Professional Documents
Culture Documents
CE 403
DR Akhilesh K Maurya
Microscopic Relation
The basic microscopic relation is between speed and distance headway. One of the major
discoveries of traffic science was the relation between microscopic speed-distance headway
behaviour and macroscopic speed-density (and hence speed-flow or flow density)
relationship. In this chapter we first describe various characteristics of the speed-distance
headway behaviour and some of the models developed.
Characteristics:
This behaviour essentially arises from the two needs
(i) The need to maintain a desired speed
(ii) The need to maintain a desired minimum distance
Whenever one can maintain ones desired speed the interaction between speed-distance
headway become masked. However, when one travels at a speed less than the desired speed
one generally would travel at the speed of the leading vehicle and at the minimum desirable
distance headway.
Owing to this second phenomenon, one see a complex behavioral interaction between the
actions of leading vehicle and those of the following vehicle. This behaviour is called the car-
following behaviour; different models of this behaviour gives rise different macroscopic u-k
relationships.
1
9/4/2018
Brief Discussion:
1. The response of any vehicle could be only in terms of acceleration/deceleration, &x&n (t )
The stimulus could be differences in speed, deviation of present distance headway from
that desired for the current speed, etc.
Of course there will be a time lag between the stimulus and its response.
2. This stimulus-response process is approximate in the sense that there neither exists a
deterministic mapping between the stimuli and response nor the stimuli precisely
perceived.
3. The process of response to stimulus need not be symmetric. That is, if s→ r then -s
need not to imply -r.
4. Since the process of car-following is an approximate process is difficult to believe that a
very precise distance headway/velocity can be maintained by the driver of following
vehicle over a period of time. In fact, there is always some “drift” one sees in this
distance/speed are one maintains. That is even when two vehicles are traveling at the
“same” speed and at a “constant” distance headway there exists some variation.
Distance headway
lim e(t ) = f (t )
t →∞
drift
where f(t) is an oscillating
function with more or less
constant magnitude.
Time
2
9/4/2018
Distance headway
perceptual threshold
+ ve 0 - ve Relative speed
3
9/4/2018
5. Even if following vehicle (FV) finds that it is traveling at same speed of leading vehicle
(LV) it may slow down or speed-up depending on whether the distance headway at that
point acceptable or not to the driver.
For Line 1: IDH= 180 ft, IS=50 ft/s, LV neither accelerates nor decelerates
For Line 2: IDH= 30 ft, IS=50 ft/s, LV neither accelerates nor decelerates
6. Car-following behaviour is generally stable over time and space. That is any perturbation
e(t) to the system of a pair of vehicles eventually dies down to f(t); and as these
perturbation are passed on from one system to the other they die. The first is called
local stability and the later is called asymptotic stability.
Local stability
lim e(t ) = f (t )
t →∞
Asymptotic stability
lim en (t ) = f (t )
n→ N
4
9/4/2018
• Pipes’ model I
• Pipes’ model II
• Forbes model
• GM models
• Fuzzy inference model
Pipes’ model I:
A good rule for following another vehicle at a safe distance is to allow yourself the length L
for every u mph of speed at which you are traveling.
xn (t ) = xn +1 (t ) + {b + Tx& n +1 (t )}+ Ln
L
where T=
u
xn (t )
xn+1 (t )
n+1 n
L
b+ × x& n +1 (t )
u
Ln
1
x&n +1 (t ) = {xn (t ) − xn +1 (t ) − b − Ln } L
T where T=
u
∴ &x&n +1 (t ) =
1
{x&n (t ) − x&n+1 (t )}
T
10
5
9/4/2018
Assumption: Drivers react to the rate of change of visual angle. Further the length of LV is
considered negligible compare to the distance between the LV and FV.
xn (t )
xn +1 (t )
n+1 θ n W
xn (t ) − xn+1 (t )
&x&n +1 (t ) = −Cθ&(t )
θ W /2
note tan =
2 {x n ) − xn +1 (t )}
(t
11
W
θ = 2 tan −1
2(
n x ( t ) − x n +1 (t ) )
CW
&x&n +1 (t ) = × ( x& (t ) − x&n +1 (t ) )
{xn (t ) − xn +1 (t )}2 n
Forbes model:
Assumption: Drivers choose to keep a minimum time gap from the rear of the LV, and that this
is equal to the reaction time, Δt.
Therefore
Ln
hn +1 (t ) = ∆t +
x&n (t )
12
6
9/4/2018
Assuming that all vehicles are of the same length one may define time headway by using the
rear end of the vehicle as the reference point
Ln
hn +1 (t ) = ∆t +
x&n (t )
d n +1 (t )
now hn +1 (t ) =
x&n +1 (t )
d n+1 (t ) L
= ∆t + n+1
x&n +1 (t ) x&n+1 (t )
∴ d&n +1 (t ) = ∆t × &x&n +1 (t )
1
∴ &x&n +1 (t ) = {x&n (t ) − x&n+1 (t )}
∆t
13
GM model:
Assumptions:
• The GM model is a stimulus-response model of car-following.
• Relative speed (with LV) is the only stimulus.
• The degree to which a given stimulus affects the response is a function of the distance
headway (from LV) and its speed.
xn (t )
xn+1 (t )
n+1 n
7
9/4/2018
In summary, the GM model does achieve local and asymptotic stability. However;
1. SDH obtained using the GM model is sensitive to initial distance headway, initial speed
and the perturbation pattern of the LV.
2. The GM models implicitly assume that the initial condition is a stable condition.
3. The GM model has no limit on the acceleration and deceleration rates.
16
8
9/4/2018
Assumptions:
• The stimulus is perceived only linguistically by the driver who then utilizes a set of
approximate driving rules (which is the outcome of a person’s driving experience and
attitude) to infer an approximate response.
• Driver’s response depends on three stimuli – relative speed, distance headway and
acceleration/deceleration of the LV.
17
9
9/4/2018
Rule formation:
Rule: Premise
Premise variables of a rule are the distance headway between the LV and FV (DS), relative
speed of the vehicles (RS), and the acceleration (or deceleration) rate of the LV (ALV). Each
of these premise variables are grouped into many natural language – based categories (see
the following table). Each of these categories is fuzzy set.
ALV
DS RS
acceleration deceleration
1 very small FV slower strong strong
2 small FV slightly slower somewhat strong somewhat strong
3 adequate Near zero normal normal
Categor-
ies 4 more than adequate FV slightly faster mild mild
5 large FV quite faster very mild very mild
6 very large FV faster none none
19
Rule: Consequence
Consequence of a rule is the FV’s reaction in terms of acceleration or deceleration rate
expressed in fuzzy quantity (AFV) which can be further represented by a natural language
term such as VERY STRONG DECELERATION.
If a category of DS in a rule is ADEQUATE, the AFV is computed as follows:
If the category of DS in Rule i is different from ADEQUATE, the value of AFVi is modified.
The modification is done by sliding the membership function of AFVi to the right or to the
left (making it larger or smaller) according to DSi’s deviation from the category ADEQUATE.
That is, AFVi is determined by
where βDSi is the number of categories for which DSi deviates from ADEQUATE (it can be
positive or negative number depending on whether the deviation is to a longer distance or
a shorter distance, respectively), and φ in this case is 1 ft/s2.
20
10
9/4/2018
Rule: Structure
Each rule is a conditional statement in the sense that, given a set of conditions represented
by the promise variable, the consequence is predicted.
21
Following figure shows the process of deriving the conclusion based on the compatibility
between input (DS, RS and acceleration or deceleration rate of the LV) and the premise of a
rule i. In this particular example two rules are applied to the same input.
22
11
9/4/2018
Properties:
• This model represents an approximate behaviour.
• This model incorporates asymmetry in response. For example, a rule which applies for
large positive relative speed may suggest a response of medium acceleration; whereas,
a rule which applies for similar conditions of headway and actions of the LV but for large
negative relative speed may suggest large deceleration.
• This model considered more than one stimulus (which affects the actions of FV) which
allows it to explain the closing-in and shying-away behaviour (see the figure below).
• This model also shows the drift behaviour (i.e. at the stable condition the distance
headway between the LV-FV pair does not remain constant but it oscillates around the
stable distance headway).
For Line 1: IDH= 180 ft, IS=50 ft/s, LV neither accelerates nor decelerates
For Line 2: IDH= 30 ft, IS=50 ft/s, LV neither accelerates nor decelerates 23
• To study the asymptotic stability a five-car platoon is simulated using this model.
Following figure shows distance headway variations with time for all pairs of two
consecutive cars in this platoon. It can be seen that the perturbation (variation in
distance headway) reduces as one proceeds down the platoon. This shows that the
simulated stream is asymptotic stable.
24
12
9/4/2018
Comprehensive Models
These models consider both the longitudinal control and lateral control behaviour of
drivers. The longitudinal control is achieved by controlling the vehicle’s speed (i.e., through
acceleration/deceleration as in previous models) while the lateral control is achieved
through proper choice of steering angles. In reality both these activities (longitudinal and
lateral control) are inter-dependent and goes on concurrently.
The only comprehensive models are those developed at IIT Kanpur (for e.g., see Gupta et al.
[1998] and Chakroborty et al. [2004]). These are force field (or potential field)-based
models.
• All roadway and traffic features are viewed as obstacles, either dynamic or static. Each
obstacle poses a threat to the safety of the driver. Hence, each obstacle emanate a
positive potential (repulsive force field) around it which repels the driver. However,
every goal emanates a negative (or attractive force field) potentials which attract the
driver;
• Shape and strength of the potential field emanated by an obstacle depend on the
properties of the obstacle. For example, the potential due to a parked vehicle is less
pronounced than the potential emanated by a truck coming in the opposing direction;
• The potential at a point on the road is assumed to be the algebraic sum of all the
potentials from the various obstacles and goals;
• The potential at a point is perceived as a threat to a driver’s safety; the threat increases
with speed; hence, it is assumed that the sustainable speed (a speed at which a driver
feels comfortable) at a point is inversely related to the potential at that point;
• Road is divided in many transverse cross-section and vehicle is assumed to move from
one cross section to next;
• Given the lateral position of the vehicle, its orientation, and the maximum steering
angle of a vehicle, only certain points on the next cross-section are accessible by that
vehicle. These points of the next cross-section are referred to as the accessible points
(see the figure in next slide).
• Each driver wishes to reach his/her destination quickly, he/she chooses the path which
minimizes the potential (and hence maximizes the speed).
The two submodels (steering response model and acceleration response model) are
presented to predict the response of the drivers through (1) steering control and (2) choice
of acceleration or deceleration rate is given driving situation.
26
13
9/4/2018
U ( x* , y * ) = min U ( xn , yn )
n n ∀ ( xn , yn )∈ An
27
V (t ) − Va (t ) V (t ) − Va (t )
&x&(t + ∆t ) = α β (t )k1 s − k 2U& (t ) + (1 − α ) s
∆t ∆t
where α is a parameter which indicates whether or not the vehicle is driving in a condition
constrained by other dynamic obstacles (which, it is felt, drivers view differently than static
obstacles). The value of α varies from 0 to 1; α=0 if the driver is in a ‘‘free flow’’ like
situation, α=1 if the driver is in a complete ‘‘constrained flow’’ situation.
β(t) is a sensitivity parameter for the first term in brackets. It is assumed that the driver is
more sensitive in situations where the driver feels unsafe, i.e., β(t) is higher if the driver
faces a potential (at time t) which is higher than the potential for which the actual speed (at
time t) is the sustainable speed. Keeping this in mind the following expression is proposed
for the determination of β(t):
U (t ) − f −1[Va (t )]
β (t ) = 1 +
d
The function f(.) is a function relating potential at a point to the sustainable speed at that
point and d is a constant to ensure that β(t) does not become negative in practical
situations. 28
14
9/4/2018
In a case, if the only obstacle is the road edge (i.e. with very little or no traffic, no lane
marking etc) then the minimum potential point (x*), for any y, is given by
b2 ln(a1b1 / a2b2 ) for keep left policy
x* = w+
b1 + b2 b1 + b2 b1>b2 and a2>a1
The above equation illustrates that U * ( y ) is a constant with respect to y if the width of the
road w is constant.
29
PPF emanates by static obstacle should look like a symmetric hill (with the obstacle in
middle) with higher potential near the obstacle and lower potential away from it (as shown
in figure).
30
15
9/4/2018
31
Modeling
approaches
16
9/4/2018
Spe ed
Flo w
uf q max
u0
q max k0 kj
Flow Density
The quantity qmax is the capacity of the facility for which the above figure is drown.
Speed
However, more recent understanding of
the relation acknowledge that it is very
difficult to represent the congested
regime through a single relationship;
there exists a jump around the qmax and
that slopes on the u-q curve in the free
regime is quite flat.
Flow 33
Ideal Capacity: Maximum number of passenger-cars that can expected to cross a point or line
on an ideal road in a unit interval of time. Ideal road section is one which has ample width (at
least 3.5 m wide lanes), wide paved shoulders (al least 1.8 mm wide) and zero gradient.
34
17
9/4/2018
Capacity Analysis
Speed
Speed
u0
However, more recent understanding of the relation acknowledge that it is very difficult to
represent the congested regime through a single relationship; there exists a jump around the
qmax and that slopes on the u-q curve in the free regime is quite flat.
There exists a speed, u0, where flow is maximum or reaches its full capacity.
35
Capacity Analysis
36
18
9/4/2018
HCM method
New HCM method suggests capacity values
based on the free-flow speed in actual
driving conditions. Free-flow
speed
values
BFFS is set at 70 mph for urban facilities and 75 mph for rural facilities
37
HCM method
Step 3: Determine the reduction in free-speed due
to lane width, shoulder width, no of lanes, no.
of interchanges. Free-flow
speed
values
FFS = BFFS - fLW-fLC- fN - fID
u
Where:
Step 4: Sum up all the reduction and subtract the sum from the ideal/base free-flow speed.
Step 5: Based on free-flow, determine the capacity from above figure in pcphpl.
Effect of vehicle mix and driver population is incorporated in conversion of the existing
traffic volume from vehicles/hr to an equivalent no. of passenger cars/hr.
38
19
9/4/2018
Speed-Flow Curves for Basic Freeway Sections, from Highway Capacity Manual, Dec 2000,
Exhibit 23-3, pg. 23-5.) 39
Speed-Flow Curves for Multilane Highway, from Highway Capacity Manual, Dec 2000,
Exhibit 21-3, pg. 21-4.)
20
9/4/2018
HCM method
• Adjustment Factor for Lane Width (fLW)
HCM method
42
21
9/4/2018
Level-
Level-of-
of-service, LOS
Level-of-service (LOS) offered by a particular express-section at a given time is dependent on:
(i) Demand at that time
(ii) Capacity of the road.
To convert the existing or forecast demand volumes, q, to an equivalent flow rate f (in pcphpl) under
ideal conditions:
q
f = k=k1 k=k2 k=k3
PHF × N × f hv × f p k=k4
k=k5
u3
u2
u1
A B
C D
E
u
F
q, (in pcphpl)
This result is used to enter either the standard speed-flow curves of freeways or multilane
highways. Using the appropriate free-flow speed, the curves may be entered on the x-axis with
the demand flow rate, v,, to determine the level of service and the expected average speed 43
Level-of-service, LOS
44
22
9/4/2018
Level-of-service, LOS
http://en.wikipedia.org/wiki/Intersection_%28road%29 http://www.sharoma.com/roundabout.htm
23
9/4/2018
Traffic stream
Space-mean speed (u)
Flow (q)
Density (k)
q = uk
Think units: [veh/hr] = [mi/hr][veh/mi]
24
9/4/2018
k crit = k j 2
q max
Flow (q)
kj
uf
uf
Speed (u)
Density (k)
25
9/4/2018
Flow (q)
Density (k)
q x
k t
26
9/4/2018
q x
k t
SHOCKWAVES
27
9/4/2018
28
9/4/2018
29
9/4/2018
x
IV
III
II
30
9/4/2018
x
IV
III
II
x
IV
III
II
I
t
31
9/4/2018
x
IV
III
II
I
t
I II III IV
uf = 60 mi/hr x
IV
kj = 240 veh/mi III
II
qmax = ?
I
t
32
9/4/2018
uf = 60 mi/hr x
IV
kj = 240 veh/mi III
II
qmax = 3600 veh/hr
I
I t
u1 = 55 mi/hr
k1 = ?
q1 = ?
uf = 60 mi/hr x IV
kj = 240 veh/mi III
II
qmax = 3600 veh/hr
I
t
I II
u1 = 55 mi/hr u2 = 0 mi/hr
k1 = 20 veh/mi k2 = ?
q1 = 1100 veh/hr q2 = ?
33
9/4/2018
uf = 60 mi/hr x IV
kj = 240 veh/mi III
II
qmax = 3600 veh/hr
I
t
I II III
u1 = 55 mi/hr u2 = 0 mi/hr u3 = ?
k1 = 20 veh/mi k2 = 240 veh/mi k3 = ?
q1 = 1100 veh/hr q3 = 3600
q2 = 0 veh/hr
uf = 60 mi/hr IV
kj = 240 veh/mi u4 = undefined
k4 = 0 veh/mi
qmax = 3600 veh/hr
q4 = 0 veh/hr
I II III
u1 = 55 mi/hr u2 = 0 mi/hr u3 = 30
k1 = 20 veh/mi k2 = 240 veh/mi k3 = 120
q1 = 1100 veh/hr q2 = 0 veh/hr q3 = 3600
34
9/4/2018
I II
u1 = 55 mi/hr u2 = 0 mi/hr
k1 = 20 veh/mi k2 = 240 veh/mi
q1 = 1100 veh/hr q2 = 0 veh/hr
I II III IV
I II
u1 = 55 mi/hr u2 = 0 mi/hr
k1 = 20 veh/mi k2 = 240 veh/mi
q1 = 1100 veh/hr q2 = 0 veh/hr
I II III IV
35
9/4/2018
I II
u1 = 55 mi/hr u2 = 0 mi/hr
k1 = 20 veh/mi k2 = 240 veh/mi
q1 = 1100 veh/hr q2 = 0 veh/hr
I II III IV
A B
uA kA qA uB kB qB
uAB
A B
36
9/4/2018
uA kA qA uB kB qB
uAB
A B
Vehicle speed from the left
→ →
q AB = u kA
AB relative to shockwave
= (u A − u AB )k A
uA kA qA uB kB qB
uAB
A B
Vehicle speed from the right
→ →
q AB = u kA
AB relative to shockwave
= (u B − u AB )k B
37
9/4/2018
A uAB B
→ →
q AB = q BA
(u A − u AB )k A = (u B − u AB )k B
u A k A − u AB k A = u B k B − u AB k B
u AB (k A − k B ) = u A k A − u B k B
u AB (k A − k B ) = q A − qB
q A − qB
u AB =
k A − kB
A uAB B
q A − qB
u AB =
k A − kB
38
9/4/2018
u1 = 55 mi/hr
k1 = 20 veh/mi I
1100 − 0 q1 = 1100 veh/hr
u I , II =
20 − 240
u2 = 0 mi/hr
= −5 mph II k2 = 240 veh/mi
q2 = 0 veh/hr
I II III IV
u2 = 0 mi/hr
k2 = 240 veh/mi II
0 − 3600
u II , III = q2 = 0 veh/hr
240 − 120
u3 = 30 mi/hr
= −30 mph III k3 = 120 veh/mi
q3 = 3600 veh/hr
I II III IV
39
9/4/2018
u3 = 30 mi/hr
k3 = 120 veh/mi III
0 − 3600
u III , IV =
120 − 0 q3 = 3600 veh/hr
I II III IV
u1 = 55 mi/hr
I k1 = 20 veh/mi x IV
q1 = 1100 veh/hr III
II
u3 = 30 mi/hr
I
III k3 = 120 veh/mi t
q3 = 3600 veh/hr
I III
40
9/4/2018
IV II
Density (k)
III
Flow (q)
IV II
Density (k)
41
9/4/2018
IV II
Density (k)
x
IV
III
II
I
t
42
9/4/2018
If q1 and q2 , and k2 and k1 are nearly equal then in the limit we can write:
dq
uw =
dk
Note that the speed of the is basically the slope of the line joining the two flow and density
conditions on a q-k plot.
slope of this line
q
(1) is u w
q1
q2 (2)
• Forward moving shock wave
• Stationary shock wave
• Backward moving shock wave
k1 k2 k
85
q2 − q1 16 × 75 − 1000
usw1 = = = 3.39 km/hr
Distance
k2 − k1 75 − 16
q − q 1400 − 1200
qmax=1400 usw2 = m 2= = −6.45 km/hr
shockwave 4 k0=44
u0=1400/44
k0 − k 2 44 − 75
qm − q1 1400 − 1000
q2=16x75
k2=75
usw3 = = = 14.29 km/hr
u2=16 k0 − k1 44 − 16
0 − q2 0 − 1200
usw4 = = = 16 km/hr
0 − k2 0 − 75
q1=1000
k1=16
u1=62.5 0 − qm 0 − 1400
usw5 = = = 31.82 km/hr
0 − k0 0 − 44
Time
86
43
9/4/2018
To find the maximum length if the one must realize that the platoon grows till the time
Shockwave 2 develops.
The rate of growth of the platoon is the | relative speed | between shockwave 1 and
shockwave 4.
The platoon grows at the rate of (16-3.39) kmph or 12.61 kmph.
The platoon grows unabated for 10 minutes (the duration for which the truck is there).
The time it takes the platoon to dissipate = 2.1/9.84 = 0.213 h = 12.8 minutes
87
Problem 2:
For the problem 1 plot the location of the front of the platoon and rear of the platoon
versus time. Choose appropriate reference frame.
Choosing distance = 0 at the point at which the truck enters and time = 0 as the time at
which the truck enter the enters.
Distance (km)
2.67
-6.45
16 60
60
1.29
3.39
0.565 60
0
0 10.0 22.8 Time (min)
88
44
9/4/2018
Problem 3:
For the problem 1 plot the length of the platoon versus time.
platoon (km)
Length of
2.1
-9.84
12.61
60
60
0
0 10.0 22.8 Time (min)
89
Intersection:
An Interruption is a location where three or more roads carrying traffic in different
directions cross. The space which is common to all these roads is referred to as the
intersection.
•Signalized intersection
•Unsignalized intersection
Type of Signals
90
45
9/4/2018
Flow Characteristics:
Interruption to traffic flow at a signalize intersection is orderly and deterministic. In this
interruption pattern following process become important for analysis:
• Arrival process
• Departure process
• Delay and queue analysis
Arrival process:
Arrival process at intersection could be of three kind:
(i) Random arrivals
(ii) Grouped arrivals
(iii) Mixed arrivals
91
Random arrival:
Such arrival pattern is seen on isolated intersections (i.e. no intersection is present in the vicinity, 3-4 km).
In these case inter-arrival times (time headways) are often distributed more or less according to negative
exponential distribution
( λ t ) k e − λt
P( N t = k ) =
k!
P ( H 1 ≤ h ≤ H 2 ) = e − λH 1 − e λH 2
Grouped arrival:
Such arrival pattern is seen at intersections which are located close to (say < 2 km) another upstream
intersection.
Arrival process seems to be uniform and vehicles can be assumed to arrive at reasonably constant headways.
Vehicle released from upstream intersection reached in platoon.
Mixed arrival:
Such arrival pattern is seen at intersections which are located at intermediate distances (say from 2-4 km)
another upstream intersection.
It is not purely random not purely grouped. Because of large distance many of released vehicles may disperse
from platoon vehicles and come independently.
92
46
9/4/2018
Departure process:
If the headway were measured as vehicle entered the intersection an interesting pattern is
observed. The first headway would be defined as the time between the initiation of the
green signal and the first vehicle’s front bumper crossing the entry line. The second
headway would be defined as the time between the first vehicle front bumper crossing the
line and the next vehicle front bumper crossing the same line and so on. The pattern
typically observed is:
headway (between the
nth and (n-1)th vehicle)
t1
t2
t3 t4
h
1 2 3 4 5 6 7 8 9
n, the position in the queue
From the above figure two features emerge: (i) the headway stabilizes to a value h referred
to as the saturation headway; (ii) the initial headway are larger than h.
93
Saturation headway represents the maximum number of vehicle that can crass the intersection
during the green time.
Initial headways are larger than the hs because of perception reaction time and the extra time
taken to accelerate to a reasonable speed (note that later vehicles more or less achieve this
speed when they cross the specified point as they start from a distance further upstream
from the specified point). In the sense some time is lost due to the fact that initial vehicle
takes longer time than hs . Sum of these excess times is referred as start-up lost time, ls
ls = ∑ ( hi − hs )
∀i
The quantity is of the order of 2s. A typical data set from studies is shown in the following:
headwa
Greenshields (1942)
4
y
1 2 3 4 5 6 7 8 9
n, the position in the
queue 94
47
9/4/2018
Nearly the end of the departure process some time is also lost. This happens because invariably some part
of the amber time remains unutilized because vehicles come to a rest when some part of the amber time is
still remaining. This loss of time is referred as movement loss time ( or clearance loss time), lm.
Clearance lost time (or movement lost time) is the time between the last vehicle from one
approach entering the intersection and the initiation of the green signal for conflicting
movements, assuming condition in which demand is present to utilize the non-red times in
both approaches.
This quantity is of the order of 1.2 to 2.8 s. Longer yellow + all red times led to longer
clearance lost times.
Saturation flow rate is defined as the flow rate from a lane in the intersection assuming that
each headway is equal to the saturation flow rate headway (hs) and green exists for the entire
time period.
Hence if “hs” is the saturation flow rate headway (or saturation headway) at an intersection in
seconds then s,the saturation flow rate is:
3600 s is in vphgpl
s=
hs
However, “s” is just a theoretical quantity and what is the of prime importance is the
capacity. 95
In order to understand how the capacity is related to s consider the following example:
Example: A given lane at a traffic signal has been observed to have the following
parameter: hs=2.0 sec./veh, start-up loss time =1.5 sec., and clearance loss time=1.5 sec.
The signal provides the lane with 27sec. of green, 3 sec. of yellow and 30 sec. of red during
each 60 sec. cycle of the signal. Determine the capacity for that approach?
3600
• In one hour there are = 60 cycles.
60
• Hence total lost time in an hour = 60 (1.5+1.5) = 180 sec.
• Time available for the movement per 60 sec. is 30 sec. Hence, in 3600 sec only 1800
sec. movement is allowed.
• Out of 1800 sec., 180 sec. is lost. Hence, time in which movement take place at 2.0
sec./veh is 1800-180=1620 sec.
• Hence, capacity of the lane is
1620
= 810vph.
2.0
96
48
9/4/2018
Vehicles Cumulative
A(t)
arrivals, A(t)
R G R G R G R
97
98
49
9/4/2018
Now if we assume that A(t) has a constant slope of “v”; i.e. vehicles arrive at a constant rate
of v vehicles/unit time and the maximum rate at which they are discharged is “s” vehicles
per unit time and if the number of vehicles that arrive during a cycle are cleared during the
green period then one can easily determine the average delay an individual vehicle faces
thus:
The modified diagram of A(t) and D(t) in this case is:
Vehicles
A(t)
D(t)
R G R G R G
Time
C C C
99
∫ Wi di
s
Total delay during cycle = Wi
0 a b t Time
Or, total delay = Area of Δ abd (1) g
C
Now Wi= - (arrival time of ith vehicle – departure time of ith vehicle)
i
Arrival time of ith vehicle = [Note slope of A(t) is v]
v
i
Departure time of ith vehicle = (C − g ) + [note slope of D(t) line in the relevant portion is s]
s
∴ v−s
Wi = (C − g ) + i (2)
sv
Now t is the time where A(t) and D(t) line meet; hence
vt = s{t − (C − g )} or s(C − g )
t= (3)
(s − v)
100
50
9/4/2018
Hence,
vt i (v − s )
Total delay = ∫ (C − g ) + sv di
0
[Note i* = vt ]
(C − g ) v − s 1 2 s 2 (C − g ) 2
= (C − g )vs + × ×v
s−v sv 2 (s − v) 2
vs 1 (C − g ) 2
= (C − g ) 2 − sv
s−v 2 s−v
(C − g ) 2
= {2sv − sv}
2( s − v)
vs (C − g ) 2
=
2( s − v )
Now, total number of vehicles that arrive in the cycle is “vC”.
sv(C − g ) 2 1
Average delay = Total delay/ Total number of vehicles = ×
2( s − v ) vC
2
Average delay = (C − g ) × s (4)
2( s − v ) C
101
The average delay in equation (4) is generally termed as “uniform delay” (UD) as it is based
on the assumption of uniform arrival.
Note that in equation (4) v is the volume and s is the saturation flow rate.
Note that the equation (4) can be easily obtained by using 0.5 x base x height to calculation
of area of the Δabd and then dividing it by “vC” – the total number of vehicle that arrive
during the cycle length.
d
vs (C − g ) (see previous figure)
Note i * = vt =
s −v
i*
1 vs(C − g ) vs (C − g ) 2
a b
Hence, area = (C − g ) =
2 s−v 2( s − v )
(C-g)
102
51
9/4/2018
vs(C − g ) 2 1 (C − g ) 2 s
∴ Average delay =
2( s − v )
× = ×
vC 2( s − v ) C
(6)
103
gs
Let us study how one could estimate delay if over saturation (i.e. v > ) exists for a finite
period of time T. It must be understood here that the over saturation C is not due to
stochastic disturbances but due to a hike in demand for a certain period of time. So the
following analysis is completely deterministic.
Consider the following case:
Arrival flow
v
v2
v1
0 T τ Time
104
52
9/4/2018
In this case arrival /departure diagram would look like the following:
vehicles
A(t)
D(t)
Z
v2
y
v
v1
g T τ Time
0
C
During 0 to τ the signal is over saturated. Let us estimate the average delay during this time.
The total delay during this time will be the area shaded with dots + the area shaded with
lines.
105
Now, total delay due to area shaded with lines cab easily obtained by assuming the dashed
line as an arrival pattern and using the uniform delay equations.
The slope of the dashed line can be obtained by looking at any one of the small triangles.
Say if slope is σ, then Cσ = sg
g
or σ= s
C
Then by substituting “σ” in place of “v” in Equation 5 one would obtain on an average how
much time a vehicle has to wait due to the part of the figure shaded with lines.
C (1 − g / C )
2
=
gs / C
2 1 −
s
C g 1
= 1 − = (C − g ) (7)
2 C 2
which is half of the red light period
To compare the average delay in the over saturation case, one has to add the average delay
due to the over saturation component to the average delay due to the uniform component.
106
53
9/4/2018
Now average delay due to over saturation component, ADosc can be obtained (note in this
discussion “wait” refers to the “waiting due to the over saturation”).
Consider the vehicle that arrives at T. This vehicle has to wait for “Z” units of time (see the
previous figure).
The vehicle that arrives at time = 0 has to wait 0 units of time.
The waiting time of vehicles between 0 and T the “wait” increases linearly (since “wait” is
the difference between two straight lines – the A(t) line and dashed line.
“wait”
Z
time
2
T
Now look at the vehicles which arrive between T and τ.
The wait time decreases from “Z” (for the vehicle that arrived at T) to 0 (for the vehicle
arrive at τ).
“wait”
Z
T and τ is 1 Z
time
2
T
107
1
Hence, one can say that the average waiting time for all vehicle arriving during 0 to τ is Z
2
Now y
= Slope of the dashed line, σ Note σ =
g
s
Z C
y
∴ Z=
gs
C
gs gs
But y = vT − T = Tv −
C C
gs
Tv −
C v gs
∴ Z=
gs
= T
− 1
cap where cap =
C
C
T v
∴ 1
ADosc = Z =
2
− 1
2 cap
(8)
C (1 − g / C ) T v
ADos = + − 1
2 2 cap
108
54
9/4/2018
In reality, however more often than not arrival is not deterministic, it is stochastic as
discussed earlier. One it is assumed that arrival is stochastic the previously given relation for
average delay cannot be used.
Under the following assumptions, the delay for such stochastic arrivals have been obtained
by “Webster”.
Assumptions:
(i) The number of arrivals in a given time interval has Poisson distribution and that the
distribution does not change with time.
(ii) The departure headways are uniform.
g
(iii) v < s ; i.e. the system is not saturated,
C
(iv) The system has been running long enough to have settled into a steady state.
Under there assumptions “Webster” developed a delay equation which is given as equation
(7). This is the best known delay equation.
g 2 v
2
C 1 −
1 5g
2+
C + cap − 0.65 cap v
3 C
The first term of “Webster” equation is derived Equation 4 and the second term can be
obtained analytically through steady state queuing analysis and is often referred “overflow
delay” or “random delay”. The third is a correction term obtained using simulation studies
and generally effect a 5 to 15% reduction in the estimates of d obtained by summing the
first two terms. Hence, as an approximation the third term is often omitted and the sum of
the first two term is multiplied by 0.90.
Also note that Webster equation for obtaining cycle length is based on optimizing d from
Equation 9.
There, however, exists a problem with the Webster and Webster – like models which
assumes steady state conditions. They invariably over estimated the delay when v is close to
gs / C . This over-estimation is due to the fact that it assumes steady state operation which
would imply that v is close to gs / C for sufficiently long period so that steady state is
reached. If, in reality, such a thing happens then the delay estimates from Webster like
equations would hold not be too bad. However, this high demand never exists for that long
period that steady state reached. Hence, the discrepancy between mathematically obtained
results and real world results.
110
55
9/4/2018
Example:
On an approach to a signalized intersection, the effective green time and the effective red
time are 30 s each. The arrival rate of vehicle on this approach is 360 vph between 0 -120 s,
1800 vph for 120 – 240 s, and 0 vph for 240 – 420 s. The saturation flow rate for this
approach is 1440 vphgpl. The approach under consideration has one lane. Assume that at
time = 0 s the light for the approach has just turn red.
Plot of arrival rate of vehicle versus time Plot of cumulative number of arrivals and departure
of vehicle versus time.
111
Q3. Determine the average delay to the vehicles arriving between 0 – 120 s, 120 – 240 s
and 0 – 240 s.
Between 120 – 240 s the intersection is operating under oversaturated conditions. The
arrival is deterministic and uniform. Average delay can be calculated using following
equation
(C − g ) + T v 60 − 30 120 1800
− 1 =
ADos = + − 1 = 105s
2 2 cap 2 2 720
56
9/4/2018
Average delay between 120 – 240 s can be also obtained from following figure:
Average delay= (Area of Triangle III + 5x Area of Triangle IV)/No. of arrivals from 120-240 s
0.5 ×180 × 60 + 5 × 0.5 ×12 × 30
= = 105s
60
Average delay to all vehicle between 0-240 s can be obtained dividing the total delay (faced
by all vehicle) by the number of vehicle.
n d + n2 d 2 12 ×10 + 60 ×105
Average delay = 1 1 = = 89.2 s
n1 + n2 12 + 60
113
Q4. Determine the delay to the fourth and the sixtieth vehicles that arrive at the
intersection.
The arrival rate of vehicle from 0-120 s is 360 vph or 0.1 vps. Assuming that fourth vehicle
arrives before the expiry of 120 s, the time of arrival of the fourth vehicle is 4/0.1 = 40 s.
Departure rate of vehicles is 1440/3600=0.4 vps. The time of departure of the fourth
vehicle, assuming that fourth vehicle gets discharged during first green, is 30+4/0.4=40 s.
Therefore the delay to fourth vehicle is = departure time – arrival time = 40 – 40 = 0 s
The same observation can be made from above figure. The delay to the sixtieth vehicle
can also be read from figure as 144 s.
As can be seen from figure, the maximum queue length is 36 vehicles. At time = 240 s, the
queue length first becomes equal to 36 vehicle.
114
57
9/4/2018
Q7. Determine the percentage of time for which there exists a queue on this approach.
As can be seen from figure, there is no queue from 40 – 60 s and from 100 – 120 s. For the
rest of the time, there is a queue at the intersection. Hence, the % of time for which there is
no queue is (40/420)100 = 9.52 %.
Hence, the % of time when there exist a queue is 100 – 9.52 = 90.48 %.
Q8. Determine the average queue length between 120 and 420 s.
Average queue length = (Area of Triangle III + 5xArea of Triangle IV)/(total time from 120–420 s)
115
∫ q(t )dt
t
Vtotal
Average delay =
Total number of arrivals
R G R G R G R
Cycle 1 Cycle 2 Cycle 3
m
0.9 × I × ∑ qi Time
i =1
Average delay =
Vtotal
116
58
9/4/2018
Capacity analysis:
ci capacity of lane i
Gi
ci = si × Gi Green time for lane i
C si Saturation flow on lane i
Saturation flow depends on (i) no. of lanes in the lane group and width of lanes or
alternatively the width of lane group, (ii) gradient of the lane, (iii) percentage of turning
traffic, (iv) vehicle mix, (v) number of parking manoeuvers, and (vu) number of bus
stoppings.
117
Level of service:
118
59
9/4/2018
Various warrant conditions are defined for signalization. Detailed standards exists. However,
in this class, we are only going to look at the warrant conditions without going into details.
If vehicular volumes are “high” for a “reasonable period” of the day is “most” of
the approaches then signalization is warranted.
If the volume on certain approaches one quite high and the pedestrian volumes
wanting to cross those approaches are also high then signalize.
119
Some times none of warrants may be satisfied fully, however, if two or more of
warrants 1, 2, and 3 are satisfied to a reasonable extent then a signalization may
be warranted.
Although there other conditions which can also justify the use of signals like:
•A minor intersection between two intersections
•Flow pattern on an intersection is highly peaked with high volume observed only for 4 to 5
hours of a day.
120
60
9/4/2018
Terminology:
Cycle: one complete sequence of signal indications
Cycle Length: total time for signal to complete one cycle
Phase: part of cycle allocated to any combination of traffic movements receiving the right of way
Interval: period of time during which all the signal indications remain constant.
Change interval: the “yellow” and/or “all-red” intervals which occurs at the end of a phase to provide
for clearance of the intersection before conflicting movements are released.
Green Time: time within a given phase during which the green indication is shown
Lost time: time during which the intersection is not effectively used
Effective green time: time during which a given phase is affectively available for stable moving
platoons of vehicles in the permitted movements. It is equal to the green time plus the change
interval minus the lost for designated phase.
121
Signal Phasing
Phasing is the sequence by which the various movements both vehicles and pedestrians are
being served at a signalized intersection. The objective of phasing is the minimization of the
potential hazards arising from the conflicts of vehicular and pedestrian movements, while
maintaining the efficiency of flow through the intersection.
Greater the number of phases, better separated are the conflicting flows.
However, increasing the number of phases hinders efficiency while improving safety.
Safety improves (with large number of phases) because conflicts are eliminated, however,
efficiency falls because delays increase due to:
(i) more lost times (in start-up and unused yellow times), and
(ii) minimum phase duration requirements.
122
61
9/4/2018
There exists no algorithm by which phasing may be selected. It is purely an art based on
certain guidelines.
A. Keep the phasing scheme as simple as possible (like start with simple two phase
system)
B. Increase the number of phases if pedestrian or turning volumes is high.
The following diagram illustrates three most basic phasing scheme: (i) Two phase
operations, (ii) Three phase operations, and (iii) Four phase operations.
Vehicular Pedestrian
Pedestrian traffic traffic
traffic
Vehicular
traffic
Phase A Phase B
123
Pedestria
n not Vehicular All red
allowed traffic
Sometimes the right-turning volume from one road may be large and this may require a
separate phase.
Vehicular
traffic
Pedestrian traffic
Phase A Phase B
Phase C
THREE PHASE OPERATION
124
62
9/4/2018
If turning volume in either direction is heavy then a four phase operation may be warranted.
In all the above phasing schemes, note that if a right-turn is protected then no pedestrian
movement is allowed during the protected phase.
The four phase scheme shown above or the second of the two three – phase scheme
shown here works best if a turning lane exists.
125
A cycle is a complete sequence of signal indications; cycle length is the duration in which the
whole set of phases at a signalized intersection takes place once.
The appropriate cycle length is generally obtained using Webster’s equation. This equation
yields results close to an optimal cycle length, however, we shall not go into the details of its
derivation. We shall simple state it here. The details will be stated while discussing the delay
equations at an intersection.
Least delay point
(optimum cycle
• Cycle length’s effect is not monotonous length)
Vol. on approach
• there exists a cycle length for which delay
Average delay per vehicle
V1 > V2 > V3
to vehicles is the least
• Sensitivity of cycle length near optimum V1
cycle length is very small
• On both side of optimum cycle length V2
sensitivity is different with respect to cycle V3
length
63
9/4/2018
1.5L + 5
C= p
1 − ∑ (V / s) icr
i =1
C: Optimal cycle length, in seconds
L: Lost time during a cycle. Sum of the start-up lost time and the clearance lost times.
p: total number of phases in the cycle
(V / s ) : critical flow ratio for phase i
i
cr
p
L = ∑(lsi + lmi + lri )
ls: startup time loss
i=1
lm: movement time loss or
clearance lost time
lr: all red time loss
127
1670
725 765 Phase A
335
250
Saturation flows:
Th = 1800 vphgpl Phase B
Th, LT = 1700 vphgpl
Th, RT = 1650 vphgpl
To obtain the critical movements in each phase one proceeds in the following manner:
335 250
Phase A: max , = max{0.20,0.15} = 0.20
1650 1700
Hence the Th, RT movement from west is critical.
128
64
9/4/2018
If for the above problem, lost time per phase is given as 4 s then one could determine
1.5(2 × 4 ) + 5
C= = 50 s
1 − (0.20 + 0.46)
A point worth mentioning here is that empirical research show that cycle lengths within a
± 30% from the “optimal” length estimated using Webster’s formula perform close to the
optimal.
129
Step 1: For each phase compute the yellow / amber time requirement using dilemma
zone calculations. For the same phase generally the same amber duration is
provided. However, for different phases different amber times can be given.
Step 2: The (cycle time - ∑(amber time + all red) ) is allocated as green in proportion to
the critical flow ratios in every phase.
Step 3: Check whether the allocated green times meet the requirement from the
pedestrian standpoints. If it is does not meet the requirement then increase the
cycle time in steps of 5 s till the requirements are met.
130
65
9/4/2018
Let us look at the previous example again. Assume lane width is 3.66 m. Also assume that 3 s
of amber time is provided per phase.
Increase C to 60 s
A 60 – (2x3) = 54 49x (0.20/0.66) ≈ 16 3 7+(4x3.66)/1.2 =19 s 16 s ok
*During Phase A pedestrians have to cross 4 lanes. Cycle Length = 60 s All red time = 0 s
Tp = 7+(4x3.66)/1.2 = 19 s
Green time Amber time
Phase A 16 3
During Phase B pedestrians have to cross 2 lanes.
Tp = 7+(2x3.66)/1.2 = 13 s Phase B 38 3
131
• Consider the scenario where a vehicle moving at speed limit decides to stop when the
light turns amber.
The distance required to come to stop, xs
v02
xs = v0δ b +
2d c
where, v0 is the speed limit (or design speed), δb in the reaction time for breaking, dc is
the comfortable deceleration rate.
Note that any vehicle whose distance from stop line is less than xs when the light turns
amber will not be able to stop.
• Consider the scenario where a vehicle moving at speed limit decides to cross the
intersection when the light turns amber.
The distance required to be crossed during an amber time of τ , is
132
66
9/4/2018
ac (τ − δ a ) 2
= v0δ a + v0 (τ − δ a ) +
2
ac (τ − δ a ) 2
or = v0τ +
2
Since, v0 is the speed limit it is assumed the ac , the comfortable acceleration rate,
should not be operative as no body will accelerate. Hence, distance traveled is only v0τ
∴ v0τ ≥ xg + W + L
or xg ≤ v0τ − W − L
Any vehicle farther than xg will not be able to cross the intersection during amber duration.
133
xg xg
Can not go Can not go
Can go or can stop Can not go and can
(Option Zone) not stop (Dilemma
Zone)
(b)
(a)
Now clearly in situation (b) where xg < xs should be avoided. Note xg is the only term which
is function of τ.
Hence, at least xg = xs ; let this happen when τ = τ min
In this case, v02
v0τ min − W − L = v0δ b +
2d c
v0 W + L
τ min = δ b + +
2d c v0 134
67