11

process on an explanatory model for road-user be- section. Biometrlka, Vol. 54, 1967, pp. 657-658.
havior, rather than on a merely statistical model, 12. K. Harders. Leistungsfahigkeit nicht signalregelter
ought to fav-o r this responsiveness. stadtischer Strassenknoten. T. H. Hannover,
thesis, 1976.
REFERENCES 13. W. Siegloch. Die leistungsermittlung an Knoten-
punkte ohne Lichtsignalsteurung. Strassenbau und
5. S. Nordqvist. Gators och vligars kapacitet. IVA, Strassenverkehrstechnik, Vol. 154, 1973.
TFK-medd, No. 39, 1958. 14. E. Grant. Traffic Capacity of Small Roundabouts.
6. Highway Capacity Manual. HRB, Special Rept. 87, Traffic Engineering and Control, Vol. 10, No. 12,
1965. 1969.
7. Capacity of At-Grade Junctions. Organization for 15. F. Wagner. An Evaluation of Fundamental Driver
Economic Cooperation and Development, Research Decisions and Reactions at an Intersection. HRB,
Group TS, 1973. Highway Research Record 118, 1966, pp. 68-84.
8. G. Jessen. Ein Richtlinienvorschlag fiir die 16. P. Solberg and J. Oppenlauder. Lag and Gap Ac-
Behandlung der Leistungs-fahigkeit von Knoten- ceptance at Stop Controlled Intersections. HRB,
punkten ohne Signalregelung. Strassen Verkehrs Highway Research Record 118, 1966, pp. 48-67.
Technik, Vol. 12, No. 7-8, 1968. 17. A. Miller. Nine Estimators of Gap Acceptance
9. V. Surti. Operational Efficiency Evaluation of Parameters. Proc., 5th International Symposium
Selected At-Grade Intersections. HRB, Highway on the Theory of Traffic Flow and Transportation,
Research Record 321, 1970, pp. 60-73. Berkeley, 1971.
10. J. Tanner. A Theoretical Analysis of Delays at an 18. R. Ashworth. The Capacity of Priority Type In-
Uncontrolled Intersection. Biometrika, Vol. 38, tersections With a Non-Uniform Distribution of
1962, pp. 383-392. Critical Acceptance Gaps. Transportation Re-
11. J. Tanner. The Capacity of an Uncontrolled Inter- search, Vol. 3, No. 3, 1969.

Part 3. Capacity of Signalized Intersections

Karl~Lennart Bilng, Swedish Transport Research Commission,
Royal Academy of Engineering Sciences
Calculations of signal timing and capacity have been performed in Sweden The manual was completed and published in Swedish in
by using methods based on the 1950 and 1965 Highway Capacity Manuals. 1977 (4).
These methods, however, give results that differ by as much as 50 percent An overview of the different phases of the work is
from observed flows. A comprehensive development was therefore under·
taken by the Swedish National Road Administration from 1971 to 1976. presented as part 1 of this paper.
After extensive literature reviews, theoretical analyses, and field studies,
a method was developed for calculating signal timing, capacity, queue SCOPE AND OBJECTIVES
length, proportion of stopped vehicles, and delay . The method is based
on calculating saturation flows separately for each lane, which makes the The manual was designed for calculations of capacity
method applicable to all geometric designs and phasing schemes. Adjust- (defined as maximum flow at given conditions), queue
ment factors for vehicle-actuated control are also included. Study em· length, proportion of stopped vehicles, and delay.
phasls was on the relations among approach width, lane markings, and
capacity: conflicts between left-turning and opposing flows; and con· These calculations serve to describe the consequences
flicts between turning vehicles and pedestrians. The signal-timing method of a given set of geometric and traffic parameters.
is based on the minimum average delay criterion proposed by Webster. No recommendations for design standards are given,
because the manual does not include all aspects, such as
For 20 years the calculations of signal timing and ca- safety, capital costs, and external effects on the en-
pacity done in Sweden have been based on a manual de- vironment, that would have to be considered.
veloped by Nordqvist in 1958 (5). This manual was, in The manual does include forms and examples for
turn, largely based on the 1950 Highway Capacity practical. application. The purpose of this paper is to
Manual (HCM) . Some attempts have also been made to present an overview of the proposed method and some
promote the use of the 1965 HCM (6). · of the material dealing with signalized intersections.
These manuals, however, have been found to have The method is based on theoretical models supported
some serious drawbacks. They do not give optimum by a limited number of field surveys. Each model ex-
cycle time, and capacity values often differ from ob- plains the behavior of traffic in a critical conflict or
served flows by as much as 30-50 percent. Further- geometric design that affects the discharge rate at the
more, no guidance is provided for calculating more stop line. By identifying the true reasons for each
complex ge ometric designs and signal phasing schemes. effect, adjustment factors such as city size and location
To overcome these deficiencies, a comprehensive within the city can be neglected or given only minor im-
study including the development of calculation methods portance. This should greatly improve the accuracy of
for roads as well as for different types of unsignalized the results as well as the possibility of their being
and signalized intersections was initiated in 1971 by the reproduced.
Swedish National Road Administration (NRA). Chief
investigat or for the \vork was Professor Stig Nordqvist METHOD PROCEDURE
at Vattenbyggnadsbyran (VBB); I was responsible for
the part covering signalized intersections; and Arne Each lane is treated individually in the method. This
Hansson was responsible for unsignalized intersections. means that the manual can be used for practically all
12

intersection layouts and phasing scheme designs. lane, and the critical conflict, defined as the point hav-
The computational procedure involves the following ing the highest total of ratios q / s 1 of adjoining lanes,
basic steps: (q / s 1)m•• , is identified (Figure 10). The sum represents
the degr ee of saturation of tJie inte rsect ion if ther e are
1. Preparation no losses in effective green time.
a. Determination of phasing scheme, lane division, e. Calculation of cycle, split, and green times .
and lane types. These calculations, carried out according to Webster
2. Signal Timing: First Round ~) , give the signal timing that r esults in minimum
b. Calculation of the saturation flow (s) for each lane. average delay for fixed t ime operation of the signals.
Saturation flow, s, is defined as the highest stable flow 3. Signal Timing: Second Round
in vehicles per hour of green (vphg) during existing f. Renewal of steps b-e for improved accuracy.
conditions. First a base value of s is obtained. It takes The saturation flow for lanes having turning traffic in
into account the proportion of turning vehicles nnd the r.onflir.t with oppoRing vP.hicle flows or having pedes-
degree of conflict with other vehicles and pedestrians in trian crossings with green in the same phase is a func -
the intersection with green in the same phase. This tion of the length of the green time and the cycle time.
base value is then adjusted for conditions other than The second round includes procedures for more precise
normal regarding the width, slope, length, and mark- calculations of s for such lanes. The signal timing and
ings of the lane, the proportion of heavy vehicles, and other parameters are derived in the first round (b-e)
so forth. as input. This is only necessary if consequences for
c. Distribution of the flow on different lanes. If the individual lanes, such as queue length or delay, are to
flow in a lane, q 1, is not known, q 1 is calculated so that be calculated.
equal values of q/ s 1 are obtained for the adjacent la ne 4. Consequences
or lanes with the same direct ion in the approacb (Fig- g. Calculation of capacity, que ue length, proportion
ure 9). This assumes that a driver arriving at the ap- of stopped vehicles, and delay. Once s, q, and the
proach selects the lane that will minimize his or her signal timing obtained from steps a-f are known, alculat -
delay before he or she crosses the stop line. ing the de manded consequences is a simple ai·ithmetic
d. Identification of the critical conflict point in the procedure . Capacity is defined as the l:lighes t st able
intersection. The q/ s 1 ratios are calculated for each flow in vehicles per hour (vph) d ul'ing existing conditions .
The manual is divided into a num ber of sepa1·ate work
Figure 9. Method used to distribute the flow 111uJJ1e11t~ Lhet.L a.i·c i1u.111be;red c0r.1s~~;....iivcl:-; ir.. !!ic o~dc~
between lanes in the same direction. in which they are to be performed. In Figure 11 a flow-
chart of the method is illustrated.

IMPORTANT FEATURES
'2
~ Lane Configur ation and Clas sification

The models that form the basis of the method assume

that each lane carries a single flow of vel:licles through
Q f/h Figure 11. Flow chart for the manual.
ql ; Q • '1
S1 + 52 + SJ
1. Gearetric design

q2 ; Q • •z Traffic volures
'1 • 'z t '3 Lane division
2• Selection of ?JaSe schane
q3 ; Q • '3
s 1 + s2 + s 3 3• Oet:ennl.nll ti.On of lane types

SIGIAL TIMIN> 4. calculations of saturation flows (s)

FIRST KXW 5. Adjust:lle1ts of s
Figure 10. Definition of critical conflict. 6• Influence on s of reduced length of
awroach lanes
7. calculation of fl0"1 distribution (q)
degree of saturation q/s, .L; (q/s)
CRITICAL
am split

CONFLICT
8, calculation of signal timing

SIGNAL TIMING 9• Control of mirumum green, etc

SJ:XX:ND RlllID 10. calculation of s
ll. Adjust:lle1ts of s
12. calculation of fl0"1 distribution,
degree of saturation, and split
l3. calculation of signal. timing

17 . Dalay

16. Proportion of
vehlcles
 13

the intersection. It is therefore important to make sure possible, one should assume a minimum lane width of
that actual traffic behavior corresponds to the lane 2.5 m.
markings. If this is not the case, the calculations The relations among capacity, approach width, and
should be performed with the number of lanes actually number of marked lanes in the approach have been
formed by moving traffic. evaluated by means of before-and-after field measure-
If field observation in the actual intersection is not ments at intersections with different lane arrangements.
Figure 12 shows examples of results from these experi-
ments and indicates that capacity is a function of both
Figure 12. Relations among saturation flow, approach width, approach width and number of lanes. The model deals
and number of marked lanes. with this by making separate calculations of the capacity
Saturat io11 flow
for each lane and by adjusting for lane widths other than
vphg
3 .0 m and for absence of lane markings (Figure 13).
To simplify handling of the method, the lanes are
5000
classified as different types, d ~pending on the presence
of turning traffic and the degree of conflict experienced
by the traffic in the lane. In Figure 14 seven different
4000 lane types are defined; lanes with turning traffic that
conflicts with both opposing vehicle traffic and pedes-
trians discharged in the same phase may be assigned
3000
to more than one type, for instance to D/F or E/G.

Base Values for Saturation Flow

1 I

2000
, I
The base value for the saturation flow (s)for lane type A
/'" (only through traffic) is set at 1700 vpbg and for type C
(only turning traffic witllout conflict) at 1500 vphg. For
type B (some turning traffic without conflict) s varies
1000
between 1700 and 1500 vphg, as a function of the per-
centage turning (0-100 percent).
For lane types with some degree of conflict between
turning traffic and opposing flows or pedestrians (types
12 45676610 D, E, F, G), determining s is more complicated .
Approach vidth m
Conflicts Between Left-Turning and
Opposing Flows
Figure 13. Adjustment 1.llVT1TTimTTTT1TTT1rnTmTTITITTrnTTTI11
factors for saturation From practical experience and limited field measure-
flow. 1.0 ttttttfllftto~tttlHtttttttttltttHftttH
ments it was concluded that a primary source of error
0 .9 tttttttllttttttltttlltlttttttiti:l:HiHftttH in older manuals was inadequate handling of conflicts
between left-turning vehicles and opposing vehicles on
0 •8 tttttttllttttttltttlftftt+ltttfl'\ott:Hfttttl two-way streets. Thie is particularly important for
normal, two-phase controlled intersections where all
0.1 o.z 0.3
left turners have to face this conflict. In order to over-
come this weakness, a thorough analysis was performed,
Proportion trucks and through buses and a model, suggested by Gordon and Miller in 1966
~
...,0 (20), was applied for stepwise calculations of the dif-

:~11111
ferent stages of the conflict (Figure 15). These s tages
~ are as follows:
j 1. First part of the green phase, gk, is when queue
."..," ·10 10
discharge from the opposite direction blocks left-turning
""
<
(S) vehicles in the lane. During gu only through or right-
Downward ...,_ ~ Upward turning vehicles, Nk, can be discharged from the lane
Slope and then only until the lane becomes blocked by queueing
left turners.
2. Remainder of green, gg = g - gk, is the time dur-
ing which left-turning vehicles can pass when accept-
1.0
able gape occur in the opposing flow. The total dis-
charge during this stage is noted as Ng = gg x s,.
0.9 3. Inter-green period is that time during which ve-
hicles stalled in the intersection can pass, N.,
0.8 Provided that the signal timing is known from the
first round of calculations, gk can be estimated as the
0.7
time required to discharge the queue that has been
1.5 3.0 3.5 4.0 4.5 (m) formed in the opposite direction during the previous red
Lane width phase:
lanes with turning traffic
lanes with only through traffic
(8)
marked lanes
unmarked lanes (assumed)
14

Figure 14. Classification of lane types.

TYPE DESCRIPTION EXAMPLE

A ONLY THROUGH TRAFFIC

i
B
SOME TURNING TRAFFIC
WITHOUT CONFLICT
r
c ONLY TURNING TRAFFIC
I
D
SOME TURNING TRAFFIC IN
CONFLICT WITH PEDESTRIANS

ONLY TURNING TRAFFIC IN

r•• I

+
I •
E CONFLICT WITH PEDESTRIANS I

F
SOME 'l'UHNlNG 'l'RAFFI C IN
CONFLICT WITH OPPOSING TRAFFIC
• +{
ONLY TURNING TRAFFIC IN
G CONFLICT WITH OPPOSING TRAFFIC
'1
D/ F
SOME TURNING TRAFFIC IN
CONFLICT WITH OPPOSING TRAFFIC
AND PEDESTRIANS

ONLY TURNING TRAFFIC IN

t.;UNl'L.U..:·1· WJ.'.l.'ti U.t'.t'Vo::>J.!'IV J..tu-\r s:: .1.'-
:·1,
++ ...,
I
E/G AND PEDESTRIANS
+
Figure 15. Calculation of saturation (1 I (2 I I 3I
flow for lanes with left turning

'
Nk N gg s N
vehicles. g g r
L.J- r~
.
! i
J ,_,[1
.
!

gk gg

Figure 16. Number of 15

I
where
vehicles, Nk, that can be
discharged during gk. •
30 g length of green phase in the opposite direction;
10 c cycle time;
q, flow vph in opposite direction; and
• s. saturation flow vphg in opposite direction .

The term 1/ (1 - q / s 0 ) accounts for the cumulative ef-

£ fect of vehicles ar';.iv ing at the approach before the
" queue bas been completely discharged. Numerical
....""
0
,-i
o. o 0.2 0.4 0.6 0.8 1.0 values for all variables are obtained from the first
round of calculations for the opposing lane with the
"'~ (a) Proportion of left turns
highest q/ s ratio .
...
0 With gk known, the number of vehicles, Nk, that can
0
be discharged is a function of the proportion of left-
"' 15
.~ turning vehicles, p1, in the lane and the number of left-
'\3 0
A
turners that can queue in the intersection without block-
ing other vehicles in the same lane. Nk can be solved by
I'
10 using general probability theory with the following re-
IS
'°.... sults in Figure 16, where (a) is the case where no left-
.... turning vehicle can queue without blocking the lane:
I
N- 1
Nk = ~ [ix P1 X (I - P1) 1] + N x (I - p 1)N (9)
I l= J
0 I I
a.a a.4
0.2 0.6 0.8 and (b) is the case where one left-turning vehicle can
(b) Proportion of left turns queue without blocking the lane:
 15

Figure 17. Accepted critical gap, ag, in the conflict

between left-turning and opposing vehicles. CROSS-STREET ~IIDTH
bk(m)

510 >la
OPPOSITE
APPROACH WIDTH ~8 4.8 5.3
bf (m) >8 5.3 5. 8

Figure 18. Saturation flow a = 4.3 s The s aturation flow s . during g 8 can be expressed as
s 9 during g9 • 9
0.5
(I I)
0.4
where ar is move-up time, which is 1/s.
o .o~
0.3 In the field studies ar was foWld to be closely cor-
0.2
•.. related with a. (ar = 0.54 a 8). In the manual this is pre-
0 .2
sented in a separate graph for each value of a. (Figure
18), where critical gap a 8 = 4.3 s.
0.1
...ta
o .3
0.4
The value for N is then obtained as
HEEEBE
~ 1000 1500 2000 2500 vph (12)
Total. opposing flov
The number of vehicles discharged during inter-green,
N,, is obtained from measurements or observations of
the space for queueing vehicles inside the intersection.
N-2 In Sweden an average space of 8 m per vehicle is as-
Nk = ~ [i(i +I) xpf x (1 - p 1 );] sumed. Care should also be taken to ensure that the
l=t inter-green is sufficiently long to allow these vehicles
+(N - l)xN (p1 (1- p1)N·t] +N x (1-pi)N (10) to clear the intersection before the next phase begins.
Otherwise the intersection can become seriously
where N is the maximum number of vehicles that can be blocked.
discharged during green (N = g x s ). The discharge, Nu The resulting saturation flow for the lane in vphg is
during g, is a function of gg, the proportion of left- obtained as
turning traffic, p1, and the possibilities for left-turners
to cross in gaps in the opposing flow. This conflict is s = (3600/g) (Nk + N• + N,) (13)
similar to the one that occurs in an unsignalized inter-
section. The capacity for left turns is in this case a The complex and thorough procedure described above
fWlction of the accepted critical gap a. and the distribu- can only be carried out when signal timing and satura-
tion of gaps in the primary stream. tion flows have been estimated in the first round of
Measurements of a. were performed at 10 signalized calculations. The manual gives two simplified graphs
intersections with a method described by Hansson in for the first calculation round (Figure 19) for this pur-
part 2; a. was found to vary as a function of the width of pose that are based on series of calculations with the
the opposite approach, bf, and the cross street, bk, as second-round procedure for normal types of intersec-
presented in Figure 17. For busy intersections in large tions. In these graphs s is determined by using the
cities, the values obtained for a 8 were 0.5 slower than ratio of left-turning vehicles and the total opposing flow
the values in Figure 17. as inputs.

Figure 19. Saturation flow for lanes I I .

11aa 11ao ........
with left-turning vehicles in conflict
with opposing flow. 15aa '\ I'\
I
.... _ 1sao
::'.!
- \'\ .... ,_
' ,...
"'"1'
.....
100 ,\'
' ....
100

,, '' ' "' .b ''

"'-
200
t!l ' .....
100

i
bO

1aaa ''
'- ,.... ,._ I
r-...
... ....
I"~ 300

400 .c:
go
.c:
~
1000
i\ I '\ '\.I'
\':" 1\..1\,. ...
:'\ ..... ...
..,
...
300

.c:
go
"
""-
...
\ :>

d
:>
0 \ r-...
' .....
500
:>
....0
0
....
....
\ '\
\\ "
""",'... 1'....
' """ ~
......' .._,
-
800
.... d
..0". .

·-·.
' ' ....
~ 100 700

.....0" 5aa
800 !:O
....." sao .... BOO 00
..". .
..... ~ .... ...... 000
~ ~ ~ 0 00 ... 000
... r- 100 "'A.
0 .,.." ,.... 1 100 "'0A.
~" I I
1200
!S !I}
1 200
0
A

"' a.a a.z a.4 a.6 a.8 i.a a.a a. z o.4 a.6 o.a i.o

Proportion le~ turn s Proportion le~ turns

~o le~-turning vehicJ e can One left-turning vehicle car.

queue without blocking the lane queue vithout blocking the lane
16

Figure 20. Calculation of saturation SPACE

flow for lanes with right-turning
movements.
PEDESTRIAN TRJl..JECTORIES

A B c D
! ME

GREEN T ME

Conflicts Between Turning Vehicles q" pedestrian flow, in number of pedestrians

=
and Pedestrian Movements per second, two-way; and
N = number of turning vehicles that can pass
In two-phase signal control for i;ntersections of two-way during inter-green.
streets, right-turning as well as left-turning vehicles
are usually affected by pedestrian movements in the TA - T 0 are functions of the length of the crosswalk
crosswalk of the leg into which the turn is being made. and of the size and d'iffusion of the pedestrian platoons
This is particularly true for Sweden, where right turns formed during the red phase. The equation, however,
on red are not permitted. An attempt was therefore includes too many vai·iables to allow graphic presenta-
made to analyze these conflicts ancf to develop a theo- tion. It has therefore been evaluated for a numbe1· of
retical model as a basis for a computational procedure typical cases and the results have been put together in
in the manual (3). Table 3.
Collecting data with time-lapse photographs com-
bined with inductive loops for vehicle detection was
performed in four intersections with the described con- Effect of Restricted Len h of
flicts. A theoretical model was then developed to Approach Lanes
describe the behavior of a vehicle and a pedestrian as
a function of the relative time advantage aby for the In built-up areas, the length of the curb lane is often
party first arriving at the collision point. The reason limited by parking, bus stops, or narrowing street
for this was that it was felt that the right-of-way for width . If the curb lane in an approach is too short,
the pedestrians during green was not fully respected capacity can fall. Two different cases can be dis-
by the turning vehicles. The model was therefore de- tinguished (Figure 21).
signed to enable tests of alternative hypotheses of this
behavior. Case A. Curb Lane Serving Right-Turning
The results from the field studies were evaluated and Through Vehicles
manually, and aby was derived by probit analysis. Th~
vehicles usually required a positive time advantage of A reduction in saturation flow, s, of the curb lane
3.2 s if they were not to give way to a pedestrian (stan- occurs if the available lane length, 1, is smaller than
dard deviation 4.3 s). the space occupied by the maximum number of vehicles
A capacity model based on division of the green phase (8 m per vehicle) that can be discharged during the
into four parts was developed. Figure 20 shows A and C, green time, g.
where no vehicles can pass because pedestrian platoons
have formed during red, and B and D, where a random 1/8 < (s x g)/3600 (15)
arrival of pedestrians into the conflict zone is assumed.
Turning vehicles can pass if their time advantage is The reduced saturation flow s' becomes
sufficiently great.
The capacity of the turning flow can be expressed as s' =(3600 x 1)/(8 x g) =(450 x 1)/g (16)

(14) Case B. Curb Lane Serving Only

Right-Turning Vehicles
where
In this case the queue formed during red in the nearby
Kr = largest number of turning vehicles that lane might bloc!.< the curb lane from being used to its
can pass per green phase per lane; full length. The likelihood of such blocking occurring
TA - T 0 = starting point of intervals A-D; is a £unction of 1, g, and the ratio of right-turning ve-
hicles in combined curb lane and adjacent lane. The
number of right turns, N, that can be made during a
green phase has been solved with probability theory
Figure 21. Restricted CJ\.SE JI. CJ\.SE B (Figure 22).
length of approach lanes. The adjusted saturation flow s' is obtained as

s' = N x (3600/g) (17)

1 1
Ifs' < s for the curb lane, s' should be used in fur-
ther calculations. In this case the curb lane and the
nearby lane are considered one Lane with a saturation
flow equal to the sum of the individual saturation flows
for each lane.
 17

The method can also be applied with minor modifica- Figure 22. Number of 10 . z-u.1~ 00 · ...,-.
tions to lanes for left turners. vehicles per cycle that can
9
... . ,. ·n .
be discharged from a curb ':" ... ' . I
I
lane for right turns with
••
..
z ;' • .: I

.....,
8
Signal Timing- reduced length. . ·~ . I , '.I
/ .
The signal timing method the manual uses was developed "
;.,
u •.· 1- · .· / J I-
I ,, . , .... e
.,...
. .'' ii
by Webster (19) and is based on minimum average delay. 6 '
'"· .'.. I ·1
An important factor in this method is the estimation of
the loss of effective green time per cycle, F. For this
.
0.

.....,
-~.· .. : 1 7
-~ l.J II
- I.·

..... ' ~·
..,
.
purpose the manual includes methods for determining ....u 1 I/ / tt .
4 ..
safe clearance times between phases set according to ~
... ,, t.7
~
71/ . ...
..... I
rules established by the Swedish National Traffic Safety
Administration.
Traffic signals in Sweden display green plus amber
z0
0

;
r!. 7,
(./
// ,
I
·~
~...,-

--1
...
- u ., a
at the end of green. Field measurements have shown :... ., .. - -~
that this interval is fully utilized by drivers and that it ' .. , I.
has therefore been included in the effective green in 0.0 0.2 0.4 o.s o.a 1 . 0
the manual. An inter-green interval thus consists only Ratio of right turns
pr
Table 3. Saturation flow of
Green Time (s)
lane with turning vehicles in
conflict with pedestrian 10 15 20
movements. Two-Way
Pedestrian Length of Crosswalk (m) 230
Flow
Turne per Hour 7 14 21 7 14 21 7 214 All Widths

0 250 1700 1700 1700 1700 1700 1700 1700 1700 1700

21500
0.05 250 1580 1650 1640 1650 1670 1650 1660 1670 1680
500 1500 1620 1600 1610 1660 1640 1640 1670 1670
750 1390 1570 1580 1560 1570 1580 1600 1600 1630
1000 1330 1490 1560 1510 1490 1510 1550 1540 1600
21500 1290 1490 1560 1380 1120 1490 1450 1500 1540
0.10 250 1530 1620 1600 1610 1650 1640 1630 1660 1670
500 1400 1580 1530 1540 1620 1600 1590 1640 1650
750 1260 1480 1500 1460 1500 1490 1520 1550 1580
1000 1150 1380 1460 1380 1370 1390 1450 1450 1520
21500 1090 1380 1460 1170 1280 1350 1280 1350 1420
0.15 250 1430 1580 1560 1560 1630 1600 1590 1630 1650
500 1270 1510 1460 1470 1580 1540 1540 1610 1620
750 1070 1390 1400 1360 1410 1400 1440 1470 1530
1000 960 1230 1360 1250 1230 1270 1350 1340 1450
21500 890 1230 1360 1010 1420 1220 1130 1200 1300
0.20 250 1390 1560 1520 1520 1620 1580 1560 1610 1640
500 1200 1480 1400 1410 1550 1510 1490 1590 1600
750 1010 1320 1340 1290 1350 1340 1380 1430 1490
1000 870 1160 1280 1160 1140 1180 1270 1260 1380
21500 800 1160 1280 890 1030 1120 1020 1110 1210
0.30 250 1300 1500 1450 1440 1580 1530 1510 1580 1610
500 1040 1390 1280 1300 1490 1420 1410 1540 1560
750 850 1200 1210 1150 1240 1220 1270 1330 1410
1000 700 1000 1140 1000 980 1020 1130 1120 1260
21500 630 1000 1140 720 860 960 850 950 1060
0.40 250 1230 1460 1390 1380 1550 1480 1450 1550 1580
500 930 1320 1190 1200 1430 1350 1330 1490 1510
750 720 1100 1110 1030 1150 1130 1170 1250 1330
1000 590 880 1030 880 860 900 1010 1000 1160
:>1500 520 880 1030 600 740 840 730 830 940
0.50 250 1180 1410 1340 1330 1510 1450 1420 1520 1550
500 830 1240 1100 1120 1380 1290 1260 1440 1470
750 640 1000 1020 940 1070 1050 1090 1160 1270
1000 500 790 930 780 770 810 920 910 1070
21500 450 790 930 520 650 740 640 730 850

0.60 250 1130 1370 1300 1280 1480 1410 1380 1490 1520
500 750 1180 1030 1050 1320 1230 1200 1400 1440
750 570 940 950 870 990 980 1020 1120 1210
1000 440 710 860 710 690 730 840 830 1000
21500 390 710 860 450 570 670 570 680 770
0.80 250 1040 1300 1110 1200 1410 1350 1300 1450 1480
500 640 1070 910 930 1230 1120 1090 1320 1360
750 470 830 820 750 910 870 900 1020 1120
1000 360 600 730 590 580 820 720 710 880
:.1500 310 600 730 380 470 550 460 550 650
1.0 250 950 1240 1170 1120 1360 1290 1240 1410 1420
500 550 980 820 840 1150 1030 1000 1250 1300
750 390 740 730 660 830 780 800 930 1040
1000 300 510 640 510 490 530 630 620 780
"1600 260 510 640 300 400 470 390 470 570
18

of all red and all red plus amber between green in two distribution according to Equation 18.
conflicting phases. The manual also includes recommendations for tim-
The total loss of effective green time per cycle, F, ing of traffic -actuated controls (minimum and maximum
is obtained as the sum of the inter-green intervals be- green, extension intervals) based on previous research
tween the conflicting phases. (21) .
The cycle time, c, is obtained as
Capacity , Que ue Length, Proportion of
{18) stopped Vehicles, and Delay

where All the measures dealt with under this heading are first
calculated on a per-lane basis. Totals and averages
F sum of inter-green intervals per cycle, for an approach or for the whole intersection can also
and be derived according to instructions in the manual, but
!l(q/ s i) ma x sum of q/ s for the critical lanes, i.e., they are not always meaningful.
the lane s that make up the critical con- When the signal timing is established, lane capaci-
flict point in the intersection. ties,Ki1in vehicles per hour are derived as

The distribution of green times that best minimizes K; = g;fc x s1 ( 20)

delay is derived from
The degree of s aturation, q / Ki, becomes highesl and
(19) equal for all the critical lane s. If it exceeds 1.0,
queues build up infinitely, and even at values above 0.8
The green times are checked against minimum green substantial queueing occurs. The average number of
time requirements, and the timing is adjusted if neces- queueing vehicles in lane i at the beginning of green, Ni,
sary. This is done by adding the extra green time re- is calculated according to Miller (22 ) :
quired to F and calculating a new cycle time that is dis-
tributed among the phases without disrupting the optimum (21)

Figure 23. Form A plan sketch, traffic

volumes. and assumptions.
~·

1q;r6

.,~,,
,4,P,Pro«,, :' HN ~ ~riTi -
I
-
I \ r~
.- ----
-,
I.
.... H .. l'l-w
7'r"'~Jfr1'r.,, ../law ~l"h .._______
I I . i20
1:20 640 110
-X?m .'". 1 :L..., I k
~ 2'70 I

!
"
j
J J
" 1! I~
I .
eoo [
1' y.-- 380
i
.s J"1. I , I
700
~- ---')

v
" A-r?_. 01ul-,.: ~:E. I
t
I g- ..+rc:c:-C _,, )
' ~~
..
-.
- ~~~
J r

l
~

,,, ..,o--i
I _;T.O_~
12-"+-
J .O
I
...... ,,
-
-

__~-4.
" 117

- ~~
:>

l'
c

·-
"
-
1
....
AJ'; o'•• ch : "P-..W ..
I
- 8 00
~---7

1' I
I
qo
1300 I
I
w
I

~ t' !Mm
1i r
____:!.
' ·: I
130 ---->
(It
~l II "1"" j 50 BIO 320

Con frolfrr ;
;=;)( e lf 1- hri e
110 ~
~- lli.&
-~-~ '- s..,::~;·•eh : +-15

A?pr-oo..ch .S,P ~ lfwl/ Truck, . ....

.fl.rd'-'5,1.,
I Zicy d <S I Slope
i;i.i
/'1,'n, ~,,

""°" "a+'*
,.,,,, , /,,,~ .
?~~~:,
(~) • "f1·-c/Mn (h
bv:tes
'"
('~"

HS
I

50
I
"lo

10
I . -
5

OJ
. 7.o
•
R.E 50
50
10 -
-3
9 Zo
._ /jN 10
~- 7. 0
- RW 50 5 - q 7. 0

--1
 19

Figure 24. Form C,

calculation of signal
timing, first round.

Figure 25. Form C,

calculation of signal
timing and capacity,
second round.

~-' 3 D!f_ o.oo,~ - --=.-::._ - 1420 _.55.?. _ J Izq ~~o ~~

1_,__2__, __3_ .~ q~ 1700 '--_...=_lqH,__ !26z : 630 . i~ o,556 .z~~2:~l-- 1'l, -no@
1
,__
•1 3 3 E I o 1040 - 0,14 - 350 2~ :
1-R_E-i--4
-11-lt-_2-l-''=%G,£_~1 ~ .~~~ -~ - - - 860 270 ~ 0442. 23 23 3°30/~
I• ' 5 I+
2 DI
/F B OzG
t22o/
1'"'50 .\'61: -
I
i- - - /300 400 .~ 23
4"'0
-/
r~
o~
G I +2 £ C l,o i~~b
13
- T- - - 910 l:ZO 0.J3Z 2.~ 340 O;as
HN 7 3 I DI
1-----1--1-.-i--:.+E 039 ~ - 1-
i
/,oz - ")IO 200 2'1 _430
·-~ 1
067
_ ,__,_
I e3 A O,o 1700 - : - I.oz - - J730) ~Q 0,300 2C) ~ ae0--- 4),·(.6
q 3 £ Io 510 - - I oz - 120"' 29
RW 10 I p;:A~ l,o ".%c l, o~ l,i:>3 - - TIO CJO 0,11• 14 _1 80 0,=
-t---'-~.....:...A
....:....-1-0
::.'"'-o.+-l-7'
_0_0-i...:J,_03 I, 03 - - /800 /30 14 410 Q32
1--"--11--
I :._
/ - - !.-
12 I ;= l.o 780 1,o.<; / 03 - 840 /10
..___.
qss
l---+---1---1------1--'--+--r---1
1
~
L____j'--.--l---l.-~-L----'----1--J_-J_._Jc__----'--- · · --
I I
20

Figure 26. Form D,

calculation of
queue length,
proportion of
stopped vehicles,
and delay.

1--~-+~-1---1-~-1-~-1----,, -+~-+-~+---+--+--~~~~-+~-1------1~-;-~~-~~-----~-~

l--~-+--l-~-+~-+-~~1--~-t--l-1-~--t--·~t-----1·~-+-~·~t--~-+~;-~~~~---l~-~-~--- ~

- o.e:a7[__,_1__1_1- 11-t___.._I~=--.__--_ £-o~-~--

j-=---1---t-_l___--~tJ~j

where APPLICATION

Nu= (2x 1 - l)/[2(1- x,)] x1 =(<Ji x c)/(s1 x Si) > \.4 In the manual (4) the chapter on signalized intersections
Nii= 0 X; "' Yi
a
occupies about fourth of the whole book. The im-
portant relations are all presented in graphs or tables
Ni; = q 1 x (c - gi) (22) in order to simplify use of the manual. Computer pro-
grams are also being developed for the same purpose.
Ni represents the number of vehicles remaing from The layout of the manual is in the form of stepwise
the previous green phase and N2 the number of vehicles calculations (Figure 10), each step leading to a set of
arriving during red. N is then adjusted upward by figures to be filled into the corresponding column on
means of Poisson's curves for required probability of the form for the calculation. Four different forms are
overload. provided for signalized intersections:
Queue length, L 1, is derived by multiplying the
number of queueing vehicles by the average space per 1. Plan sketch: traffic volumes and assumptions,
vehicle . The manual also includes adjustments of Li 2. Determination of inter-green,
as a function of the proportion of trucks and buses in 3. Calculation of signal timing and capacity, and
the lane. 4. Calculation of queue length, proportion of stopped
The proportion of stopped vehicles, p,1 , in lane i is vehicles, and delay.
derived from
Figures 23-26 illustrate the use of the forms for an
P,i = (r1 + N, x t;)/c (23) intersection in the downtown area of Stockholm. Fig-
ures 23 and 24 represent the first and the second
where N 1 x ti represents the time from beginning of rounds of the signal timing that are filled into form C.
green until the Nith vehicle in the queue has started The time required to carry out a complete applica-
to move. Measurements indicate that ti normally tion, such as the one in the example, may vary between
equals 1.0. 3 and 6 h, depending on the skill and practice of the
Average delay, d" is calculated according to Web- engineer. The most time-consuming part is the
ster (~: second round of signal timing, which, however, is only
necessary if queue lengths or other measures for in-
d1 = 0.9 [c(l - Xi)/2(1 - X1 x x1)] + {x?/(2q, x (1 - xi)]} (24) dividual lanes are requested. If only signal timing and
degree of saturation are asked for, the calculations
where >.. 1 equals g/c. For xi, see Equation 22 above. should not take more than an hour.
Adjustment factors for queue length and delay in traffic- The manual has been introduced in a number of work-
actuated controls based on the corresponding values at shops for traffic engineers in Sweden and has been
fixed time controls are also given in the manual. greatly appreciated for its exactness and flexibility.
A number of important questions still have to be solved
by future research, however, for example, the effect
of bicycles in the vehicle flow, the influence of different
 21

types of traffic-actuated controls, and the influence of with color for emphasis and the use of examples and
rain, snow, and darkness. forms to assist the analyst. I am generally favorably
A thorough validation of the methods is also desir- impressed by the Swedish manual and Mr. Peterson's
able and may take place after the manual has been used paper. It is, indeed, a forward step in the capacity
for some time. field.
REFERENCES I do take some minor exception to Mr. Peterson's
statement that "it has been clearly shown" that actual
19. F. Webster. Traffic Signal Settings. Her Majesty's capacity is higher than that arrived at by different
stationery Office, London, Road Research Tech- methods. In my discussion of the other two papers, I
nical Paper 39, 1958. shall indicate why I differ. Also, I believe that pro-
20. L. D. Gordon and A. J. Miller. Right Turn Move- viding design guides or levels such as levels of ser-
ments at Signalized Intersections. Proc., 3rd vice is an advantage because it sets standards. "The
Annual Conference of the Australian Road Research choice of that degree of accuracy which is right both
Board, Vol. 3, No. 1, 1966. socially and economically" can be a controversial
21. K. -L. Bang. Optimal Control of Isolated Traffic choice if it is not based on engineering fact.
Signals. TRB, Transportation Research Record
597, 1976, pp. 33-35. PART 2 BY HANSSON
22. A. J. Miller. Settings for Fixed-Cycle Traffic
Signals. Operational Research Quarterly, Vol. Mr. Hansson's paper presents a method of calculating
14, No. 4, 1963. capacity, queue length, and delay at intersections con-
trolled by yield or stop signs. The method is based on
Discussion a queuing model (M/G/l) whose most important input
parameter is the critical headway determined from
Robert C. Blumenthal, Robert C. Blumenthal Associates, field observations at 18 intersections in Sweden (Table
Boston 2 in the paper).
Data gathered for the 10 yield and 8 stop approaches
PART 1 BY PETERSON provided critical headway (right, straight, and left)
and move-up times. The paper notes that "values ...
Mr. Peterson initially reviews the historical background are based on these measurements for Swedish condi-
leading to development of the 1977 publication on calcula- tions, as well as on previous studies elsewhere."
tion of capacity, queue length, and delay on road traffic Table 1 outlines the values.
facilities. He notes that "it has been clearly shown I have reviewed the information in Tables 1 and 2
from measurements in several countries that actual and cannot find any direct relations. Table 1 values
capacity is higher than that arrived at by application are the basis for calculating intersection operations,
of the methods" in referring to previous methods. so it would be interesting to know how they are derived.
No design standards, such as levels of service, are A breakdown of Table 2 reveals 12 locations (8 yield
provided; rather, the final decision as to the design is and 4 stop) where speeds were 50 km/h, 3 (1 yield
suggested to depend on factors of road safety, cost, and and 2 stop) locations at 70 km/ h, and 3 locations (1
environment. Thus, Mr. Peterson states that "the yield and 2 stop) at 90 km/h. I wonder what level of
choice (or design) should be based on that degree of statistical confidence can be placed on values derived
accuracy which is right both socially and economically." from one or two locations.
Major additions to the literature are calculations I suspect the additional data came from previous
for unsignalized intersections, pedestrians (midblock), studies. In 1974, the Organization for Economic Co-
bicycle traffic (bike paths), and short weaving sections operation and Development (OECD) published Capacity
of 40-60 m. of At-Grade Junctions (7). Table 4 of that publication
The manual itself is a fine publication, well laid out (here Table 4) summarizes critical time gap values.

Table 4. Summary of values for the

Through Right Turning Left Turning
critical time gap found by various
researchers or used in the national Passenger Passengers Passengers
design manuals. Source Cars Trucks Cars Trucks Cars Trucks

Grabe, Worner 4. 1-5.2 6.0-6.5

Krell 6.4 (second vehicle: 1.82 t; third 2.53 t; etc.)
Ashworth (median value) 6. 5
Ho!wegen (valu e at a major
road flow o r 1000 v /h) 5.4 5.0 6.0
Kell (median value) 5. 8 5.4 6.3
Owens (median value) 6.3
Thomasson (median value) 8. 6
Knoflacher 6
French design standards' 6
German guidelines for traffic
signals
v = 90 km/h 7-8 6-7 8-9
v = 50 km/h 6-7 5-6 6.5-7
English design standards 4-8 8-12
Dutch design standards 6 (7) 6 (9)
(+2 s
reaction
time)

Swedish manual
v = 90 km/h 7.0 7.2 7.5
v = 70 km/h 6-6.5 6-6.5 6.2-6.8
v = 50 km/h 5.2-5.8 4.8-5.5 5.3-6.0

• In some cases the time needed to complete the maneuver for one carriageway with two lanes is given.
b Sight distances are calculated taking into account the time required for merging.
22

I have added the Swedish manual values to the bottom chosen by Hansson, that the results are a perfect
of this table, and find they generally compare well. match. Sites 1 and 2 have lower critical gap times,
(Note, however, that the values in Table 4 are simplistic. resulting in higher capacity for the right-turn move-
The original texts generally classify time gaps by ment.
specific configurations.) The next figw·e (Figure 11 from the OECD, here
There are no sample calculations in the paper. How- given as Figure 29) compares capacities predicted by
ever, there are two examples in the manual (attached). five methods, with critical gaps of 5 and 7 s for a
One is a T-jwiction, where both roads have speeds of straight movement in conflict with two-way primary
50 km/h (Figure 27). Traffic flow from the two-lane, movements. As may be seen, there is a great deal of
one-way secondary road is 200 vph turning left, and variance in predicted capacities. The OECD review
200 vph turning right. The two-lane, two-way primary concludes with the remark, "The foregoing review of
road carries 500 vph each way. Capacity is calculated methods of estimating the capacity of major/ minor
as 570 vph for thP. right turn ann :mo vph for the left priority junctionlil ha.& lilhown that the topic is a. very
turn. complex one, and that despite the considerable amount
For comparison, I have another OECD exhibit (Fig- of research which has already been completed, a great
ure 12 from page 35, attached here as Figure 28) . This deal still remains unknown."
figure shows three locations in Britain where com- Mr. Hansson's paper and procedures add a method for
parisons were made between observed flows and those calculating unsignalized intersections that appears to be
predicted using Tanner's formula (.!Q, .!.!)· I note in simpler than others used in some areas.
site 3, where the critical gap time is identical to that I would like to see an example or two in the paper so

Figure 27. Example 1 from the

Swedish capacity manual.
KO l'. S•.; 1!,C, tJf J. U !~t.. J " !' 1:•,.!.L[q ,,,.,-, 16 . /Z. CJ6
t;t,.., ; ~-~!:!.L l:tfr. ,, u : T.LJ.

Fl~ :t-11
tJ ?aJ
(1 1 n )
Yo C=Sa:;

I
'" ZC07
71d: (!;hf. n.t.7.f"' I
\ lg"~ZI -0._ - IJ

I v \> <:J--~ Sa?

I .q= CJ
I

_
A !;b
_,__.. I ~ I~ ~C> - - tc
I

rs
Orll!ltUlf•r,.r - ber!!"'lng P.!i:erfo~sf<;t'll: stcn lhl Q{J v~jr.lnc.;;f.silikt 0
:~!:~:.:~::;t!:)~~~;~)~~~: ::(?ri..,,rvS~I
- ,~. u.t , !f,.n.U f:lti:r tl1Uul\~ N"'.4

- tn:a.~• 11•frrfH11tnt;•r

'l"lor4rvC5tM ~redd (r:a)

.. ~C?. . {5: 1r:..~~f'llg)
Tlt?rtsstorl•k
r.oro;r.i"lj'.f.ur•.i \::i) (E,.:11 in,): ...4."'l?: .. thn~st.1;:10
Liiif f'tlrtii.er;!r.51nt!lle

1--,---~-~----lfl tl de
tl vit r-
or~llt
:< ritis~1 ~tU­
lids-
h'i:ll'od
r,;r~ :;-
lld
:. t) t:- S£k..
' nt- ever-
r ~ !':':jS· ordni·~~
lt~rl"i-
~~r11
~o!iLt--
i;e l-
till-
ftrt.r;n s
~t rrl -
9.;;r..::d
bth\t·
l'.t;l!CI•
tet
r.-.dfl d
. ••

..
s-r•:I for- Int- :itlJ::. t- M~!j'-·
Ji 11-
f.?r"t
l"' lt1
hrt
~;. k"'~•l
~. O r!JH
icr".::onsstrti1J
'•
(f/ h) ;; r ,,..1 d-
~ons-
1trtirr.-
n l "I•,'§:•
!j'rl.1
nln91 •
1r1d
511d

(f/h) (r)
',<•>·/
(e ) (s.!
'"
(<) (s!
v.i rOt n h t ,,,,,, ~ - t
S ~ 't . ~ frC~ ~Orta
ri\f• '1 k&r-
ntl\g (f/h) t.r~. t:ia
st ~;i

MoJ.:11 7.2 7.l 7.4 7.S 7. 6 1.1

'·' 1.9 7.11 7.ll i.ll

h
A /"
v
O'h I h Z().J J~t' J.7 6..J 0..JS
tJ ,,.
o.,, 1 I/ la'J I® 6.o /.2.o 0.67 tJ.67 300
;,
c r
v
h
/) ~- - - /"
//

-' ''' '.I.--

• - ~ --''----'"-
-
--'---''-----''--- ' -"-,__"_.___•·-·__.__:i_1.....;.:"'--'--·~_j I~ :> I ;~
- --~~-~-~ 11
 23

the method could be used by traffic engineers. Thus, PART 3 BY BANG

comparisons could be made. At present, the paper pro-
vides a great deal of background, which has value but Mr. Bang has presented a well-written paper outlining
does not help the reader make practical use of what is the Swedish capacity manual method for calculating,
obviously a well-developed method. I hope the Swedish not only capacity, but also queue length, delay, ap-
manual can be translated to fill the missing elements. proach grades effect, proportion of stopped vehicles,
and signal timing. The method makes use of previous
findings by Webster of Great Britain on signal timing
and Miller and Gordon of Australia on conflicts between
Figure 28. Comparison of predicted and observed traffic left turns and opposing flow and signal timing, as well
flows (l). as previous Swedish work, especially by Arne Hansson
15CO
on conflicts between vehicles and pedestrians.
:;:
Original procedures are shown for calculating

',,, T~·
SITE I
~
':_ IOOO
.,. 0 Obntv<id ~.J/e 1. Effect of conflicts between left turns and opposing
The-oreric.,/ C:!.rve

g traffic on two-way streets where (a) a waiting left turn

... ...... will block the lane and (b) a waiting left turn will not
I
'....,i.t1nor road

5CO block the lane;

''....,.o o 0 2. Reductions in the capacity of right turns by
" -~ -..q,
-- 0 0

---- 0
0
pedestrians in the crosswalk; and
3. Effects of restricted length of curb lanes where
(a) through and right-turning traffic are in the lane and
1500 (b) the curb lane serves only right turns.
:;:
The paper contains an example of the procedures for
~I~
SITE 2
0 Obsernd d.!ta a four-way intersection, in downtown Stockholm, with
Th11or11llcal c:ir11e
three-phase signal control. There are three examples
! .... 5..2s 13 1 -1.91. fJi ~ 2.7s

I
:;
500
in the manual: two-phase, three-phase-the one in this
paper (Figures 30 and 31)-and four-phase.
For comparison, I calculated capacities for two of
the three intersections using the HCM and then com-
pared the results (Tables 5 and 6). Tables 5 and 6
Ua/ar-roart flew O,. (~·e.'1 I h} show the comparisons. My calculations were rough,
11C11 but are adequate for comparative purposes. Overall
:;:
intersection capacity appears to correlate fairly well.
SITE 3 Location A, M-Stad, shows an overall variance of 6
11000
~
O ObSlttvect ddt•
Th~tarlc•I CIJfllfl
percent, or 220 vehicles. In this example, the HCM
' ...... would predict more overall capacity than the Swedish
!
I 500
.....o-...
0 °""°-ig.q. ~
O"'"O ..... O._ -o -o
0 ___ _
method. Location B, Stockholm, shows an overall
variance of 4 percent. In this instance, the Swedish
manual predicts more capacity than the HCM.
G''-~~~~.......L~~~~~-'-~~~~--''--~-
D 1111'..0 1500
--- The individual approaches do not compare as well,
with variances of 2-50 percent between the methods.
One of the major reasons for variations on the high
end is signal timing. Table 7 reviews the level of ser-

Figure 29. Comparison of capacities (1).

14CO

llCO

I • crWcal ~'31' (ucands} leCOldl"'J '" NttthMIJJ1'd'I melltods

--01calc'Jlari0tts
o • sland.Jrd d•11farJon
Ol!Jd;:J di•lrfb!JliOll
(•ssurr'r'Jnorr.i.J/J
tncondJ}

--- JJCC01f.!ir.7 ,~ Tan ..,,,,.\ (Errglitt>J m./hod

a.s tr.odif.~ tiy Ashworth (2a}

•r:cOHdf(:!] to rt.. Get'"iin guid•/Jndl

- - - - I~ traflic Slilf/JIS (fJ}

JCO

2ol0

100

1000 1!00 2000

Total m:tjOT-r!J3rl lloN (OJ (11B,,/h)
24

Table 5. Capacity comparison for location A:

Volume
four-way two·phase intersection with all two-way Width HMC
streets. Approach (m) L s R Level E Capacity" Difference

BN 5.5 100 200 200 750 590 -160 (-0.21)

BS 5.5 50 150 150 940 470 -470 (-0.50)
AV 6. 5 110 760 150 1000 1200 +200 (0.20)
AO 6. 5 80 850 170 1060 1270 +210 (0.20)
Total 3750 3530 -220 (-6)

"Pedestrian volume: N-S - 600-700, E-W • 200-150.

Table 6. Capacity comparison for location B: Volume

four·way three•ph11se intersection with 1111 two-way Width HMC
streets. Approach (m) L s R Level E Capacity" Dillerence

HN 9.6 110 640 120 1340 1310 -30 (-0.02)

HS 8.0 50 810 320 1260 1340 +80 (0.06)
RV 9.0 90 130 110 700 780 +80 (0.06)
RO 9.0 380 290 120 i090 1150 +66 (D.11)
Total 4390 4580 +190 (+4)
'Podeltrlan volume: N·S • 1300-600, E-W = 700-600.

Figure 30. Location A from the Swedish capacity SIGMALREGLERADE KORSNINGAR ,....,., i1 - ~ta.d .
manual.
"~·•·• A vt:i en - B ~a.ta.n
PLANSKISS - FORUTSATTNINGAR
'""''" Nils Nil=on
,.,_ 76 -11- 0_:8
= - - - - - - --l
rhdir 11111;
1,; "1"t76
EM ma.xtlmme. T:> r • h '"'~' \ •cnr. 10:;-~·· •~r-a.u;,r""'-·•~. H "'"' r ' ·'" " •
APr-"~o>.,l.J,
"
Ti4:

llLLrA.!::.T: ~N
! ;:
~~1~~~:~~ ~ ::~ ~~~ !!;~!.::;::;,.' ~ ;;.~ ;~ ~ ~ f ;:~~:t •·:, ;
"'---- 170
200 200 100

I
-(---- 850

Jl l 'I'
1700 , . - -
80

Um [!_ +
~2gQ~ ,t.ilr~ull.tH
6-S.
::::;re TIL.L"FA~T ' A8
;'f - p~
I
!- - - - - - ' .., +'- ~
- I ;}.f
- I

·-
3. -;-- J.b
I .!l::> ~

i - - J .t>
~s ~
~- 7.
6.
- - - - - -;
I '\ ~ /' I

TILLrAe.T : AV
1'
~£. 1 /6m .~/SJQ~

5001
JI 0 ______:/' -v
7GO---? <~
'§
l i r
50 150 150
~
Slj,. ~.Hrlt :

T111 r.,t
150~
Tp •• n1m111g: TIDSST'Vlle!t-..IUJ.C:.

ll.IUldltUlr:M•
"/<>
M.stl ll.flJ ro,.•-tn
!\)_
- -·
li.t.'\1111
ScAl.v"
Tl\...l..FAe.T : ~s
"-- \htn t:IOl'JO

j'{j~ja;",,J".(.J ,.,....,_
lll•·r~i'lll .......... ~~ .... - .. ~
I
,,\.1tr1n1t '·''°"
/\ ,?l':!a'- l<l ,,..
S~t:l,'•>U« tf
, ~.
V 1fO

>12u'"$ ~iC~Gli:-.,.
l•I
•11?.>0·ftf·
f,/j,. ~ ... ~.
ld
' " fl:.f'
Is.)

- ,,Jit~ . SLe;>/! Pc1l. Gi:l:i\-<" ~~ -V.:Llll.

, • •
. .. h
'
B.S 50
I

10 .JA 0 .5
'
a_o I
A ·o 50._ ~~- JA + :;; '7.S t> 0
"BN _ 50_ _ ZQ_ __;:ii·." _ f?,_5 Ro
:2;"v- - "50 5 :JA -3- 75 5_o
·- -
I
-
- -- - - -·-- ,___ :
i
--- -
-- - -- - -· -- ,_J

........ l ...
 25

Table 7. Level of service comparison for locations

HCM
A, B,andC. Degree of
Location Approach w. Volume G/ C LOS "E" V/ C LOS Saturation

A BN 18 500 0.49 750 0.53 C/ D 0. 86

BS 18 350 940 0.37 c 0.74
AV 21.3 1020 0.40 1000 1.02 E 0. 86
AO 21.3 1100 1060 1.04 E 0.87
Total 2870 3750 0.76 D
8 HN 31.5 870 0.36 1340 0.65 D 0. 67
HS 26 .2 1180 1260 0.93 E 0.82
RV 29 .5 330 0.18 700 0.47 c 0. 58
RO 29 . 5 790 0.29 1090 .Q,1! D 0. 82
Total 3170 4390 0.72 D
c MN 21.3 550 530 1.04 E
MS 14.8 280 380 0.74 D
IY 19.7 1080 1100 0.98 E
IO 19.7 1270 1050 ill E
Total 3180 3060 1.04 E

Figure 31. Locat ion C from the Swedish capacity

manual.
SIG N ALP.EGLERAD C: KORSNINGAR
.. _, OSK.A~SHAU-
1 ~ ..:,11 .. a: INF'A~TSLE:DE.N -MS.DE:LV.

PLANSKISS - F6RUTSATTNINGAR
"'~"'lli•·~ x
lnf)N u:
Kl...:B
h l lol'I:
1'?1'G - II - 23
,.....,,,,., ,....... ',1 .. .-.:
-
Ar:
ti• : l"M
£"1
1q74 AMO
TIU...'l"A'e:T : MN r •- - •·c1t .. h •.r•r<ui,..,1r •rt.1
1 i,..rl r.i; I l:r",111-; (1-,.).
i!-o1r-;1r-;Hl~iltr.,
r.r ~. ~.. ~ ,~.11 1 ,:.,utr,
l ~rhlt

,._.,_..,.,ia..,o:t,.,
f:rt~1 ·
I h.l.A'uter C;( ~ r.f ~ I ·
l~rh l h'.tr«'.l ><r, t tJ ~li,.~tr ,
fc:t •• 1'1'"1t u ~rhr ;'l l
oei. tl•")lr&l!Ulc..:r::.
I L_._300
50 170 180 I ICJO
130 220 200
~ +--550
j l L
500
750
~--~
),,L
~
a. -12.
2soi
-400 I

"'
--
"'TILI-FA~r:
CJCJo
50
qo

16
1 mhds/eo'EW I

L
i
- - - - - ,(J, II.
~
Z. -=.. ti"==:I
~ ,,~
,,-- J.O s
- 2Jq·!';:R-::!...}
~
r
- J~:r __.=;=
-
9: /J. //. ~ - -- - -1i
J
~ 1....... 500
T1LL-FAl2:.T• IV ~~ 150
+--+
") 250 t
250 1400
180 ---1' ~

l j r
!.f;
,~
I 130---?
7.,0
70
:s
~
I.

TILL'F~T: MS
40 120 40
40 160 GO
~
110 ---:II
S~rf~ ;J•rlt:
Ttt ~1 tt1"l11j1: T~A'F'li:::.."&,...,...~ I NG . Shh: t :IDC.O

fllHart Mttlli;"'lts ~rOiM ""!IP1

tr1r1~
t.MO' f'tl,.\-tu
(/\1lt••fl- lw"'"'
l~J
tit•. trdll~I·
fi:lrsw ...»
ltl11. "l'J~hlp•
ti• nr ... ~-
....
l'!Vhl a 1.7.';• ..... l•I M

IQI:

MS
' SD
•
0,10
• .
.:J.A,,
-
•
7
•
3,5
'
r6 70 O. oe. .::JA~ q 7o a I V-v
MN
IV
50
70
010
O,oc;;.
:::JA
.::7r',. of<-
- q
7 5o
7o ... r
. -, ....-r=v e

-
----

..... ..
26

vice expected with the signal timing shown in the ex- superior sensitivity and consistency of application than
amples. Location A would have been timed differently, do the procedures in the HCM. For instance , Swedish
based on HCM analysis, and would have evened out the calculations are based on the analysis of the individual
capacity differences. Delay was not calculated for vehicle streams entering an intersection, rather than
location A. Therefore, there was no opportunity for aggregated intersection approaches. This enables the
comparison of that item. user to identify the critical lanes as well as the critical
I should note that neither capacity nor delay was approaches of the intersection. It also provides the
calculated for location C (Oskarshall), the four-phase, analyst with an overview of the total performance
actuated intersection. potential of the intersection, through the identification
I wish to congratulate Mr . Bang on a well-written of the degree of saturation.
paper and on his contribution of a methodoiogy for In currently practiced "critical lane analysis" pro-
calculating effects of conflicts between left-turning and cedures, the user must have access to lane volume data
oppoiing flowi, conflicti between turning vehicles and or make. assumptions re.garrling the. rliRtrihution of thP.
pedestrians, delay, and queue lengths. approach lane volumes, which can produce serious in-
Analysis by lane and movement should give the traffic consistencies in the results of the analysis. The Swedish
engineer an opportunity for specific insights into an in- manual provides a unique lane-volume distribution tech-
tersection's operations. It will obviously not be a nique whereby the lane flows are distributed in propor-
favored method for planners or those who seek a quick, tion to the saturation flow values of the adjacent lanes .
simplistic method. This technique allows all users to perform this calcula-
Although the comparison of approach capacities ap- tion in a consistent manner.
pears to indicate substantial variance between the The Swedish manual also provides a technique for
Swedish manual and the HCM, two factors in the ex- quantifying the effect of the conflicts between left-
amples could have significance in reducing the variance: turning vehicles and opposing flows and the conflicts be-
(a) signal timlng, as previously noted, and (b) use of tween turning vehicles and pedestrian movements on the
approach width versus lane. Please note the single lane performance capabilities of the intersection. This sim-
approaches at location A of 5.5 m. The HCM (and plified lane classification system enables the user to
others) treats this width as two lanes (although not identify the type and degree of vehicle and pedestrian
specifically), thus providing greater capacity. conflicts and to select corresponding values of satura-
tion flow for each lane type in a consistent manner.
!vfust ilitcrse.:tiu11 d.ii.d.l ysis tech.niqucs ai·c una.tlc tv
calibrate the utilization or capacity of a curb lane if
parking is not prohibited for a specified distance from
the intersection approach or exit. The Swedish manual
Jack A. Hutter, Traffic Institute, Northwestern provides procedures for calculating the reduction in
University, Evanston, Illinois saturation flow values for curb lanes of restricted
lengths for the two cases where the curb lane serves
Because of time and space limitations, my discussion right-turning and through vehicles or the curb lane
will be directed only to the paper pres ented by Mr. Bang. serves only right-turning vehicles.
This informative paper provides a detailed overview of The second object of this discussion is to raise ques-
an analysis procedure that offers some original and tions regarding various aspects of the Swedish intersec-
unique techniques for evaluating the performance cap- tion analysis procedures, questions brought about by my
abilities of signalized intersections. lack of practical experience in applying these new tech-
The primary object of this discussion paper is to niques to real-world situations, and give to the author an
highlight those attributes of the Swedish manual that opportunity to share his insights and experiences regard-
may be applied to the development of an improved U.s. ing the use and application of the manual by practicing
HCM. traffic engineers in Sweden.
In discussing Mr. Bang's paper, I shall carefully The lane~flow distribution technique is designed to
avoid the term "intersection capacity analysis," be- produce consistent results by all users. The question
cause the Swedish manual provides potential capabilities is whether this technique has demonstrated that it will
for evaluations beyond the determination of signalized simulate actual lane-volume distributions under a wide
intersection capacity. Additional analysis techniques range of geometric and traffic flow conditions . Also,
include it is stated that the method is based on theoretical
models supported by a limited amount of field survey.
1.Calculation of optimum signal timing, One must ask if the experiences of the practicing engi-
2.Detl!rmination of queue length, neers in Sweden in the application of the techniques sup-
3.Calculation of the proportion of stopped vehicles, ported the validity of the models.
4.Measures of delay, and There is an increasing need for American engineers
5. Determination of the degree of saturation, similar to have models that provide quantitative measures of
to a volume-to-capacity ratio that provides a measure of queue length, stopped vehicles, and vehicle delay for
the level of operation of the critical lanes or approaches use in environmental and fuel conservation studies.
of an intersection. Here the question is whether the techniques in the Swedish
manual provide for reasonably accurate and consistent
Another significant feature of the Swedish manual is simulation of these variables, insofar as previous re-
the important mathematical relations that are presented search would tend to indicate that flow distributions and
in graphic or tabular form for ease of use by the practic- traffic performance by lane are highly variable and
ing engineer. In addition, the layout of the descriptive erratic quantities when measured on a cycle-by-cycle
procedure is in the form of stepwise calculations, basis.
supplemented by clearly labeled work forms, which It was reported that calculations require from 1 to 6
assist the user in following a fairly detailed and com- h per intersection, depending on the number of measures
plex analysis procedure. desired and the skill and experience of the user. There -
There are several calculation routines in the Swedish fore, one must know the level of acceptance and use of
intersection analysis procedure that appear to have the procedure by practicing engineers in Sweden.
 27

Arthur A. Carter, Office of Traffic Operations, Federal There probably is no one answer to which is best; it
Highway Administration depends on the particular user's needs.
In this connection, I sympathize with Mr. Peterson
The authors should be commended for undertaking re- concerning his quandary regarding the right degree of
search into highway capacity in their own country, accuracy. Here in the United States, we have the same
rather than depending on earlier, increasingly obsolete problem. I suspect that the final answer one day may
American criteria, or criteria from other countries. be three entirely separate manuals, at different levels
Given the performance differences that we see just of precision and complexity, for planners, designers,
from locality to locality even within the United States, and operators.
I suspect that only concepts, not absolute values them-
selves, are freely transferable from country to country. PART 2 BY HANSSON
Having been technical editor of the American 1965
Highway Capacity Manual, I shall discuss in general Mr. Hansson's work on the performance of unsignalized
terms how these Swedish efforts relate to the American intersections is much needed. Relatively little has been
state of the art. done in this field, and that which has been done is mainly
on a case-history basis. I assume that four-way stops
PART 1 BY PETERSON are not included.
This chapter apparently uses the term "capacity" in
First, I was not surprised by Mr. Peterson's obser- its general sense to cover all levels of service, rather
vations that capacity flows in several countries appear than its specific saturation flow sense, because cau-
greater than indicated by U.S. criteria. Typically, I tions are given against its application where the minor
find that foreign flows exceed ours, possibly because flow is at a load factor (which, as they define it, is
of smaller vehicles and different driver behavior. similar to our volume-to-capacity ratio, rather than in
Therefore, I am not particularly concerned about find- the U.S. sense) exceeding 0.8 or 0.9. The observation
ings that disagree with either our 1950 or 1965 HCM, in the report that such conditions normally would war-
in terms of absolute numbers. rant traffic signals tends to substantiate the 1965 HCM
I am more interested in concepts, trends, and pro- statement that the true capacities of unsignalized inter-
cedures. In this regard, the literature review and sections are usually of academic interest only.
analysis of some 900 references, accomplished in 1973, The reference to M/ G/ 1 queueing system and M/ M/ 1
seems particularly impressive. It would appear to be queueing model may well be meaningless to many
superior, in both numbers and scope, to anything yet readers of this paper. It would have been helpful to
done in the United States. Translation into English describe these models in general terms.
might be recommended. I am somewhat concerned by the rather large num-
Although Mr. Peterson indicates that a complete ber of assumptions and deliberate omissions, in con-
manual has been developed, his emphasis appears to trast to the overall considerable detail involved in the
be on urban intersection operations, the subject of the method. We have no answers regarding the influence
two other reports presented here, with briefer mention of such factors as primary flow headway distributions,
of urban networks, bicycle flows, and pedestrian traffic. the combining of flows, move-up time, conflicts be-
He does not specifically cover the basic area of rural tween flows, relative sizes of flows, and load factor as
uninterrupted flow, either two-lane or multilane, or they define it, but I feel that the related assumptions
basic freeway flows. I assume that these fundamentals and omissions may have a significant effect in this
are included in the manual, presumably in the road unsignalized case. This points up the need for valida-
sections and street sections categories of the first tion, as suggested by the author.
table. I would be interested in learning more about Of greatest concern to me is the author's decision to
their handling. omit consideration of platooning produced by nearby
I am confused by Mr. Peterson's comments regard- traffic signals, on the basis that the effect, like that of
ing pedestrian crossings, which seem to stress the several other factors, cannot be quantified. In this
between-intersection case at the expense of the at- country, one of the principal factors governing the per-
intersection case. This appears to be in conflict with formance of a particular unsignalized intersection is the
Mr. Bang's paper, which emphasizes the at-intersection proximity of nearby signalized intersections and the
situation. [The authors of this paper revised the draft relative phasing of the nearest ones on either side of that
on whichthese discussions were based. Therefore, some under consideration. While I agree that the effect is
of the points in this discussion are no longer relevant.] nearly impossible to quantify, I submit that it is a
Mr. Peterson describes the use of models based on principal reason why some of the work in this country
queue theory for much of the work as an important in- has ended up quoting specific cases rather than general
novation, with only limited use of empirical evidence. criteria.
Although I am probably biased, I continue to feel that With all other conditions being identical, I see one
in this capacity field, practice frequently varies too specific location operating well, with periodic breaks
widely from theory for us to make empirical evidence in the major flow occurring simultaneously in both
secondary. I strongly support validation with empirical directions to permit free crossings, while another never
data. is free of at least one direction of the major flow. Pos-
The "description of consequences" concept appears sibly, adjacent signalized intersections are less com-
to be parallel to our level-of-service concept, but ex- mon in Sweden than in the United States, but I question
tended to better cover congested conditions, where de- that this factor could be overlooked in practice in the
lays, stoppages, and queueing occur. Its orientation United States. I agree, therefore, with the author's con-
differs somewhat from our procedures in that, in its clusion that primary flow headways may be the weakest
straightforward form, design and demand volume are link in the method.
both known and performance is unknown. Apparently, Also of great importance in this country is the nature
application to design problems or to volume determina- of the primary road-two-lane two-way, multilane
tions is a trial-and-error procedure. This is in con- undivided, or multilane divided. This element is con-
trast with our current procedures, which are most sidered in the method, and we are interested in their
directly oriented to flows obtainable with a given design. inability in the validation studies to confirm higher
28

critical headways for crossing four-lane roads than for utilized green intervals) is inadequate where most
two. Possibly, the influence of adjacent signals, more needed.
likely to be found on four-lane roads than on two, is It is not clear why the Swedish measures developed
involved. on a by-lane basis are "not always meaningful" when
expanded to a complete approach or whole intersection;
this concerns me.
PART 3 BY BANG The conclusion that both width and number of lanes
are significant is interesting. This point is contro-
Mr. Bing's work on signalized intersections is a valuable versial, not only within the United States but interna-
contribution to the state of the art, particularly with tionally, where viewpoints range from the 1965 HCM's
respect to its consideration of turning movements overall approach width criterion to Australia's basic
related to pedestrians. This is one area where the number-of-lanes proced\ll'e.
public expects and seems lo asswne lhal we have much Regarding the pede!!t.l'i:m eliecl, I Iind the tabulation
more information than we actually do. form of presentation quite good and a step toward what
I was surprised to note that the principal Swedish I would like to have, to answer citizens' questions. I
intersection capacity criteria, prior to the subject work, wonder, though, w·h ether such national criteria could
were based on the original American 1950 HCM rather ever be established for the United States as a whole,
than the 1965 edition, which is the principal reference given the widely differing degrees of respect shown each
in use today. Apparently this is because Sweden was other by drivers and pedestrians from one city to
in the forefront in the 1950's and developed its own another. It would seem that city factors would have to
procedures soon after ours were published. be developed, much as both we and o\ll' Swedish counter-
It would be useful to learn which evaluations were parts dislike them.
made in Sweden after the 1965 manual was issued that It is interesting to note that several of the same fac -
caused them to decide to start over rather than try to tors that have escaped easy solution in this country also
adapt its procedures. Such information would be par- defy solution in Sweden, including bicycles, actuation of
ticularly timely as we begin the initial steps of prepara- signals, and weather.
tion of a new edition of the American manual. The problem solutions described conform with Mr.
The Swedish procedure appears to emphasize a dif- P terson's introductory comments; they relate to situa-
ferent range of operational levels than does our manual. tions where the traffic volumes and geometrics are
While they indicate that their work omits design levels known, and the nature of the resulting operation is de-
and centers on saturation flow or capacity, which is our sired. It is not indicated whether or not the method can
level of service E, they suggest that 0.8 times satura- be used effectively for other situations, where either
tion flow be used in practice as a feasible maximum. volumes or geometrics are the unknown. (The sample
In effect, then, they closely relate to our level of ser- problems that exist in the manual U!ldoubtedly assist
vice D, which of necessity has become our design level greatly in providing au understanding of the proced\ll'es.)
for many urban applications even though level C is con- At first glance, the typical problem solutio.n time,
sidered more desirable. 3-6 h, looks long. However, given that this involves a
The Swedish procedures then appear to extend from detailed solution including queue lengths and so on,
our level D into level F, with a description of con- something much beyond our current procedures, and
sequences throughout including level F, as compared that a basic signal timing solution is possible inan hour,
to our levels of service A through E, with only general it is proQably reasonable.
reference to broad breakdown of level F. This is rather Finally, it is wrlortWlate that this work, like nearly
characteristic of foreign signalized intersection capacity all other work in this field in recent years, must be
investigations and procedures as compared to current tagged as "needing validation. " I hope users will soon
U.S. procedures; they typically concentrate on satura- do sufficient testing and evaluation so that validity can
tion flow to a greater extent than do U.S. procedures. be more firmly established.
I agree that any new American method must include
consideration of stops, queueing, and delay, which are
the elements of intersection performance most visible Publication of this paper sponsored by Committee on Highway Capacity
to users. Our load factor (that is, percentage of fully and Quality of Service.

Weighing Vehicles in Motion

A. T. Bergan, G. A. Sparks, and G. Dyck, College of Engineering,
University of Saskatchewan, Saskatoon

A scale for weighing vehicles in motion was developed at the University The relationship between vehicle and axle loads and the
of Saskatchewan. This scale has been successfully operated unmanned at structural requirements of bridges and roadways has
two locations in Saskatchewan for the past 2 years. An expanded evalua· been and continues to be an area of particular interest to
tion program is currently under way in which the scale will be installed
and evaluated in Ontario, Quebec, and New Brunswick by a project
those concerned with the provision and maintenance of
committee of the Roads and Transportation Association of Canada. bridge and roadway facilities. The need for comprehen-