ML20149E696

Applicant Exhibit A-15,consisting of Undated Article, Traflo:New Tool to Evaluate Transportation Sys Mgt Strategies, Published in Transportation Research Record 772

ML20149E696
Person / Time
Site:	Seabrook
Issue date:	11/05/1987
From:	Andrews B, Lieberman E AFFILIATION NOT ASSIGNED
To:
References
OL-A-015, OL-A-15, NUDOCS 8802110242
Download: ML20149E696 (7)

Text

e gs n ,

U W W 3// WE Tr naportation Research Record 772 U l

// g ,

RAPRESENTATIVE RESULTS ACKNJWLEDGMENT -

the TRAFLO Level II model was validated on a network The development of a model such as that described in -

12 downtown Washington, D.C., that consisted of 96 this paper represents the contributions o f, many links and 51 nodes and represented a wide range of people. In particular, we want to acknowledge the geometrics. Validation runs were made for morning contributions of Cuido Radelat and George Tiller of peak and off-peak periods, and a wide range of turn the Traffic Systems Division of the Federal Highway movements and traffic volumes was reflected. Each Administration, Barbara Andrews and Manf redo Devila tua was executed for 32 min as a sequence of eight of KLD Associates, Inc., William McShane of the 4-min time periods. Sperry Systems Management Poly *.echnic Institute of New York, and Fred Wagner

. M11da ted the model, reporting results on a of Wagner /MCee Associates. This work was performed lish-by-link basis for each of the eight time under a Contract with the U.S. Department of p;riods (p . The field measurement of networkwide scansportation.

cv: rage speed over the 32-min morning peak period en 9.71 miles /h compared with a model estimate of RETERENCES 10.29 miles /h. For the off-peak period, the model estimated an 8.79-mile /h average speed, which 1. E.B. Lieberman and others. Macroscopic comparco very favorably with an observed speed of Simulation for Urban Traffic Managementt The 8.73 miles /h. TPArto Model. Federal Highway Administration, U.S. Department of Transportation, vols. 1-7, PRCORAM ErrIC! mCY 1980.

2. D. I . Robe r t son . 'TRANSYT's A Traffic Network The Level II model was executed on a CDC 7600 Study Tool. Transport and Road Research corrpu t e r at the Brookhaven National La bor a tor y in Laboratory, Crowthorne, Berkshire, England, Upton, New York. Computer time for the model Rept. LR 253, 1969.

d: pends strongly on the site of the network. Buns of the validation network of 96 links indicate a ratio of simulated time to computer tiee of approximately 160:1 and a cost of less than $8 for a 32-cin simulation and ' fill

• time of 6 min. The computer memory requirement is reasonable. For IBM computers, lesa than 250 K bytee ate tequireds on Iwuainu of tom sm,xv transorcJ by (Dmouriro en ira @c Flow Theory CDC machines, less than 40 K words are needed. sst (hwiensiin c.3 e, _

cD 9 .

e% c7 1

TRAFLO: A New Tool to Evaluate Transportatipff a #g System Management Strategies if4 111 'M IJ. Q

f. B. LtE 3E RV AN AND B.1 ANDREWS The TR A F LO model. wh.ch combines tbe attnbutes of tref fic simulatmo the mobility of people and goods. The appilcation enth traffic enseenment, is desenbed. T R U LO was developed as a loof for of the transportation management process requires wie in transportenen # nning and traffic entneenne to test transoortation the ability to quantitatively assess alternative enanee. ment iteewpi Iv es a notte ste tystem. programmed in FORT R AN transportation management strategies to identify eet eenmts of fies comc >nent modeis thei nterf ace with ene another to those that best satisfy the stated objectives.

form an ente,eted sycem. Fou, of the madeis s,mytei, t,af fic operations.

The scope of the process involves both the end the fi fth st on eowilibnum tenfoc aiugnment rentet. The epereeing characteresters of the component semvist on modets are described. These tran e rtat W N ann W aM uaf f W ngineering models ers espaNo of simulatmg trafbe on one er more of the foHcmeng disciplines. The involvement of these two networ k s freenevt. corndort eist inc4ude the freemavhemptservete toad disciplines reflects the intrinalC dependence of comWen. weben and oubvebei ertenes s, end god networks representing th, behavioral responses (tr ip generation, distribution, centred bwuness essicts of urben centers. Also desenbed es the teeffic and assignment and smodal choice) on the performsnce g

owgnment component. e h ch cam be used 6n conrunction with the s mu. of the transportation system as expressed in terms

(

t letion components to determine the response of a trattat sviwm to e rat travel time, cost, ano accesaibility.

transportetsom manegement istotegy. It is clear that the need to develop effective transportation management strategies impiles a I requirement to develop analytic tools for that In recent years, events have shifted attention to purpose. Furthermore, these tools must be

( safe, effi-tent, and sufficiently broad in scope to meet the objectives l the need for providing econos.ical movement of people and goods on existing of the transportation management process for both highway facilities, rurthermore, there is a growing disciplines identified at>ove, u reness that factors suc h as air and noise One tool that is particularly effective for

! pollution and the conservation of energy must be evaluating transportation management strategies weighted heavily in any decision process involvinj applied *o a dynamic

. environment is traffic j ' the nation's transpor tattor6 system. ,

sisolation. Simulation molelo provaJe the eseans for g

'These considerations have led to the emergence of evaluating a wide spectrum of traffic management g

trLnsiertation management as the basis for improvin9 schemes within the framework of a controlled

' ~

kw# GSO211024P 071105 - - '

l PDR ADOCK 05000443 o een I' { # ' '%[- '

~

o O Transportation Research Record M2 s

10 J a

expe r ime nt . nie simulation approach is far more This tool provides the traf fic engineer with the .

[ i f appealing and practical than a strictly empirical information needed to explore candidate operational l

-2 solutions to resolve bottleneck conditions, expedite approach, for the following reasons mass transit operations, or satisfy other ' t'SM W >

obj ec tive s . He or she can apply the simulation

1. It is much less costly.

3 model repeatedly as an integral part of an iterative

2. Results are obtained in a fraction of th e design procedure.

J time required for field experimentation. '

Of course, any design improvement implemented by ]

f 3. The data generated by simulation include measures of effectiveness that cannot, in a the traffic engineer can ' feed back' and influence -

the results obtained from the planning process. It practical sersse, be obtained empirically. interdependence, noted earlier, that

- 4. Disruption of traf fic operations, whteh often is this ,

accompanies field experimentation, is ccept etely requires a strong interaction between the two disciplines. The TRATLO model, which is designed to  :

avoided. provide the capability described above, can act as a 6

5. Many transportation management strategies involve significant physical changes that are not primary mechanism for encouraging this strong __

I interaction and providing the inf ormation needed for i si-acceptable for esperimental pur W ees.

designing effective transportation management l 3

s t r a teg ie s . f g Most projects undertaken by transportation ' -"

planners address a time period that lies in the The TRAFLO model was conceived as the tool to fill this role by the Traffic Systems Division of l X f uture and thus require estimates of transportation demand. Based on these e stima te s, trip tables that the Federal Highway Mainistration (FHWA), in h g delineate traffic demands between origin and consultation with that agency's planning Division } *y'

[ and with personnel of the Urban Mass Transportation (

c destination tones within a region are developed. It Administration { UMT A) . This paper describes several 4; L is then necessary to identify a transportation

• management strategy that will satisfy the mobility, innovative vunvepts anu cesagn features ancur Wrated *

.'. in the a del, which has been implemented as a b enviror. mental. and economic objectives perceived for To do this, a traffic coeputer program.

g that future time period.

-- a s s ig nment model must be applied to estimate the {

d is t r i bu tion of traffic demand over a regional GENLPAL MODEL DESCRIPTION i g

network, consistant with the projected trip tables. _g-An assumption implied in every traf t te ansignment TPMID is a valuable tool in the transportation

{ mMel is that the traffic environment is steady management process. Its design includes features d-5 that permit the analyst to conduct a wide variety of  :-

state. That is, it is assumed that the specified a2 trip tables reflect constant traffic volumes on each studies on large roadway networks of general I

con! !gu r a tion. These networks may contain  ?!

[ network link and that any dynamic effects do not influence the assignment process. Furthermore, all components such as freeways, corridors that include j l ,

estimates of flow impedances included in the the freeway / ramp / service-road complex, urban and suburban arterials, and grid networks representing

- assignment co$els are also based on the assumption the central business district (CBD) of urban f-of steady-state conditions, and it is a s s t. sed that dynamic interactions between traffic on adjoining centers. The analyst has complete flexibility to  ;

c nfigure the netwurk according to his or her a

/ _

I network links may be disregarded. De efficacy of needs. The network may consist of any one or more (

f the results provided by a traffic assignment model depends on the degree of validity of these of the components 'sentioned above. j w

h underlying assueptions and on the accuracy of the Since the TPM LD model is actually a system composed of well-defined component models, the g

g I estimates of traffic impedance. Finally, the y

transportation planner has no means for verifying analyst can also select those component models that are most responsive to his or her needs. This -i the estimates of travel time on each network link flexibility enables the user to apply TPM10 in the >

that are provided by the traffic assignment model.

Simulation models, on the other hand, are most cost-effective ranner. d The simulation models that constitute the TPM14 -

f specifically designed to describe the dynamic effects of trattic flow. Factors that impede prcqram describe traffic flow macroscopically. Past expertence with other models has demonstrated that ]

traf fic are explicitly represented at a high degree

it would be Wasible to retain sufficient accuracy of detail. Consequently, simulation tools can T

provide a de ta n led description of the dynante fer evaluating transportation management strategies ]

if the less detailed macroscopic representation were  ;

h performance of traffic over a network. 1 The availability of an analytic tool that used. It has also been concluded that TPMto should combines the attributes of traffic samalation with proviue a hierarchy of macroscopic detail for h simulating traf fic on urban streets. That is, the l t those of traf fic assignment will greatly expand the user can select among three levels of simulation  ; 9 opportunity for the development of new and -

g detatis The more detail, the greater is the  ;

innovative transportation management concepts and designs. Trans prtation planners and engineers will accuracy obtained and the higher is the associated computing cost. This hierarchy would permit the n2 longer be restricted by the lack of a rechanism user to decide on the optimal trade-off between the 2 for fully testing these designe prior to field l de monst r a t ion .

accuracy required and the computet resources at his {

y Such a tool is of value to both transportation or her disposal. Pegardless of this selection  ;

planners and traffic e ng inee r s . It gives the process. TRATLD is far eiore economical in every l __

pierner the opportunity to examine the net result of respect than any of the existing microscopic models. i a design based oa t r a n sme t a tion system management The selection of an existing traffic assignment l OTM) principles. These results are expressed as mWe l for inclusion in TRAFLD was based on the idea  !

that the most satisfactory traffic assignment measures of effectiveness (MOEs), which describe -

traf fic operations on each network lins. With this models, from a mathematical viewpoint, are those  ;

[ d e t ailed inforeation, the planner can reexamine the that use Wardrop's principles of equilibrium q).

These models apply cptimisation theory to calculate =

estimates of travel time and accessibility that were involesi earlier in the tranapstation management the assigneent .of traffic over a network, taking .

iepedances. As  :

process (e.g.. when preparing the trip generation into consideration all link 4

expected, these models a:e computationally more data).

r.. '..- o , . . . , . . ., 3;- ,_ , , y . . , . ,q.

=

i ,

- .- . . . a

e

• I /h l O {

r Transportation Research Record 772 11 [

I comp 1;n, although computational coats are subnetworks, each of which may consist of several rZ sonable. The accurate estimation of link noncontiguous sections). The user must also specify 4 impedances is necessary if the potential of these ' interface nodes" at the juncture of the various models is to be fully realised. subnetworks.

Since the TPArLO software is designed to be machine independent, it was necessary to select a Urban i.evel 1 Mndel , ,

' model already co$ed in FORTRAN. Fortunately, such a  ;

model does exist Q) and, in fact, has been The Urban Level 1 model is the most detailed of the  !

car:f:11y validated (1). This model is similar to macroscopic simulation models. Since it treats each  ;

the equilibrium traffic assignment model currently vehicle in the traffic stream as a separate, [

embedded in the UROADS module of the Urban identifiable entity, the representation of traffic 1 Tr:nsportation Planning System (UTPS) package can be constdered microscupic. The treatment of the h dev31oped by UMTA. traffic stream, however, which is intermittent, or j event based, can be considered macroscopic. I MMC 2 PROGRAM FEATURES By treating each vehicle in the traffic stream f individually, it is possible to explicitly  !

Th3 TRAFiD program is actually a software system distinguish tetween different t y pe s of vehicles j thit consists of five functionally inoependent (automootles, trucks, and buses) and to treat each .

mod)1s. The logical structure of TRAFID is designed type according to its respective operating t3 permit these independent models to inter f ace with characteristics. Hence, the interaction of these fc one (nother so as to form a coherent, integrated vehicle types and the impact of lane channellration I system. Four of the models simulate traffic of bus-only or truct-only streets, and other (perations over a spect!!ed ne ork of roadways the detailed traffic management strategies, can be fifth model is an equilibrium traffic assignment studied in adequate detail. Furthermore, much of  ;

modil. the stochastic nature of the traffic-flow process i can be explicitly represented. {

Perr'senting the Traffic Invironment Each vehicle is processed (i.e., moved) as [

infrequently as possible. This f requency depends on in crder to use any of the simulation models in the the conditions encountered by the vehicle TRAT1D program, the user must specify the following teeediately downstream. The less impedance a f:atures 7f the physical traffic environments vehicle encounters, in the form of queues and no-go ,

signal indications, the fewer processing steps are t

1. The topology of the roadway systems required to move a vehicle a given distance. I

3. The geometrics of each roadway component: Associated with each vehicle is an ' activation

3. The channelisation of traffic on each roadway time" (AT), which is expressed in terms of the '

components simulatt'on clock t i e.e . When the simulation clock '

4. Motoriat behavior, which, in aggregate, time equals the vehicle's AT, the vehicle will be d:ttr ines the operational performance of vehicles Frocesseo (i.e., moved). The vehicle is generally in th3 systems moved to a point downstream--either on its current

5. Circulation pattern of traffic on the roa3way link or onto a receiving lint--and its new location, systems speed, and AT are calculated. This vehicle then

6. Traffic control devices and their operational remains ' dormant

• until the simulation clock time ch:stcteristics: advances to this new AT, whereupon the vehicle is

7. Volumes of traffic entering and leaving the again processed. (In contrast, a microscopic model roadray systems such as hETSIM (1) moves all vehicles every time

6. Traffic compositions and step and generates detailed trajectories.]

9. The configuration of the mass transit system, When a vehicle is processed, the determination of 1.0., bus routes, but stations, and frequency of its new location, speed, and AT (i.e., its status) s:rvice. depe nd s on conditions downstream of its starting point. A small tumber of scenarios (or ' cases *)

In using the traffic assignment sodel, the user must have been identified that, in aggregate, span the cleo specify the trip table that defines the volum* entire spectrum of possible conditions. For each cf traffic traveling from each origin to each such case, explicit analytic expressions have been casociated destination. derived to compute the vehicle's new status.

To provide an efficient framework for defining spilltack conditions that arise from inadequate thno specifications, the pnysical environment is capacity on one or more network 11nts are also ripr:sented as a network. The unidirectional links properly accounted for.

(f the network generally represent roadway costo- -Urban tevel !! MMel n:nts--either urban streets or freeway segments.

Th] nodes of the network generally represent urban The Urban Level !! model is an extension and refine-int:r:ections or points along the freeway where a ment of the flow tcdel used in the TRANSYT signal geometric property changes (e.g., a lane drop or a optimitation pr cq r am (1). This flow sMel in the chang) in grade). 14 vel 11 simulation represents the traffic stream in Figure 1 snows an example of a network represen- the form of movement-specific statistical histo-t: tion. The freeway is defined by the sequence of g r a ssa . Figure 2 shows this histogram representa-links (1, 2) , (2, 3).....(5, 6). Links (8000, 1) tion, which preserves the platoon structure of the cnd (6, 8001) are entry and exit links, r e s pec- g,,ggg, ,g,,,,,

tiv:1y. An arterial extends from node 7 to node 15 The 14 vel !! Icq ic constructs a total of five and to partially subsumed within a grid network. such histograms for each (tu n) movement on each I

r.ach of the four simulation models in TRAfta network links d: scribes tr af fic operaticas in a subnetwork. That it, the user may partition the analysis network into 1. The ENTRY his tcq r sa describes the platoon cubnetworks if he or she wishes to apply more than flow at the upstream end cf the subject lint. This

, one-alaalation ardal concurrently Of the network histcgram is sieply an aggregation of the appropr!-

consists of a f reeway and urban streets, it must to ate OUT turn-movement-sfecific histograms of all partitioned (at least) into freeway and urban feeder links.

,. , j .; . , . ,-V- a. 4, s ,- . ~ ca s  !*

. - . _ - , - - n. - - _ _---

l L 7

.2

~ .*: -

i

. i.

t . ....<. s

. ;u 5 =u;-.

>>,: ,^ .

o -- - i 2 .

L u

~  : : g t ;*c : -c.

=  ;

8 ,

l-M..

E 5

~ - I::g{3 9

. - .w.g::- :

i. 'F -

.o A : -ag:

5, -

. a on-g 2 4,e o = .u .

v t.

• =

2c ..S -

f.-.V t

8 - 6, t i

% ,. 5= .,

.tv/ m. .,

• I*

[ :g s s cio  :. U { -

s , , u t

c g w

.a 9*7

<, i e.r

. c r. .

< r,e .~

e

r2 ca ,

"1 t~

'l Q

• e 1 S .

• n

. .r

1. :

.* c . .*

2- E

• U'NO : * '

E "J'

EP o, 5*..e '

at '

e r4 c ri x 5 --

=  :~ e. .

aon

. mcc fg sI .

5

--.=.**2 o "g -

e..a.v a  %

a . .

• . .j2g * * *8 Cir s 't gsa =3- .

'l

( 00 00

5. $ $.'

0  !?

0ce .

I r 94 lcy .

o 5eoo.e .

o+.# 7 4.,,

-- o-. .e-2. -.

.~

U, m

, , <% gj;31;j 2a. 2.-:: g

-...s -

t s-

=c.;sggg .

a..n. u- .m>

!. 2:: ::t a . -

/ .

=  :-  :, .: <

.4 <

e s

. c.-. ...c.- . ,-

s e r. .

a -I .

' % i,.- i.- g s.!!2* ?.

% 1 1,. i. - F c,. su. ..- . :.

}  :

.".2

?n.,

o

.?

g b 'E 0

f-s

! j ff hh3 .- r L 1- *uE  ::s "c*.*$1 . *'3 2 '

.\ .

- -' . T1 5 t~} ---ag .

• .P y.

-i.

- } ,

is

-n

3a,a - - C 2 a

'I.. . ;8 n

. (. ., - _ s

4 Transportation Research Record 772 13 volumes for each turn navement, given that the cor.- all the link-segment-specific values of volume are tr:1 is 'go'. time dependent.

4. The QUEVE histograms describe the time- In addition, it is necessary to recognise that v: tying ebb and growth of the queue formatten at the the delay emperienced by vehicles at in t e r s ec tion s stop line. These histograms are derived from the is a function of the turn-movement Fercentages, the '

1 t"raction of the respective IN histograms with the presence or absence of oncoming traffic opposing SERVICE histograms. left turners, and the channelisation of lanes on the S. no Ot?T his tog r ase descrite the pattern of ilnk.

traffic discharging from the subject link. Each of the IN histograms is transformed into an OUT histo. Freeway Model gram by the control applied to the subject lint.

Ea;h of these OUT histograae is added into the (ag- T EFID, the freeway traffic simulation model in-gregate) ENTRY histogram of its receiving lint. ciuded in TRAT!4, is an extension and refinement of the MACK sodel developed at the University of Note that this approach provides the 14 vel !! Southern California (J). This macroscopic simula-mod:1 with the ability to identify the characterge. tion edel represents traffic in terms of aggregate tics of each turn-movement-specific corponent of the measures associated with sections of freeway gener-tr:ffic s t r e afs. Each coeponent is serviced at a ally less than 1 mile (1.6 km) in length. The ag-different saturation flow rate, as it is in the real gregate seasures used are flow rate, density, and  ;

world. Furthermore, the Level !! logic is able to space mean speed within the section. The formula-recognise when one component of the troffic flow is tion is based on a fluid flow analogy to traffic cnc:untering saturation conditions even if the ope r a tion s.

cthers are not. The earliest mMeling work Q) used a conserva.

Algorithms provide estimates of delay and stops, tion eqJation and an equilibrium speed-density r ela-r:flecting the interaction of the IN histograms with tion. In FET!4, the equilibrium speed-d en sity tha SERVICE histograms. 14 vel II logic also relation is incorporated into a dynamic speed equa=

prrvides for representing bus testfic as separate tion. Another extension allows vehicles to be dis-cntities (although at a lower level of detail than tinguished by type in three categories (a) automo-Lev 31 !) and for Frcretly treating sp111taca biles and trucks, (b) buses, and (c) carpools.

c ondi tion s . tes t of the capabilities of FETI4 are shown in Figure 1. for each freeway section, there is a Uttan level !!! MMe t variable for entry flow rate, exit flow rate, density, and space mean speed. These variables are l Lav:1 !!! logic is designed for major arterials that distinguished by vehicle type according to the three act as collectors, distributors, circulators, or categories given above.

connectors. As a collector, an arterial would serve Vehicles enter the freeway subnetwork either at ts feed traffic from, say, an outlying region (or the upstream end of a freeway segment or by way of suburb) to a region of higher traffic density. As a on ramps. In the on-raep case, it should be noted distributor, the arterial would serve a reverse that FREr!4 represents only the movement on the tr10, servicing a high demand level at one end and f reeway pain line so that vehicles are introduced at di;tributing this traffic to cross streets the raep gore and immediately merged. Vehicles exit throughout its length. An arterial that serves the freeway subnetwork at the downstream and of a primarily to provide access to adjoining traffic freeway segment or by way of off ramps. In the g:nerators can be called a circulator. Finally, a off-rapp case, FRETI4 represents movement only up to connector arterial links two high-den si t y areas, the of f-ramp gore so that mvement down the ramp is each of wh.ch would be modeled in greater detail (at represented within the adjoining subnetwork.

Lev:1 1 or Level !!). Traffic is associated with two types of lanes:

A user may determine that, although an arterial (a) special-purpose lanes that can be designated to ptsys an important functional role, as described allow use by buses and/or carpools only and (b) c bov 3, a detailed antysis of its traf fic operations re3ular lanes that can accommodate all other traffic 11* e outside his or her realm of interest. For and all vehicle types, including buses and crample, a planner may wish to determine travel time carpools. Traffic is not associated with a clong the arterial from various points to a particular lane but is considered to be uniformly p;rticular location te.g., a shopping center or a distributed over the special-purpose and regular tail station). A traffic engineer may wish to lanes, separately. The nutber of lanes of each type Ctirmine whether congested conditions will occur as is arbitrary.

part of a quien precursor study t o f ind out whether The network that can be represented is quite c more data 11ed analysis is necessary. gineral. Disjoint segments or more general disjoint p are acc W at d. NewaMo-f reeway connec=

To satisfy such needs it is not necessary to c plicitly simulate traffic elements either as t to that involve merge and diverge points, as well as several connected freeways, can be accosamod a ted .

individual vehicles or as platoons. It is FMF14 provides for the representation of an incia Cufficient to calculate the MCEs associated with a dent on the freeway by allowing for the specifica-traf fic environment that is described in relatively grors, aggregate terra. tion of a r educed number of available lanes and a Many investigators have develcped explicit constraint on the flow rate past the incident site.

CC lytic espressions that relate delay ,to traffic volume, control settings, and saturation flow rate Traffic Assianment MM el by using various techniques and asserting a variety of assumptions. tua tra assignm nt model *

• M ed in One formulation that is widely T PMI interfaces with the simulation models. That ccc5pted is that develeped by Wetster (6). The Lev 31 III strulation model uses an extensten of

** ' ' '

• 9 "" " I A I" *##'EI'

• specified trip table tratrix of o r ig i n-d e s t ina t ion Webster 's f ormula to calculate vehicle delay, dema d volumes) and assigns these trips over the Although the 14 vel III mM el is far less detailed

' spec e netw rk. The assigned traf fic volumes are tMn Le vel !!, it is still necessary to pre;erly scpresent the ' time lags

• in the systems that ns, an a@W bto U nk-s peM c turn Fercentages, as required by the simulation seMels.

.r

\\... - - , - s s .

• 3 o i t

( ,/

1

V ,

j 14 Trinsportation R;se:Ich Record 772 j

'i Fhuee 3 FREF L0 sapetplivet. buses and/or carpools may '

/ occupy the special purpose lanes

'l1 I

} [ { }

p vehicles exit y l V;hicles' enter [g jj on freeway l cn freeway

! h Il I!

l all other vehicles, including autos and trucks, v;hicles and, pssibly, buses and vehicles

$nter at carpools. I f they do r.ot use exit at 9 ore of the special purpose Janes. gore of on ramp occupy the regular lanes off ramp Tatde t. TR AF LO sard types.

Thi software then executes the simulation model(s) rauested ty the euser to generate statistical meMures that quantify the performance of testfic sei r.ruwp Card ine operations over the analysis network. This entire M automatic, requiring no manual M a k 'ad'Pwnt R u n inuf uus pr oc ess is M""**"'* * "

int:rvention beyond the initial preparation of the fl *," '[ 'S') i tu,nm, me., men n  ;]

input data.

It is the inclusion of a traf fic assignsent model p ,,,,, y i , nm, mm . ,a n 26 th:t makes TMTio a tool for the transportation f reen ay saudenl ipeutnatnet 27 bat irmtwanies Loel ill D 31 piriner as well as the traffic engineer. T MTLo I provides the planner with a desctlption of the y', ' " ,

• g 3 dynamic response of a transportation system to an ActuaieJ ugrai santrol 39-41 cpplied transportation management strategy for the Inirs olume 5452 sutinet.mt dehmrer 170 Cpecified current, or projected, travel demand G3' **J " ' a's ais samot pittern (trip tabl.). The information provided by T{a f TRATI4 of fers f ar more detail and accuracy than are Time penoJ debm.ter 210 currently available to the planner.

The TRATTIC model Q) uses the U.S. Bureau of Public Roads formulation to relate tint travel tire to volume. It then calculates the link specific vslumes that minimise an ob}ective function repre* noted above, the level !! program consumed 10 e of ccnting Wardrop's first principle (i.e., user opti* computer time at a cost of $5. The computing cost cisation). These data are subsequently transformed for the Level !!! and FREFID models is insignificant into link-specific turn percentages, as requir ed by (on the order of less than 5 s) regardless of th) simulation models, and the simulation is then network sise or volume.

implemented.

IhPUT 1dQUll@WTS OFIAATDsG CMAMCTERISTICS The input stream for TMTLD is pattitioned into sets Th3 TRArtD software was developed by rigorously ap- of cards. Each aet consists of one or more cari plying structured design and programming methodo1- groups, and each group contains one or more card ogies to reduce subsequent maintenance costs and to types (see Table 1). Only those card types that are provide an ef ficient ici"M code. required for a particular application need be The computing time for the level 1 model d e pe nd s specified.

cleost linearly on the number of vehicles that Althoug h a substantial data base is required to occupy the analysis network. The relevant statistic adequately define the traffic environment under fit this model is empressed in terms of the ratio of study, care has been taken to minielse user effort.

the, nueber of vehicle seconds of travel time to For enarple, default values are available wherever co mpu t e r ties. Based on results obtained on the CDC possible, and all input data iters are integers.

7 f,00 we pu t e r , Level I provides a ratio of 12haustive dia9 ostic tests Frotect the user against tepreper inputs to the extent possible. The user 20 C00:1. Fer the validation network of 95 lines, enJ an average content of 375 vehicles, the total also has the optien to review his or ter specified cMcution time for a simulation of 32 atn plus 5 min inputs prior to the execution phase of TRATIO to of fill time was 26 s, which corresponds to a ccat f urther reduce the prospects of incorrect results.

of less than $15.

The computing times conswe.ed by the advel !! and V ALICATION Level lit soiets, as weil as the TREFlu model, d: pend strongly on the size of the network rather All three urban simulation models (14 vel J. Level

11. and Level !!!) have been carefuity validated by than on L2e traffic vo l uee . For a network of c oepa r ing sudel results with field data on a l

,cpprostmately 95 links, the ratto of simulated ttee Details are provided elsewhere

' t3. computing time f or the Level 11 soJe1 is 160r1 on statistical easts.

the CDC 7600 computer. For the validation case (1).

M. .m ., . m . r y, . . - s. . . , .p , c,, , , .ug,,,

e w

TransWrtation Rese!.tch Record 172 15 CU*AENT STA1W Cr TRAr14 4. Network Flow Simulation for Urban Traf fic Control System thase 11 (NET 31M) . Feat. Marwick, The TSAF14 program is now complete and is carently Mitchell, and Co., Washington. CCs and ELD ,

Associates. Inc., Huntington Station, NY, 1973.

und:rgoing in-house testing by THWA personnel. More d; tailed descriptions of TPATio appear elsewhere Q). hT15: 78 230 760-4.

5. D.I. Robertson. TPMSYT A Traffic Network ACK)CW1.F.DOtENT study Tcol. Transport and Road Research 1.ato r a- '

tory, Crowthorne, Berkshire, England, Rept. 1Jt The development of a piodel such as that described in 253, 1969.

this paper is the result of the contributions of 6. F.V. Webstet and 5. st . Cobbe. Traffic Signals.

d Her Majesty's Stationery Cffice, London, Road Re=

(

sany people. In particular, we want to acknowledge Cuido Radelat and George Tiller of the Traffic search Tech. Faper 56, 1966.

Systems Division of FAA Mark Yedlin and Manfredo 7. H.J. Fayne, W.A. Thompson, and i. . !satsen.

• Davila of RID Associates, Inc. William McShane of Design of a Traf fic-Responsive Control System for the Polytechnic Institute of New Yorks and Fred a los Angeles Freeway. Trans.. Institute of

. Wagner of Wagner /McGee Associates. Electrical and Electronics Engineers, Systems J This work was performed ander a U.S. Department Science Cybernetics, Vol. 550-3. May 1973, pp.

cf TransWrtation contract. A portion of the first 213-224.

8. E.B. Lieberman and others. Macrosec.pic Simula-p;rt of the paper was extracted from the Request for Tr e TPAFL4 Pryosal statement of work. tion for Urban Traffic Management j

Model. Federal Highway Administration, 0.5. De-4 R EFERENCES parte.ent of Transportation, Yols. 1-7, 1980.

1. J.C. Wardrop. Some Theoretical Aspects of Poad Traffic Research. Proc., Institute of Civil

' Engineers, Part 2, Vol. 1. 1952, pp. 325-378.

2. S. Nguyen- An Algorithm fot the Traf fic Assign-g ment Problem. Transportation Science, Vol. 4.

No. 3. Aug. 19 7 4.

3. M. Florian and S. Nguyen. Recent Experience with

. Equ!11brium Methods for the Stt.Jy of a Congested I U r ba n Area. Proc.. International Syeposium on i Traffic Equilibrium Methods. Univ. of Montreal, /%Mcerun nf this prn remsorcJ by C..mmittre vn ha//k /h They Nov. 1914. and Chracterancs b

L s

u Abndrment Hybrid Macroscopic-Microscopic Traffic i

/

Simulation Model C1 C. DAVILA ANO f. B. Lit 8tR8A And ti I: The teni l modet. e compoaeas of ee TRu to mecronop.e tratfie semv- is designed to provide a reasonably high resolution tenoa proyem den.eaed 9 e.eNeu treasportanon evitem meaepmeat of detail as well as economy of operation,

} strateges ee duceibed. The te.el l madel 6t dewysed te enpl ciety treet ideas eebedded in several existing traffic l.

(' treffw eenarcd deeices. 6acaude sti cheanehaet.on opsone. ead describe simulation models have been selected, synthestaed,

q tretfke operetsene et erede 6aterwetieae > eaasiderabio dete l. Othee features ,g gg, g ggg g gg g um g g ge L

incdude ocsasted ersai lops. rtt tum on ted. pedeitnen kterference. snd These include the System Development Corporatton j sa,tce4mk flee. Automobdes, buses, coepods, and tevos see emphcauy macroscopic eodel Q) , the TRANS model Q)e the treeted se 6,wt.e dual eatsees. The s>mulet4aa processag uses "e eat 6med ,

, hETSIM (f ormer1Y UTCS-1) model (4), and the SCOT *Q leye, whd mens ene nNors eateemetwativ, en reeve,ed ninee then at The basic concept J

enry time stes tiaterat iesaa.ael Thve. e.i masel es hybe.d in ee sWel (which is not documented).

b esame that the entites eve emeroucow but ee proceumt is mmenopic of Froceneing each wehiele only when it is ttee to 6a truement. Aa enemm ef es tenil modellat.c ss petuated. ee do so is called (in GFSS terminology) ' event-based d

6aput eseuwemenn end ens eeuret of effecte,eaess pmed hv ee mnoel transactions'. The intrinsic benettt of this are md.seted. ead peep sm eff.aeacy mad eshdeson enuam e,e esecuiwd. concept is that it greatly reduces coeputing time, g

particularly when each event le widely spaced in time.

?

h This paper briefly d e sc r i tse s the 14 v el I indele a A careful analysis of these existing acmiels h compoe.ent of the TPAFL4 macroscopic traffic revealed that it would be feasible and desirable to

. Cisulation progran Q). which has twen designed to use an event-tased aFproach in processing all cvaluate transportation systes managesent (TSM) vehicless that is, even when a vehicle is in qJewe strategies. level 1 is the most detailed simulation state, it could to ' jumped' to the stop line and yet h; mo$el within 11LAF14. It provides a. sicroscopic t he a mechannsa of the queos discharge espansiort wave 1 description of the traffic stress and a sacrosecpic coald be preserved. By treating each vehicle in the 3

description of each vehicle movement. This approach traffic stream individually, the model is able to

,h , , y 1.< s .. 'N  ; , ,

t i...i.,,.s i . , .

e ..

. . , .m. , ,

i

. ,