
>0 t
max
<t
C
/2 v
Ld
<v
M
(0), v
Fo
<v
M
(0) MSS(M,Fo)=S(t
C
)
" " v
Ld
>v
M
(0), v
Fo
>v
M
(0) MSS(M,Fo)=S(t
max
)

v
M
0 ( ) v
Fo

v
M
0 ( ) v
Ld

----------------------------- =
t
max
t
long
=
S t ( ) Max v
Fo
v
M
0 ( ) ( ) t
t
2
2t
long
--------------
. ,

| `
v
Fo
v
Ld
( )
t
2
2t
long
-------------- +
. ,

| `
0 ,

' '

=
Page 27
Fig 20: The collision region between Fo and M; Case II: switching acceleration.
II) t
adj
>0:
Here, the state space variable can be defined as follows:
(62)
Obtaining the state space equations, the isoclines corresponding to a
adj
can be found as follows:
(63)
Fig 21: Applying a
adj
to move from the unsafe region into the safe area.
where the constant c depends on the initial values, x
1
(0) and x
2
(0). Figure 21 shows the isoclines
x
1
x
M
x
Fo
l
M
=
x
2
v
Fo
v
M
=
x
1
x
2
2
2a
adj
------------ c + =
Page 28
corresponding to various values of a
adj
. The initial point has been chosen to be in the unsafe
region. By applying a
adj
, it is possible to move into the safe region in order to start merging
maneuver. It should be noted that only two values of a
adj
, i.e. 5 and 7, are acceptable here. The
other values result in Sr(t)<0 that is collision between M and F
o
. The minimum acceptable value
of t
adj
for each a
adj
is determined by the point of intersection between the corresponding isocline
curve and the safety margin.
4. Conclusions
In this paper, we analyzed the kinematics of the vehicles involved in a lane changing/merging
maneuver, and studied the conditions under which lane changing/merging crashes can be
avoided. That is, given a particular lane change/merge scenario, we calculated the minimum
longitudinal spacing which the vehicles involved should initially have so that no collision, of any
type, takes place during the maneuver.
Simulations of a number of examples were presented to demonstrate the results. We assumed in
our simulations that all vehicles, involved in the lane changing/merging maneuver, were initially
at steady state, i.e. their velocities were constant prior to the maneuver. Except for the merging
vehicle, all vehicles keep their steady state velocity during the merging maneuver. Three
different longitudinal acceleration scenarios - constant longitudinal velocity, switching
longitudinal acceleration, and modified switching longitudinal acceleration - were applied to the
merging vehicle in order to determine the safe and unsafe region as well as the MSS between the
merging vehicle and its surrounding vehicles. We observed that the switching scenario and the
modified switching scenario expanded the safe region for lane changing. Furthermore, by
considering the longitudinal adjustment acceleration for the merging vehicle, we studied the
possibility of moving from the unsafe region into the safe region. Our results together with
appropriate sensors and equipment on the board of vehicles could be used to assess the safety of
lane changing maneuvers and provide warnings or take evasive actions to avoid collision.
Reference
[1] J.-S. Wang, R. R. Knipling, "Lane Change/Merge: Problem Size Assessment and Statistical
Description," National Highway Trafc Safety Administration, 1993.
[2] J. D. Chovan, et al " Examination of Lane Change Crashes and Potential IVHS
Countermeasures," National Highway Traffic Safety Administration, 1994.
[3] D.N. Godbole, et. al. Design of Emergency for Automated Highway System: Obstacle
Avoidance Problem, IEEE Conference on Descision and Control, PP 4774-9, 1997.
[4] Zvi Shiller, and Satish Sunder, Emergency lane-change Maneuvers of Autonomous
Vehicles, ASME Journal of Dynamic Systems, Measurement and Control, Vol. 120, No.1,
Page 29
PP 37-44, March 1998.
[5] J.L. Bascunana, "Analysis of lane change collision avoidance," Systems and Issues in ITS,
(SP-1106), 1997.
[6] A. Kanaris, E.B. Kosmatopoulos, and P.A. Ioannou, "Strategies and spacing requirements for
lane changing and merging in automated highway systems," Technical Report, University of
Southern California, 1997.
[7] P.A. Cook, Nonlinear Dynamical Systems, Prentice-Hall,1986.
[8] S.E. Shladover, et. al, "Automatic vehicle control development in the PATH program," IEEE
Transactions on Vehicular Technology, Vol 40, PP. 114-130, 1991.
[9] J.A. Dickerson, et al., "Lateral and longitudinal control analysis", Technical Report,
University of Southern California, 1994.
[10] A. Kanaris, P. Ioannou, F. Ho, " Spacing and capacity evaluations for different AHS
concepts," Technical Report, University of Southern California, 1995.