Capacity at Unsignalized Two-Stage Priority Intersections

Published Transportation Research Record 1555

TRB, National Research Board, Washington, D.C. 1996

by Werner Brilon, Ning Wu, Kerstin Lemke
Abstract
This paper deals with the capacity of minor traffic movements across major divided four-lane
roadways (also other roads with two separate carriageways) at unsignalized intersections.
The center of the intersection, corresponding to the width of the median, often provides
spaces in which the drivers who crossed the first half of the major road may stop before
proceeding across the second major traffic stream.

This situation which is common at

multilane major streets is called two-stage priority. Here the capacity for minor through
traffic is larger than at intersections without such a central storage space. The additional
capacity being provided by these wider intersections can not be evaluated by conventional
capacity calculation models.
This paper presents an analytical theory for the estimation under two-stage priority
conditions.

It is based on a former approach by Harders.

However, a set of major

improvements were necessary to match the results with realistic conditions. In addition to
analytical theory, simulations have been performed which enable an analysis under more
realistic conditions. As a result a set of equations is presented which compute the capacity for
a minor through traffic movement in the two-stage priority situation. These equations are
completed by two sets of graphs which enable an easy application of the theory in practice.
keywords:
capacity, unsignalized intersection, two-stage priority, median central reserve.
Authors address:
Prof. Dr.Werner Brilon

Ruhr-University

Werner Brilon, Ning Wu, Kerstin Lemke

Dr. Ning Wu
Dipl.-Ing. Kerstin Lemke

page 1
D - 44 780 Bochum
Germany

Werner Brilon, Ning Wu, Kerstin Lemke

page 2

Capacity at Unsignalized Two-Stage Priority

Intersections
by Werner Brilon, Ning Wu, Kerstin Lemke

1.

INTRODUCTION

At many unsignalized intersections there is a space in the center of the major street available
where several minor street vehicles can be stored between the traffic flows of the two
directions of the major street, especially in the case of multilane major traffic. This storage
space within the intersection enables the minor street driver to pass each of the major streams
at a time. This behaviour can contribute to an increased capacity.

Part II
q5

Output line

STOP

k spaces for pass. cars

Input line

q1
q2
STOP

Part I

q8

Figure 1: Minor street through traffic (movement 8) crossing the major street in
2 phases. The theory discussed here is also available if the major street
provides more or less than 2 lanes per direction.

Werner Brilon, Ning Wu, Kerstin Lemke

page 3

Therefore, a model is needed which can describe this behaviour and it's implication on the
intersection capacity. A model of this type has been developed by Harders (1968). His
concept has been used here as a basis and it is described in the following derivations.
However, some major amplifications as well as a correction and an adjustment to reality have
been made to achieve better correspondance to realistic conditions.
For our derivations we look at an intersection consisting of two parts according to Fig. 1.
Between the partial intersections I and II there is a storage space for k vehicles. This area
has to be passed by the left turner from the major street (movement 1) and the minor through
traffic (movement 8). Also the minor left turner (movement 7) has to pass through this area.
We will see that movement 7 can be treated like movement 8. Therefore, for our derivations
we concentrate on the minor through traffic (movement 8) crossing both parts of the major
street. The enumeration of movements has been chosen in accordance with chapter 10 of the
HCM (1994). We assume that the usual rules for unsignalized intersections from the highway
code are applied by drivers at the intersections. Thus movements 2 and 5 (major through
traffic) have priority over each other movement. Movement 1 vehicles have to obey the
priority of movement 5 whereas movement 8 has to give the right of way to each of the
movements shown in Fig. 1 . In our derivations movement 5 stands for all major traffic
streams at part II of the intersection. These, depending on the layout of the intersection,
could include through traffic (movement 5), left turners (movement 4) and right turners
(movement 6).

Werner Brilon, Ning Wu, Kerstin Lemke

2.

page 4

ANALYTICAL MODEL FOR THE DETERMINATION OF THE CAPACITY

To determine the capacity of the whole intersection we assume a constant queue on the minor
approach (movement 8) to part I.
Let wi be the probability for a queue of i vehicles queueing in the storage space within the
central reserve. Then the probabilities wi for all of the possible queue lengths i must sum
up to 1 with 0 i k, i.e.:
k

=1

(1)

i =0

where k is the number of spaces in the storage space within the central reserve
Now we consider the central area of the intersection as a closed storage system, which is
limited by the input line and output line (cf. Fig 1). The capacity properties of the storage
system are restricted due to the aspects of maximum input and maximum output. We now
have to distinguish between different states of the system:
1. State 1 :

We first consider part I of the intersection which decides on the input to the storage area.
Under state 1 we consider situations during which the number i of vehicles in the storage
area is less than the maximum possible queue length k , i.e. i < k . During this state a minor
street vehicle from movement 8 can enter the storage space if the major streams (volume q1
and q2 ) provide sufficient gaps. In this case the capacity of part I (possible input from
movement 8) characterizes the capacity, i.e.:
c1 = c(q1 + q2)

(2)

Werner Brilon, Ning Wu, Kerstin Lemke

page 5

where
c(q1 + q2)

capacity of part I in case of no obstruction by the subsequent part II ,

which is the capacity of an isolated unsignalized cross intersection for
through minor traffic with major traffic volume q1 + q2 .

The probability for this state 1 is p1 = 1 - wk . Thus, the contribution of state 1 to the
capacity of part I for movement 8 is
cI,1 = (1 - wk) . c(q1 + q2)

(3)

Of course, during state 1 also vehicles from movement 1 can enter the storage space.
2. State 2 :

For this state we assume that the storage area is occupied; i.e. k vehicles are queueing in the
storage space. In this case normally no minor vehicle from movement 8 or vehicles from
movement 1 can get into the storage area. If, however, a sufficient gap for the passage of one
minor street vehicle can be accommodated at both parts (I and II) of the intersection
simultaneously then also a vehicle can get into the storage area. The capacity for q8 (possible
input from movement 8) during this stage is
c2 = c(q1 + q2 + q5)

(4)

where
c(q1 + q2 + q5) = capacity of an isolated cross intersection for through traffic with major
traffic volume q1 + q2 + q5 :
Thus, the contribution of state 2 to the capacity of part I is
cI,2 = wk . c(q1 + q2 + q5)
where

wk = probability that k vehicles are in the storage space

(5)

Werner Brilon, Ning Wu, Kerstin Lemke

page 6

State 1 and state 2 exclude each other. The capacity of part I is the total maximum input to
the storage area. Here the volume q1 of movement 1 in addition to the partial capacities
mentioned above has to be included. Therefore, the total maximum input to the storage area
is
Input = cI,1 + cI,2 +(1-wk) . q1
= (1-wk) . [c(q1 + q2) + q1]+ wk . c(q1 + q2 + q5)

(6)

3. State 3 :

We now consider the output of the storage area. Here we concentrate on part II of the
intersection. For i > 0 each possibility for a departure from the storage area provided by
the major stream of volume q5 can be utilized. The capacity (maximum output of the storage
area) of part II in this case is
c3 = c(q5)
where c(q5) =

(7)

capacity of part II in case of no obstruction by the upstream part I

which is the capacity of an isolated unsignalized cross intersection for
through minor traffic with major traffic volume q5 .

The probability for this state is

p3 = 1 - w0 .

Thus the contribution of state 3 to the capacity of part II is

cII,3 = (1 - w0) . c(q5)
where

w0

(8)

= probability that 0 vehicles are in the storage space

No vehicles from movement 1 (volume q1) can directly (i.e. without being impeded by
movement 5) pass through the storage area in this state.

Werner Brilon, Ning Wu, Kerstin Lemke

page 7

4. State 4 :

For i = 0 (i.e. an empty storage area) no vehicle can depart the storage area even if the
major stream of volume q5 would allow a departure. If, however, a sufficient gap is provided
in the major streams of both parts of the intersection simultaneously, a minor street vehicle
from movement 8 can pass the whole intersection without being queued somewhere in the
storage area. The possible output of the storage area from movement 8 vehicles during this
state is
c4 = c(q1 + q2 + q5)

(9)

Thus, the contribution of state 4 to the capacity of part II is

cII,4 = w0 . c(q1 + q2 + q5)

(10)

Also vehicles from movement 1 can pass through the storage area in this state. The number
of vehicles from movement 1 which pass through the storage area in this state is
cII,4,q1 = w0 . q1

(11)

Here, cII,4,q1 does not mean the capacity for q1, but the demand on the capacity. The traffic
intensity of q1 should be less than the capacity of the part II c(q5). i.e. q1 is subject to the
restriction q1 < c(q5) . Otherwise, the intersection is overloaded and due to this nonstationarity no solution can be derived.
State 3 and 4 exclude each other. Therefore, the total maximum output of the storage area is
output = cII,3 + cII,4 + cII,4,q1
= (1-w0) . c(q5) + w0 . c(q1 + q2 + q5) + w0 . q1
= (1-w0) . c(q5) + w0 . [c(q1 + q2 + q5) + q1 ]

(12)

Werner Brilon, Ning Wu, Kerstin Lemke

page 8

One might argue that the derivations of cI,2 and cII,4 neglect the travel time of the vehicles
from part I up to part II . This, however, is justified: The probability that a minor street
vehicle will meet a sufficient gap in part I and part II at time tI and time tII (with tII = tI +
t

and

with

t = travel time between the stop lines of part I and part II ) is independent of the travel time
t if t = constant for all vehicles and if the two arrival processes in the major streams are
independent of each other. Therefore, the result is the same if t has a realistic positive value
or if t is assumed to be 0.
During times when the whole intersection is operating at capacity, due to reasons of
continuity, the maximum input and output of the storage area must be equal.
Therefore
input

output

(cf. eq. 6 and eq. 12)

i.e.:
(1-wk) . [c(q1+q2) + q1 ]+ wk . c(q1+q2+q5) = (1-w0) . c(q5) + w0 . [c(q1+q2+q5) + q1]
(13)
The total capacity

cT

for minor through traffic (movement 8) regarding the whole

intersection is identical to both sides of this equation minus q1 . In addition, since negative
traffic volumes are not possible cT must fulfill the restriction:
output q1 = cII ,3 + cII ,4 + cII ,4 ,q1 q1
cT = max
0

(14)

For the easiest case of k = 1 we get

w0 + w1 = 1
Together with eq. 13 and the subsequent explanation we get

(15)

Werner Brilon, Ning Wu, Kerstin Lemke

page 9

w0 = [c(q5) - c(q1+q2+q5)] / [c(q1+q2) - 2 c(q1+q2+q5) +. c(q5)]

wk = 1- w0 = [c(q1+q2) - c(q1+q2+q5)] / [c(q1+q2) - 2 c(q1+q2+q5) +. c(q5)]

(16)

For k > 1 some more general derivations are necessary. For these derivations we assume
the following simplifying conditions:
a) Let q2 and q5 be constant over time. Then also c(q2), c(q5) , and c(q2 + q5) are constant
over time.
b) We devide the continuous time scale into intervals of duration tf = follow-up time =
average time interval between the departure of two subsequent minor vehicles which enter
into the same gap of the major flow. It is also assumed that the minimum gap between
two vehicles of movement 1 is of the same size as tf .
Let

a =

probability that a vehicle enters the central storage area from intersection part I
during a time interval of duration tf .

b=

probability that a vehicle can pass intersection part II during a time interval
of duration tf .

a and b are variables which are only introduced for the following derivations. They need
not to be evaluated later for the application of the theory. Both a and b are looked at for the
fictitious case that part I and part II would be independent intersections. The follow-up time
tf for part I and part II should be of similar duration for this derivation. We now treat the
process of the number of vehicles in the storage space as a statistical process with Markowproperties. We then can say

w0 (t ) = w0 (1 a ) + w0 a b + w1 b (1 a )

(17)

= probability that no vehicle is queueing

in the storage area at time t
This is valid because the case of an empty queue at time t can be achieved by the following
possibilities:

Werner Brilon, Ning Wu, Kerstin Lemke

page 10

either: no queue at time t - tf (prob. = w0) and no arrival (prob. = 1 - a) during tf

or:

no queue at time t - tf (prob. = w0)

and one arrival (prob. = a) and one departure (prob. = b) during tf

or:

one vehicle queued at time t - tf (prob. = w1)

and no arrival (prob. = 1 - a) and one departure (prob. = b) during tf .

By similar considerations we get an expression for the probability of i vehicles queueing in

the storage space at time t :
wi(t) = wi-1 . a . (1-b)
+ wi . a . b
+ wi . (1-a) . (1-b)
+ wi+1 . (1-a) . b

(18)

Since k is the maximum number of vehicles in the storage space we get

wk(t)= wk . (1-b)
+ wk . a . b
+ wk-1 . a . (1-b)

(19)

Due to the assumed stationarity of the process w0, wi and wk do not depend on each other at
time t .
Equations 17 through 19 form a system of k + 1 equations which can be written as
w0 (a ab) + w1 (b ab)

=0

(20)

wi 1 (a ab) wi [(a ab) + (b ab)] + wi +1 (b ab)

=0

(21)

wk 1 (a ab) wk (b ab)

=0

(22)

For abbreviation we use

A = a a b

(23)

B = b a b

(24)

Werner Brilon, Ning Wu, Kerstin Lemke

page 11

Our system of equations 20, 21 and 22 then is written as:

=0

(0)

-Aw0

+Bw1

(1)

Aw0

-(A+B)w1

+Bw2

Aw1

-(A+B)w2

+Bw3

Awi-1

-(A+B)wi

-(A+B)wk-2

+Bwk-1

Awk-2

-(A+B)wk-1

+Bwk

=0

Awk-1

-Bwk

=0

(2)

=0
=0

...
(i)

+Bwi+1

=0

(25)

...
(k-2)

Awk-3

(k-1)
(k)

=0

From the first equation we get

A w0 = B w1

w1 =

A
w0
B

(26)

From the last equation we get:

A wk 1 = B wk
wk =

A
wk 1
B

(27)

If we sum up all our equations (0) through (i) we get:

A wi + B wi +1 = 0
wi +1 =

A
wi
B

(28)

The sequence of the probabilities, therefore, is forming a geometric series where each
subsequent term is resulting from the prior term by a multiplication with the factor y = A/B.
y=

A a ab
=
B b ab

(29)

Werner Brilon, Ning Wu, Kerstin Lemke

page 12

i.e.:

wi +1 = y wi

(30)

wi = y i w0

(31)

or:

According to equation 1 the wi (i = 0,...,k) are subject to the restriction

k

=1

i =0
k

w0 = 1

i =0

w0 y i = 1
i =0

Therefore:
w0 =

1
1 + y + y 2 +...+ y k
1

(32)

The sum in the denominator is the sum of a finite geometric series which is
k

i =0

y k +1 1
y 1

(33)

Thus, and with eq. 30 and 29 we get:

w0 =

y 1
y 1

(34)

wk =

y k +1 y k
y k +1 1

(35)

k +1

Werner Brilon, Ning Wu, Kerstin Lemke

page 13

Let us now recall eq. 13 and 14 and combine those with eq. 34 and 35 . Then we get
y k +1 y k
1 k +1
y 1

y k +1 y k
[c(q1 + q2 ) + q1 ] + k +1
c(q1 + q2 + q5 )
y 1

(36)

y 1
y 1
= 1 k +1 c(q5 ) + k +1 [c(q1 + q2 + q5 ) + q1 ]
y 1
y 1

Note that in this equation the capacities c(q2), c(q5) and c(q2+q5) as well as k are treated to
be known whereas the variable y has to be obtained from the equation. As a result we get:
y=

c(q1 + q2 ) c(q1 + q2 + q5 )
c(q5 ) q1 c(q1 + q2 + q5 )

(37)

Using this result for y we can now calculate the total capacity cT for the minor movement 8
using eq.14 .

y 1
y 1
cT = 1 k +1 c(q5 ) + k +1
c(q1 + q 2 + q5 ) + q1 q1
y 1
y 1

y 1
y 1
c T = 1 k +1 [ c ( q 5 ) q 1 ] + k + 1
c(q1 + q 2 + q5 )
y 1
y 1

(38)

It should be noted that for the special case of k = 1 using some algebra we get the solution of
eq. 16 which might give some confirmation for the above derivations.
For y = 1 ( i.e. c(q1 + q2) = c(q5) - q1

) this expression is not defined. By developing the

limiting case for y 1 we get

cT =

k +1

[ k (c(q5 ) q1 ) + c(q1 + q2 + q5 ) ]

(39)

At this point it should be noted that the capacities c(q1 + q2 +q5) and c(q5) can be
calculated by any useful procedure, e.g. by formulas from gap acceptance theory. But also
solutions from the linear regression method or Kytes method (for details cf. Brilon,
Troutbeck, Tracz, 1995) could be used.

Werner Brilon, Ning Wu, Kerstin Lemke

3.

page 14

CAPACITY ACCORDING TO GAP ACCEPTANCE THEORY

The most simple formula for the capacity of an unsignalized intersection with one minor and
one major traffic stream is Sieglochs (1973) formula. Several authors (cf. Brilon, Troutbeck,
Tracz, 1995) have shown that this formula produces also realistic results if the basic
assumptions for the formula are not fulfilled. The Sieglochs formula is shown as following:
c( q ) =

where

1 qt0
e
tf

c(q)

=capacity for the minor movement

tf

=follow-up time

(40)

(veh/s)
(s)

=average gap between two successive minor flow vehicle entering

into the same major stream gap
t0

=tc - tf /2

(s)

tc

=critical gap

(s)

=average gap between two successive major flow vehicle which - as a

minimum - is accepted by the minor stream vehicles to cross the
intersection
We have now to distinguish between different tc - and tf - values:
a)

tc - and tf - values for part I of the intersection (state 1 and state 2)

b)

tc- and tf - values for part II of the intersection (state 3 and state 4)

c)

tc - and tf - values for crossing part I and part II of the intersection simultaneously

in the case of k = 0 . It is realistic to assume that a driver who has to cross the
whole major street at one time without having a central storage area needs longer tc and tf - values than in case a) or b) .
It is justified to assume that the tc - and tf - values in case a) and b) are of the same
magnitude and that especially the tf - values between both cases are nearly identical. This
assumption is important for the following derivations.

Werner Brilon, Ning Wu, Kerstin Lemke

page 15

Realistic values for the tc - and tf - values can be obtained from table 1. The given critical
gaps tc and follow-up times tf are of realistic magnitude compared with the measurement
results worked out by the NCHRP-project 3-46 (Kyte e.a., 1995, working paper #16). Here
the critical gap and the follow-up time for the case without central reserve (k = 0) are larger
then for the two-stage priority case which seems to be more realistic.
k1
k=0

i.e. a central reserve

i.e. no central reserve

of variable (with k) width

part I

part II

case c)

case a)

case b)

tc

7.0 s

6.0 s

6.0 s

tf

3.8 s

3.8 s

3.8 s

Table 1:

Typical tc - and tf - values for two-stage priority situations within

multilane major streets under US-conditions

Based on eq. 40 with the assumption that all of the tf - values are nearly identical we can say:
c(q1 + q2 + q5 ) c(q1 + q2 ) c(q5 ) q1
=

c0
c0
c0

where

c0 =

(41)

1
tf

(veh/s)

= maximal capacity, for the case of no cross traffic

This relation makes it possible to standardize all of the capacity terms by c0 . If c0 is used in
units of veh/s also the other capacity terms must have this unit. Of course, also the unit
veh/h could be used for all of the capacity terms. Then it is useful to standardize also cT in

eq. 38/39 :
where

c$T =

cT
c0

(veh/s)

(42)

Werner Brilon, Ning Wu, Kerstin Lemke

c$T

page 16

(which has to be obtained from eq. 38/39) then can be expressed as a function of

c(q1 + q2)/c0 and

[c(q5) - q1 ] / c0

Thus it is possible to indicate the results of these

derivations by graphs (cf. Fig. 2).

0.9

0.9

0.8

0.8

0.7

0.7

0.6

0.6

0.5

0.5

0.4

0.4

0.3

0.3

0.2

0.2

0.1

0.1

0.0
0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

c(q1+q2)/c0 (-)

cT/c0 (-)

1.0

1.0

[c(q5)-q1]/c0 (-)

Figure 2: Total capacity

c$T = cT c0 as a result of eq. 42 (in combination with

eq. 38) in dependence of c(q1 + q2)/c0 and [c(q5) - q1 ] / c0 for k = 1 .

It is further justified to use graphs of this type with sufficient approximation also under
circumstances which differ from the conditions of gap acceptance theory, e.g.
-

if capacities c(q1 + q2) and

c(q5)

are computed from other theories than gap

acceptance or even if they should be measured,

-

if within gap acceptance theory the critical gaps tc are different for each part of the
intersection.

The only necessary condition for the application of these graphs is that the follow-up times tf
are of nearly identical magnitude.

Werner Brilon, Ning Wu, Kerstin Lemke

4.

page 17

LIMITATIONS OF THE THEORY

With a critical view on the theory which lead to eq. 38 we see that this theoretical concept has
to be treated with care. The concept would be true if we could estimate the capacities c(q1 +
q2 + q5) and c(q5) completely according to the fact that at a time, as a maximum, only k

vehicles can enter one major stream gap both in part I and part II of the intersection due to
the restricted storage space within the median reserve of the two-stage situation.

This

restriction applies especially for c(q1 + q2) and c(q5) . (state 1 and 3, cf. above) . This
restriction does not apply for c(q1 + q2 + q5) since during state 2 and state 4 (definition see
above) the number of minor stream vehicles departing during one large gap (being provided
simultaneously in major streams 1 and 2

as well as 5) is not limited.

Each of the

conventional formulas for the capacity c(q) (e.g. the Siegloch-formula eq. 40) are , however,
based on the assumption that during large major stream gaps a greater number of minor
stream vehicles can be accommodated, which is not true in the two-stage gap-acceptance
situation (state 1 and 3) since here the number of minor vehicles per gap is limited to k in
both parts of the intersection.
To take account of this limited validity of eq. 38 different approaches have been tested. The
derivation of an analytical formula which takes into account these effects seemed not to be
possible. Only a partial approach to the complete realistic truth was possible (cf. Brilon, Wu,
Lemke, 1995). Therefore, some approximations were necessary.

Werner Brilon, Ning Wu, Kerstin Lemke

5.

page 18

SIMULATION STUDIES

Therefore, and for the test of the theory leading to eq. 38, the solution has been further
investigated based on simulations. For this purpose a simulation model has especially been
developed (Lemke, 1995). The basic structure of the model is closely related to the ideas of
KNOSIMO (cf. Grossmann, 1992). The important features can be characterised as follows:

The headways in the major streams are distributed according to a hyperlang-distribution

(cf. Dawson, 1969 ; Grossmann, 1991).

The critical gaps and the follow-up times are distributed according to an Erlangdistribution with the parameters given by Grossmann (1991) which are also used in
KNOSIMO.

Both these assumptions together relate the model closer to reality than the theoretical
derivations mentioned above.

On the other side, the following assumptions are a

simplification compared to reality. They do, however, correspond to the assumptions of the
theory described above.

No delays due to limited acceleration or deceleration of the vehicles are taken into
account.

The travel time t between the two parts of the intersection has not been regarded; i.e.
t = 0. (cf. argumentation following eq. 12).

Each minor street driver has a minimum delay of tf at the first part of the intersection,
also if no major stream vehicle is nearby. This simulates the time which a driver needs to
realise the traffic situation on the major street when he is first approaching the
intersection. This time margin is also necessary for the driver to decide if he can enter the
intersection. Such an orientation time is not applied for vehicles entering the second part
of the intersection since here a better visibility is assumed.

All traffic volumes are kept constant over time.

Werner Brilon, Ning Wu, Kerstin Lemke

page 19

The program is organised such that a constant queue in front of the first stop line of
movement 8 is always maintained. Thus, the maximum number of vehicles which can
enter the intersection can be evaluated.

This number is the representation of the capacity for movement 8 . A comprehensive set of
simulation runs has been performed for different parameters q1 , q2 , q5 .
Different attempts have been made to find an easy to be used approximative description of the
results. Several of these attempts are described in Brilon, Wu, Lemke (1995) together with a
statistical assessment of their precision. A good compromise between easy application and
highest precision seemed to be the following solution. Instead of cT we use a more realistic
solution cTr which is obtained as a good approximation to the simulation results.
cTr = cT

where

cTr

(veh/s)

(43)

= realistic total capacity for movement 8

(minor through traffic)

cT

= result from the theoretical approach obtained from

eq. 38

= adjustment factor
1
=
. k)
1 0.32 exp( 13

for k = 0
for k > 0

(44)

An even better solution for the correction term is given by the following formula. This
approach, however, has the drawback of a rather complicated use. Thus it is recommended
rather for computer applications.
= 1 0,245

e2 e5
k 1,65

with

( 2 z2 )
e2 = 2
e ( 2 z2 )
( k )!
k

2 = a + b z5 + c z52

(45)

Werner Brilon, Ning Wu, Kerstin Lemke

page 20

( z )
e5 = 5 5 5 e ( 5 z5 )
( k )!
k

5 = a + b z2 + c z2 2
z2 =

c( q 2 )
c0

c0 =

1
(veh / s)
tf

z5 =

c( q 5 )
c0

or

c0 =

3600
(veh / h)
tf

Within this set of equations the following parameters should be applied:

a = 2.788
b = -1.259
c = -0.576
These solutions for the total capacity cTr of movement 8 approximate the simulated results
with a standard deviation s (between results for cT

being simulated and those being

estimated from eq. 42) according to table 2 . Other solutions with smaller deviations but
more complicated formulas for the calculation of realistic cTr can be obtained from Brilon,
Wu, Lemke (1995). The effect of the correction term is also illustrated in Fig. 3 .
s
q1 = 50

q1 = 100

q1 = 200

=1

29

30

32

eq. 44

18

18

19

eq.45

15

veh/h

veh/h

veh/h

Table 2:

Standard deviation s for computed cT - values compared to the

simulated results for two different approaches of the correction term

Werner Brilon, Ning Wu, Kerstin Lemke

page 21

Capacity CT from without correction (eq.38/39)

[veh/h]

1200

k=1

1000

800

q1 = 100
600

a)

400
s = 36,2 veh/h
200

0
0

200

400

600

800

1000

1200

Capacity CT from simulation [veh/h]

1200
Capacity CT with correction (eq.43/45)
[veh/h]

k=1
1000

800

q1 = 100
600

b)

400
s = 7,9 veh/h
200

0
0

200

400

600

800

1000

1200

Capacity CT from simulation [veh/h]

Figure 3: Comparison of simulated capacities cT for movement 8 and calculated

cT - values . The simulated results are regarded as the true values.

a) calculation without correction term for eq. 38/39
b) calculation with correction term (eq. 43/45).

Werner Brilon, Ning Wu, Kerstin Lemke

page 22

At this point we can conclude the steps of computation which are necessary to estimate the
capacity of an unsignalized intersection where the minor movements have to cross the major
street in two stages:
q1

volume of priority street left turning traffic at part I

q2

volume of major street through traffic coming from the left at part I

q5

volume of the sum of all major street flows coming from the rigth at
part II. Of course, here the volumes of all priority movements at part II
have to be included. These are: major right (6, except if this movement
is guided along a triangular island separated from the through traffic) ,
major through (5), major left (4); numbers of movements according to
HCM 1994, chapter 10.

c(q1 + q2)

capacity at part I

c(q5)

capacity at part II

c(q1+q2+q5) =

capacity at a cross intersection for minor through traffic

with a major street traffic volume of q1+q2+q5
(all capacity terms apply for movement 8 . They are to be calculated by
any useful capacity formula, e.g. the Siegloch-formula, eq. 40)

y=

c(q1 + q 2 ) c( q1 + q 2 + q 5 )

cT =

k +1

cT ( y =1) =
cT
with

(37)

c( q5 ) q1 c(q1 + q 2 + q5 )

{y ( y

k +1

1 [c(q5 ) q1 ] + ( y 1) c(q1 + q 2 + q5 ) for y 1

[ k [ c( q ) q ] + c ( q
5

+ q 2 + q5 )

for y = 1

(38)
(39)

total capacity of the intersection for minor through traffic

1
=
. k)
1 0.32 exp( 13

for k = 0
for k > 0

Equations 38, 39 and 40 are only valid for c(q5 ) q1 > 0.

(44)

Werner Brilon, Ning Wu, Kerstin Lemke

6.

page 23

GRAPHS FOR PRACTICAL APPLICATION

The results for the theory given in this paper are illustrated in Fig. 4 for k = 1 and 2 . Here
the capacities c(q1 + q2) and c(q5) can be introduced independent of the type of formula
from which they have been determined. Another advantage of these graphs is that they can be
applied with each arbitrary value of q1 .
For example, we look at two-stage priority intersection with the traffic volumes q1=100 veh/h,

q2=600 veh/h and q5=400 veh/h. Let there be two possible storage spaces within the central
reserve (cf. also Fig. 1). The capacities for movement 8 crossing the intersection separately
can be calculated from the Sieglochs formula (eq.40). The corresponding values of tc and tf
can be obtained from table 1. Then the parameters for the entering application graph (Fig.4)
can be calculated as following:
Part I:

Part II:

1 ( q1 + q2 )t0
e
tf

c ( q1 + q 2 )

c( q 5 )

1 q 5 t 0
=
e
tf

1
e
.
38

(100 + 600 )
3.8
( 6 )
3600
2

400

= 0.119

(veh/s)

= 0.167

(veh/s)

3.8

1 3600 ( 6 2 )
=
e
.
38

(cf. eq.40)
And with

c0

1
tf

q1

100
3600

1
38
.

we obtain the parameters for using Fig.4:

c( q 1 + q 2 )
c0

0119
.
0.263

c( q 5 ) q1
c0

0167
.
0.028
= 0.53
0.263

= 0.45

= 0.263

(veh/s)

= 0.028

(veh/s)

Werner Brilon, Ning Wu, Kerstin Lemke

page 24

1
0.9

0.9

0.8

0.8

[c(q5)-q1]/c0

0.6

0.6
0.5

0.5

0.4

0.4

0.3

0.3

0.2

0.2

0.1

0.1

cT/c0

0.7

0.7

a)

0
0

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

c(q1+q2)/c0

0.9

0.9

0.8

0.8

0.7

0.7

0.6

0.6

0.5

0.5

0.4

0.4

0.3

0.3

0.2

0.2

0.1

0.1

cT/c0

[c(q5)-q1]/c0

b)

0
0

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

c(q1+q2)/c0
Figure 4: Capacities c$T = cT / c0 (cf. eq. 42 ) for movement 8 in relation to
standardized values of capacities and of q1 (calculation with correction
term , eq. 43/45)
a) k = 1
b) k = 2

Werner Brilon, Ning Wu, Kerstin Lemke

page 25

With these two parameters we obtain the relative capacity for the movement 8

c$T

= c T / c0

= 0.36

(Fig.4, b) ).

Therefore, the absolute capacity for movement 8 is

= c$T c0

cT

= 0.360.263
= 0.095 veh/s
= 342 veh/h.
If gap acceptance theory is applied (eq. 40) to estimate the basic capacity terms ( c(q1 + q2)
and c(q5) ) and if the tc - and tf - values are known, then the capacity for movement 8 can
also be indicated by graphs directly depending on q2 and q5 . Then, however, one graph has
to be indicated for each possible q1 - value. This type of graphs using tc - and tf - values
from table 1 (right columns) is given in Fig. 5 as one example.
1600
1400
C T =100 veh/h

q5 [veh/h]

1200
1000
200

800
300

600
400

400

500
600

200

700
800

0
0

200

400

600

800 1000 1200 1400 1600

q 2 [veh/h]

Figure 5: Capacities cT for movement 8 depending on traffic volumes q2 and q5

with q1 =0 (calculation with correction term , eq. 43 / 38 / 45 , k = 1).

Werner Brilon, Ning Wu, Kerstin Lemke

page 26

Of course, the same theory as it has been described here can be used to determine the capacity
of the minor left turner (movement 7) under two-stage priority conditions. If there is no
separate lane for this movement in the central storage area the so-called mixed lane formula
(eq. 10-9) of the HCM, 1994) has to be used to calculate the total capacity for movements 7
and 8 .
Delay estimations for the two-stage priority situation can be performed using the concept of
reserve capacities (cf. Brilon, 1995) or the general delay formula by Kimber, Hollis (1979).
An easy to be understood procedure for the practical application of the theory presented here
still has to be developed for the future HCM . Also some tests of this theory against
measurement data which are available from field studies in NCHRP-project 3-46 are
desirable.

Werner Brilon, Ning Wu, Kerstin Lemke

7.

page 27

CONCLUSION

The two-stage priority situation as it exists at many unsignalized intersection within multilane major streets provides larger capacities compared to intersections without central reserve
areas. Capacity estimation procedures for this situation have not been available up to now.
The paper provides an analytical solution for this problem. In addition, simulation studies
lead to a correction of the theoretical results. Based on these derivations a set of graphs could
be evaluated which enable an easy estimation of the capacity at an unsignalized intersection
under two-stage priority. These graphs are ready to be used in practice.
Nevertheless, an empirical confirmation of this model approaches would be desirable. Also
the question of the validity of the model for larger k-values should be discussed. It is
questionable if the theory also applies for a grid of one-way street networks. Also if these
questions should be addressed in the future, the presented theory is recommended for use at
unsignalized intersection in practice.

Werner Brilon, Ning Wu, Kerstin Lemke

page 28

REFERENCES

1. Harders, J. Die Leistungsfaehigkeit nicht signalgeregelter staedtischer Verkehrsknoten

(The capacity of unsignalized urban intersections). Schriftenreihe Strassenbau und
Strassenverkehrstechnik, Vol. 76, 1968.
2. HCM. Highway Capacity Manual. Transportation Research Board, Special Report 209,
new edition, Washington 1994.
3. Brilon, W., R. J. Troutbeck, and M. Tracz. Review if International Practices Used to
Evaluate Unsignalized Intersections. To be published as TRB Circular in Nov./Dec. 1995
4. Siegloch, W.

Die Leistungsermittlung an Knotenpunkten ohne Lichtsignalsteuerung

(Capacity calculations for unsignalized intersections). Schriftenreihe Strassenbau und

Strassenverkehrstechnik, Vol. 154, 1973.
5. Brilon, W., N. Wu, and K. Lemke. Two-Stage Priority . Working Paper #24 of NCHRPProject 3-46.
6. Lemke, K.

Berechnungsverfahren fr Knotenpunkte ohne Lichtsignalanlagen -

Weiterentwicklung - (Calculation Methods for Unsignalized Intersections - Additional

Development -) Diplomarbeit, Lehrstuhl fr Verkehrswesen, Ruhr-Universitt Bochum,
Aug. 1995.
7. Grossmann, M. Methoden zur Berechnung und Beurteilung von Leistungsfaehigkeit und
Verkehrsqualitaet an Knotenpunkten ohne Lichtsignalanlagen (Methods for calculation
and assessment of capacity and traffic quality at intersections without traffic signals).
Ruhr-University Bochum, Germany, Chair of Traffic Engineering, Vol. 9, 1991.
8. Dawson, R. F. The Hyperlang Probability Distribution - A Generalized Traffic Headway
Model. In: Schriftenreihe Strassenbau- und Strassenverkehrstechnik, Vol. 86, pp. 30-36,
Bonn 1969.
9. Brilon, W. Delays at oversaturated unsignalized intersections based on reserve capacities.
Preprint of a presentation at the TRB-conference, Washington, Januar 1995, to be
published in a TRB-record.
10.Kimber, R. M. and E. M. Hollis. Traffic queues and delays at road junctions. TRRL
Laboratory Report LR909, 1979.