USF Libraries
USF Digital Collections

Field performance evaluation of Peek Traffic's Idris at Leesburg mainline toll plaza


Material Information

Field performance evaluation of Peek Traffic's Idris at Leesburg mainline toll plaza
Physical Description:
ix, 17 p. : ill. ; 28 cm.
University of South Florida -- Center for Urban Transportation Research
University of South Florida, Center for Urban Transportation Research
Place of Publication:
Tampa, Fla
Publication Date:


Subjects / Keywords:
Traffic flow -- Measurement -- Equipment and supplies -- Florida -- Leesburg   ( lcsh )
Detectors -- Testing -- Florida -- Leesburg   ( lcsh )
technical report   ( marcgt )
non-fiction   ( marcgt )


Additional Physical Form:
Also available online.
Statement of Responsibility:
prepared by Center for Urban Transportation Research, University of South Florida, College of Engineering.
General Note:
"February 2000."

Record Information

Source Institution:
University of South Florida Library
Holding Location:
University of South Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 001926055
oclc - 45193384
usfldc doi - C01-00422
usfldc handle - c1.422
System ID:

This item is only available as the following downloads:

Full Text
xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd
leader cam 2200229Ia 4500
controlfield tag 001 001926055
005 20080327141553.0
008 001018s2000 flua t 000 0 eng d
datafield ind1 7 ind2 024
subfield code a C01-00422
2 local
b F544 2000
0 245
Field performance evaluation of Peek Traffic's Idris at Leesburg mainline toll plaza /
prepared by Center for Urban Transportation Research, University of South Florida, College of Engineering.
[Tampa, Fla. :
University of South Florida, Center for Urban Transportation Research],
ix, 17 p. :
ill. ;
28 cm.
"February 2000."
Also available online.
Traffic flow
z Florida
x Measurement
Equipment and supplies.
University of South Florida.
Center for Urban Transportation Research.
1 8 773
t Center for Urban Transportation Research Publications [USF].
4 856


Field Performance Evaluation of PEEK TRAFFIC at Leesburg Mainline Toll Plaza SEMI TRAILER 0 "' ill :!! 100 ] u:: # ,. : ..... : ,: "'. ... 0 2 4 . . Time (seconds) ' ; . . . ......


fiELD PERFORMANCE EVALUATION OF PEEK TRAFFIC'S IDRIS AT LEESBURG MAINLINE TOLL PLAZA PREPARED BY CENTER F O R U R B A N T RANSPORTATION RESEARCH UNIVERSITY OF SOUTH fLOR I DA COLL EGE O F ENGINEER I N G fE B R UARY 2000 T h e opinions and finding s expressed in this report are those of t h e Center for Urban Transp ortation Research (CUTR) and the University of South Florida and n ot necessarily those o f the Flo r ida Dep a rtment of T ransportation Office of Toll Operatio n s (FDOT OTO) or the project technolosy partner (Peek T raflic Inc. ) This repo r t has been prepared in cooperation wit h the the T urnpike Data Ce n ter jn Bo c a Rat on, and Peek T raffic, Inc., Saraso ta


this page IS blank


ACKNOWLEDGEMENTS CUTR would like to recognize the funding agency for this evaluation project, the Florida Depa r tment of Transportation-Office o f Toll Operations. In particular, jim Davis and Doug Martin who served as FOOT project co-managers, and Brett Massey who prov ided on-site coord ina tion at the L e esburg mainli ne toll plaza Additionally, Ms. Carri e Latimer, at the Turn pike Data Center i n Boca Raton, provided plaza l ane transaction reports to CUTR on a daily basis througho u t the eval uation period. CUT R sincerel y appreciates the op po rtunity a ffo rded by the FOOT -OTO to cond u ct this i n dependent fi eld evalua ti on on thei r be half The techno logy vendor partner i n this eval uation wa s Peek Traffic, Inc. (Sarasota, FL), represented by Peter Keen-General Manager. Michael Pietr zy k, CUTR's ITS Progra m Director, served as principal inves tiga tor under th i s contract. The other primary CUTR project staff was Sujeev a (Anu) Weerasuriya (Graduate Research Assistant), and Susan Horsman (Program Assistant). Witho u t the assistance and effort provided by the aforementioned in d iv iduals, this eval uation could not have been conducted and doc u mented. Ill


EXECUTIVE SUMMARY ldris Smart Loops Technology is an inductive loop based traff i c sensor manufactured by Peek T r affic Inc., u nde r license f rom Diamond Consulting Services. The system uses Peek Traffic's standard off-the-shelf, high per f ormance inductive loop detectors, fitted with a special device to allow communicat i on with a separate micro-con tr oller with a POSIX compliant operating system running the ldris a l gorithms. This device uses an extension of the ldris core algor i thms orig i nally developed for incident detection, and toll ing for Design Buil d F i na nce a nd Operate (DBFO) app li cations in the United Kingdom, to achi eve vehicle separation and classif i cat i on. This new l y developed a l gorithm fo r i n ductive loops was i n stalled in two different plazas to test vehicle separation capability. A preliminary field eval uation (Phase I) was undertaken at the Anderson P l aza on the Veterans Expressway in Tampa. Phase I testing was co n ducted for AM and PM Peak hours, fo r three consecutive days a total of 7 hours and 15 minute s tdriswas eva l uated against an imp r oved overhead ultraso nic device for ve h icle separation Following the successful conclus i on of Phase I, FDO T-OTO authorized Phase II testing, which was und er tak e n at t h e L eesburg Toll P l aza on the Florida Turnpike. ldris0 was installed at the Leesburg P l aza to conduct statistically signi ficant and comprehensive f i el d tests. Phase I and Phase II were for vehi cle separation accuracy only. The Phase I field results showed 99.72 percent vehi cle separation accu racy o f ldris compared to 99. 38 percent accuracy o f the improved overhead ultrasonic device. The number of ve h icles eva luated over the 7 hour 15 minutes period was 1,761 vehicles The Phase II tests showed ldriswas accurate in separating veh i cles 99.89 percent. In P hase II a total o f 45,282 vehicles we r e verified with 49 ldris erro rs. Algor i thm modificat i ons by Peek Traff i c, Inc occurred on February 27 and again on April 12, 1 998 l eading to improved results after each modificat i on. For example for t he eva l uation per iod after Apr i l 12, ldris separation acc u racy was 99.92 percent. This improved separation accuracy was also attai ned during the portion of evaluat i on that shifted from continuous veh i cle sampling to a biased sampling (i.e., hours with greater volume and mix of vehicles). v


this page IS m1ss1ng


this page IS m1ss1ng


this page IS m1ss1ng


TABLE O F C O N T ENTS AC KN OWLEDGE MENTS ..... .... .... ..... .... ..... . .... ...... .... ... ..... .... ...... .... ....... ..... ..... 111 EXECUTIVE SUMMAR Y .... ... ._ ..... ... .... .... .... ... ...... .. ..... ..... ...... . ..... .. .... ... .. ........ .. V 1.0 PURPOSE AND B ACKGROUND .................. ................. .................................. .. 1 2.0 /DRt S TECHNOLOGY .................................................................. .............. ... 2 2. 1 System Description .... .... .... .... .... ................... ........ .... ...... .... ..... .... .......... 2 2.2 System Operation .... .................................. .... ..... ....................... ...... ....... 2 2. 3 Toll Plaza Configuration .................. .... ................................................... 3 2.4 Features of ldris Smart Loops' .............................................. ................ 3 3.0 F I E L D EVALUATIONS ............................. ............... .. ................... ................... 5 3. 1 Anderson Plaza (Phase I) .................................... .................. .. ................ 5 3.2 Leesburg Plaza (Phase II) ................ .... ............... .. ................ ... ................ 7 3 .2.1 Daily Eval uat i ons ................................... .... ................. .. ........ ............ 8 3.2 2 Four Hour Random Evaluation .. .... ...... .. .......................... .... .............. 9 3.2.3 F i fth Hour Random E va l uation ............ .. ............................................ 9 4.0 EVALUA T ION RESU L T S FO R VEHICLE S EPARAT ION .............................................. 9 4. 1 Anderson Plaza (Phase I) ................................ .... .................................... 9 4 2 Leesburg Plaza (Phase II) .......... .. ..................................... ........ ............ 12 5.0 OBS ERVATI ONS OF lORis PROBL E M AREAS .................................................... 15 5.1 Time Shifts ....................................... .... ........ ....... .... ........ ..................... 15 5.2 Vehicle Configurations .............................. ....... ............... .. ................... 15 6 0 SUMMARY O F FINDINGS ..................................................... ......................... 17 APPENDIX A lORIS Product Information VII


this page IS blank


Table Number 1 2 3 4 5 Figure Number LIST OF TABLES T itle Error Table for Varying Sample Sizes for 95% Confidence Interva l Compara t ive Eval uation of Phase I Results Summary of Phase II Observations Summary of Phase II Erro rs by Period Detect ion Erro r s by ldris0 LIST OF FIGURES Title 1 Typ i calldris Installation 2 Equipmen t Configu r ation at Anderso n P l aza (Phase I) 3 ldris at Anderson Plaza 4 Determin i ng Vehicle Configuration Type IX Page 8 10 13 15 16 Page 3 6 6 7


1.0 PURPOSE AND BACK GROUND Tol l agencies use a variety of tec hniques to co ll ect revenue on toll roads: manual c ollection, automati c coi n machi nes, or automati c toll co ll ection systems with pre paid accoun ts. Each agency may use diffe r e n t toll co ll e ction structures (toll r ates), which a r e based on one or more vehicular features s u c h as number of axles, length, weight, number of occupants of the vehicle, etc. A vehicle separation and classification system is necessary to ensure proper collection of toll fares in automatic coin machine lanes. Vehicle separation, axle counts, and classification are the three vital features of any automati c vehicle classification (AVO system. Almost every to ll agency aroun d the country utilizes some form of an AVC s y s tem to ve rify and aud i t toll transactions For a detailed c ompilati on of variou s A V C systems refer to Task Order # 1 -Automatic Vell i c/e Classification Systems report by CUTR'. Florida s Turnpike is currently using overhead ultrasonic devices for vehicle separation and pressure sensitive axle detectors ( tre adles) for vehicle axle detection. Using the signals from the overhead device and the treadle, vehicle classification is performed by a separate unit called a 'treadle doctor'. These existing overhead devices and treadle compo nents have generally become a sufficien t m aintenance problem to cause the F l orida Departmen t of Tran s portation Offic e of Toll Ope r at ions (FOOT .OTO) to invest igate alternatives. FDOT-OTO entered i n to a cont r act with CUTR to assess vehicle separat i on capability of ldris, a t echnology marketed by Peek Traffic, Inc. for use in a toll plaza environment. For product desaiption, capabilit ies and other Jdris0tests refer to Appendix A. The Phase-! test was an initial field test that investigated the capability of Jdris0 in veh icl e separation The Phase-! field tests were conducted a t the Anderson P l aza on th e Veteran's Expressway in Tampa, Florid a in August 1997 The three-day i n itial fie l d test i ndicated the veh i cle separatio n capability o f ldris t o be more acc ur a t e than the improved' ove r head u l trasonic unit OTO used in the toll plaza 'Task Order # 1 Automatic Vehicle Classification (AVC) Systems, Int erim Evaluat ion Report Center for Urban Tr ansportation Research, Uni versity of South F lor ida, Tampa Fl, j uly 1997. 1


J'iefa Performance 'Evaluation: Peek, 'ITa !fie s J tfris CUTR conduc ted fie ld tests with ldris i n two differen t phases, bo t h with a specific purpose. Following the success of ldris in th e initia l tests, it was i nstalled to co nduct more comprehensive and statistically significant fiel d evaluation (Phase II) at the Leesburg P l aza on Flo rida s Turnpike in leesburg, F l orida leesbu rg Plaza provided a higher traffic volume and a good m i x of var i ety of vehicle types. The Phase II f i eld tests were conducted f rom February 17 -June 21, 19 98 In Phase Ill tests, ldris output was going to be i n t egrated i nto a new QNX/05 lane Contro ll e r to t est the operation of the vehicle separ ation capabi lity under actual operating conditions as part of FDOT -OTO lan e control system. In addition, the use of ldris fo r vehicle classific ation was to be studied. Phase Ill tests have been delayed, whil e FDOT completes th e deployment of its S u n Pass program, and Peek Traffic, I nc. comp letes dep l oyment of severa l operational ldris systems with other agencies. Phase Ill is expected to be completed i n 2000 This report summarizes results f rom the Phase I field test at Anderson P laza, and focuses on the field eva l uation tests from Phase II. 2.0 /DRtSTECHNOLOGY 2 1 System Description ldris Smart loops Techno l ogy was developed by DCS and WS Atkins, specifically fo r Design Build Finance and Operate IDBFO) contracts i n the United Kingdom. The strictness of these private initiat ives r equired a sophisticat e d detection system that could detect and monitor all freeway la ne s s imultaneou sly regardless of weather or flow cond itions. ldris Smart Loops use an extension of the core algor ithm s to provide reliable veh i cle separat i on and trigge r outputs fo r standard Violation Enforcement Systems (VES), packaged i n a single lane configuration. When equipped with special axle loops (or using existing axle sensors, e.g Trea d les), ldris Smart loops provide a comp l ete A u tomatic Vehicle Classification (AVC) System fo r pre o r post-classification app li cations. 2.2 System Operation Using advanced signa l processing, the ldris Smart Loops use the core ldris algorithm enhanced for the toll i ng industry to deal with specif i c requ i rem ents for cons i stent vehicle violation tr iggers, and axle-based cla ssification The system uses Peek Traffic Inc.'s standa rd off-the-shelf, h i gh pe r fo rma nce i n ductive loop detecto r s 2


:fieU 'vauaticn: Pee{( 'Traffics liiris in conj u nct i o n with a separate micro-controller running t he ldris a l gorithms (Octagon 586 133 MHz, rated over full NEMA environmental range). ldris Smart Loops use two 6' 6' x 6' 6' (2m x 2m) l oops separated by 6' 6 (2m) in each l ane (For axle-based classification, special axle l oops are l ocated between the main ldris loop pair). An op t ional thi rd loop may be used to provide a consiste n t tr i gger outpu t i f heavy congest i on is preva l ent after t h e toll plaza. In most cases, the third loop would not be required to provide an accurate vehicle viol ation trigge r r e .. : : I I l I''A) -... i I : ......... i. -t --j ' I ...... .,... Figure 1 T ypic al/dr is01 nstallation 2 3 T oll Plaz a C o nfiguration The i nductive loop detector genera tes raw analog loop data, which is then sent direct l y to the ldris microcomputer (586 133 MHz, rated over full N EMA environmental range) running mu l t i ple ldris0 algori t hms. The analog loop data is ana l yzed by track i ng the i nduct i ve signatures of each loop to determ i ne the. events occurring on all l oops simu l taneous l y. By analy z ing all loops simultaneously and comparing the signatures, the algor i thms can then determine exact l y what has tra n spired on each loop in the array, calculating t h e relational effec t s of each to prov i de precise and cons i stent outputs regard less of traffic flow and environmental conditions'. 2.4 Features of ldris Smart l o ops' The following are some o f the features of ldris Smart Loops in a toll ing environment: ! Accurate Vo l ume Data and Vehicle Separatio n : ldris Smart Loops include mult i p l e algo r it hms for whi ch p a tent protection is pending. These allow the Information Pamphlet, Peek Traffi c Inc. Sarasota, FL 3


:Fidtf Perfomu:mce v aluation: Peek_ Traffic's ltfri.s system to accurately proc ess veh i cles i n conditions rangi n g from stop-and-go to free flow. ldris0s un i que straddling algorithm ensures v e r y accurate s e paration, even with poor lane discipl i ne in mu ltilane express tol l installations. 0: Consisten t and Reli able Trigger Outpu ts: An i mportant requ i rement in many toll lanes today i s the ability to prov i de a consistent and reliab l e trigger at the end o f a vehicle in a l l weather conditions. This tr i gger is used for t o ll violators and i s required for speeds ran g i ng from 0 to 100 mph. ldrisSmar t L oops pr ovide vehicle triggers based on its two -l oop configuration to with in +1-1 foot regar d less of vehicle speeds or envi r onme ntal conditions ldris"' Smart Loop s a r e capable o f d i fferentiating between two vehicles tailgating and a sing l e veh i cle towing a trailer. Tail gating vehicles get two triggers while a veh i cle in tow gets a single trigger. ldris Smart L oops a r e able to track veh i cles in e i th e r single lane plaza conf i g u rations or multilane express or open road applicat i ons, trigge r ing VES systems for the approp r iate lane. ldris"' Smart Loops cope with r ecurrent output congestion w i th the add i tion of an optiona l 3' l oop at the trigger point. ! Ma i ntenance and Envi ronmental R e li ability: Loops have proven over t h e year s to be very rel iable when insta ll ed correctly, requ i ring litt l e or no ma i ntenance T he mean time between fai lu res of a prope rl y insta lled l oop system is typ i cally 25,000 hours with many loop systems lasting ten to twenty year s Being below the road s urface, loops a r e not suscept ible to vanda l ism or tampering by passers-by or damage from veh i cle passage. The transmission s i gnal of loop based technolog ies are also not affected by rain, wind, snow, ice or vehicl e spray, therefore significantly improv i n g the reliabi lity and per formance of the system in poo r weather condi t ions ! C l assif icat ion Data : When appropriately configured w ith a n ldris"' M T S Ax l e det ect o r and loops, the ldris Smar t loop system can d e t ect vehi cle axles a n d t i re w i dth enabling a complete AVC System. T h e class scheme i s app li cation depende n t based on avail able vehicle parameters (ty pica ll y numbe r o f axles, spacing, vehicl e profile, length, and presence of dua l tires). Class is provid e d as a discrete message follow i ng passage of the vehicle o r alternatively as part of a series o f messages while th e vehicle is t racked thr ough the ETC or Manual lane. Provision fo r t h ird party sensors all ows other classificat i on d i mensions to be u t ilized to determined vehicle class. 4


f[ieftl Performance 'Evaluation: 'Traf:IU s It{ris <: Speed Data: Using two loops i n each lane allows the /dris Smart loops to calculate speeds very accurately in fre e-flow conditions to within +/-3 mph Calculating speeds in slow moving congestion is much more difficult, however, +I3 mph is still achievable. 3.0 fiELD EVALUATIONS Two field eva luat ions were conducted of ldris. The preliminary (Phase I) field evaluation was conducted at the Anderson Plaza on the Veterans Expressway in the Tampa Bay area from August 6 through 8, 1997. The second f ie ld installation and testing (Phase II) was at Leesburg Plaza on the Flo rida Turnpike f rom February 17 through June 2 1, 1998. 3.1 Anderson Plaza (Phase I) Following a site visit by both CUTR and Peek Traffic Inc., the third northbound l a ne (#5) from the right shoulder was selected to conduct field tests. The vendor was t esting th e /drissystem with temporarily installed loops prior to the field tests on the same lane Since the grooved" pavement s urface of the test l ane i s not conduciv e to maintaining stable loop conditions, Peek Traffic, Inc. perso nnel cut loops into the pavement before the fie l d test began. A schematic diagram of the loop l ayout in the test lane is shown in Figure 2 The loops #1 and #2 were cut approximately l4 i nch deep and #3 and #4 were cut a pproximately 1 i nch d eep into the pavement The pavement was cut using a dry cutting plate. A member of the project team from C UT R was present duri n g the initial pavement cutt ing Th e actual field test was postponed due to the malfunction of the tr e adle in lane #5, a weaker signal from loop #4 than #3, and a dead closed-circuit t elevisio n (CCTV) camera due to a lightening strike. Once these problems were resol ved, the actual field tests were started on August 6 (Wednesday), 1997 and were concluded on August 8 (Friday), 1997. During the tests, at least one project team member from CUTR was present dur ing the first four peak p

Peiformance 'Evaluation. : PuR_ 'Traffics ]t{ris ldris"' data r eceived was impo rted into 'Exce l spreadsheets fo r processing before CUTR i n -house evaluation started. T h e plaza transactio n repo rts we r e used to est ablis h m orn ing a n d afternoon peak hou r s Once the peak hours were established, plaza video was viewed to verify each veh i cle passage reported in the ldris"' dat a as well as plaza transaction report duri n g the peak h ou r s This procedure was repeated for a ll t h ree days dur ing the A n derson P l a z a fie l d evaluat ion. Toll Booth IO'lO" Barrier Gate Loop 30' ,C) " 0/onected 32" I 4 1 r i : i 3 2 / ... ......... . ,. ... ,1 ................. l! l 1 ' 1 6 I I . Loop# 1 2' : Loop# Loop#4 Loop # 2 i ; : 6' X 6' H 6 >l'x 6 >I' 6' X 4' w! 6 w x 6 wi 6 Wx 6 w ,. x6 u I li : : ' II -; 1-...... Dicection of " 23" n 2 lh" Travel 3 >I'' 19" ,_ '\ llxpaJon Joint Vehicle Treadle Presence -Loop F igure 2. Equi p m ent Co n fi g u r ation a t A n d e rson Plaza (Phase I) Figure 3. ldris a t And e r son Plaza 6


!fie(t PeifQ/7flance 'EvaCuation: Peel(. 'Traffic's Uris Figure 4-Determining Vehicle Configuration Type 3.2 leesburg Plaza (Phase II) Statistically significant fie l d tests were conducted at the Leesburg Plaza on Flo r ida's Turnpike. Proj ect team members from CUTR, Peek Tr affic Inc., and OTO held a meeting at Leesburg Plaza and decided to use a southbound lane (#5) for the field test. The ldris ou tpu t and the plaza transact i o n record fo r the test lane were electronically transmitted to CUTR on a regular basis f ro m Peek Traffic, Inc., and OTO Data Center. These dat a along with the plaza CCTV video recordings during the entire testing perio d were used in the evaluation. During the entire evaluation, CUTR used ldris and plaza t ransacti on records and manually matched each vehicle passages on lane #5 with the CCTV videos. Since the field eva luation was focused on vehicle separation capability of ldris, each error made by the ldris unit was recorded in its entirety. Si nee the objective of the leesburg field test was to study the vehicle separation capabilit ies of Jdris, accuracy of plaza data was not verified w i th the plaza video. As in the Phase I f ie ld evaluations, Jdris0 data was imported into 'Excel' spreadsheets for proce s s i ng before CUTR in -house evaluation started. The plaza video was viewed to verify each veh ide passages reported in the ldris< data for the peak hours. This procedure was repeated for all the days of the Phase II field eva l uation The official field test was started on February 17, 1998 about 6pm. The evaluation of ldris"' can cont inue as long as upper bound of 95 percent confidence interval with accuracy specification of 99.97% will not be exceeded fo r three consecutive periods. For e xample, no more than 4 errors can occur in the first 500 vehicle passages, and no more than 5 errors can occur in the first 2,500 vetiicle passages as shown in Tab le 1. The table was used for just an early warning signal for required perform ance 7


Performance 'Evaluation: PeeK_ 'Traffic s Uris Tab le 1 Error Table for Varying Sampl e Sizes for 95% Confidence Interval Number of Errors Sample S!ze 1 2 3 4 5 250 100 00% 100.00 % 100 .00% 99, 96% 99.74% 500 1 00.00% 100.00 % 1 00.00% 99.98% 99.8 7 % 750 100.00 % 100.00% 100.00 % 99. 99% 99 .92% 1000 1 00.00% 100.00 % 99.99% 99.94'io 1250 100.00 % 100.00 % 100. 00% 99.99% 99.95% 1500 100.00 % 100.00 % 100 00% 99.99% 99.96 % 1750 100.00 % 100.00% 1 00.00 % 100 00% 99.96% 2000 100.00 % 100 .00% 100.00% 100 .0 0',1, 99.97 % 2250 100.00% 100.00 % 1 00.00 % 100 .0 0 % 99.97% 2 500 1 100.00 % 100.00o/o 100.00 % 99.98% 3 .2. 1 Daily Evaluations These daily eva l u at i ons were done ar ound the clock. The da il y evaluation period began on February 17, 1998. ldris0 did not meet the erro r crite rion stipulated in Table 1 henc e th e eva l uation was stopped and the vendor w a s notif ied. Due to errors in separating and i dentifying large vehicles such as pickup truc ks and vehicle t ranspo rte r s adjustments were made to the ldrisf!'f s algorithm s et-up parameters to better separate these ve h icles. Th e daily eva l uat i ons restar ted on February 27, 1998. The ldris0 outpu t was e valuated for any separation or uni de ntified vehicles w ith the plaza informa t ion on a 24-hour basis. On March 4 1998, CUTR observ e d that ldris unit frequently m issed v ehicle transporters and also that it was not meet i ng the error specifications stipulated i n Tab le 1. The vehicle transpor ters have rectangular meta l frames (empty in the middle) very close to the pavement surface. Therefore these types of vehicles provid e d a very unique feature rich signa t ure and a c ha llenge to ldris""s signature based algorithm. Once t he vendor expa nded the memory buffers to allow proper processing and iden t ificat ion o f the veh i cle transporters the field test was restarted on Apri I 12, 1 998. 8 --


'FreU Performance 'vauaticn: PeeK_ 'Iroff/c s Mris 3.2 .2 Four Hour Random Evaluation On Apr il 13, 1 998 it was decided that four hours per day eval ua tion would substitute the 24-hour daily evaluat i o n The daily transaction data from p l aza was used to establish morning and a ft ernoon pea k hours each day The fou r hours included were the AM and PM peak hours of the day and one hour eac h d u ring the morn ing and afternoon per iods. These two morning and afternoon hours were s e l e c ted rando m l y. The four hours a day eva lu a t ion started on Apri l 13, 1 998 a n d contin ued through April 20, 1998. Some v i deo time was not a vailab l e i n certain evaluation ho urs a nd was i ncomp l ete during some periods. (Manual r eplacement of video tapes cou l d not be perfo r med by FDOT-OTO in a time l y fashion). 3.2.3 F ifth Hour Random Evalua t ion Dur i n g the four hour a day evaluation CUTR observed that most of th e vehicles pulli n g trailers a nd he avy vehi cle traffic might not f a l l i nto any of the four ho urs. The r efore, CUTR decided to add a fifth hour, tha t is, t he pea k ho ur d e t e rmined exclusively for heavy v ehicle traffic such as, vehicles pulling t raile rs. This fi ve ho ur s a day evalu a tion started on Apri l 21, 1998 and contin ued un t il june 21, 1998. N um ber of vehicles with more than 2-axles t hat passed duri ng the fifth hour was reco rded for each day. Duri ng this per i od, some video t ime was again incomp l ete as well as some hours of the day unava il able. (Manual rep l acemen t of video tapes could not be per formed by FDO T .OTO i n a ti mely fashi on). 4.0 EVALUATION R ESULTS FOR VEH ICLE SEPARATION 4.1 Anderson Plaza (Phase I ) The preliminary eva l uation took place over a three-day period, Wednesday August 6 throu g h Friday, August 8, 1997 a t t h e Anderson P l aza on l ane #5, a manned lane in use from 9am to 9pm. Gro und tru th co u nts were obta ined from manua l review of videotape fo r each veh i cle passage, therefore t im e and budg et constraints limi ted thi s comparat i ve eva I uation to the peak AM and PM peak hours e ac h day; a t otal of 7 hours 15 minutes. Peek Traffic Inc., provided output f r om ldriscont a in ing day, t i me, l eng th and speed of each ve hicle that passed l ane # 5 fo r the 3-day continuou s eva l uat i on period. The plaza CCTV outputs and a detai led transaction repor t fo r la ne #5 for the ent ire three days was pr ovided by the FDOT-OTO. Manual review o f v ideotapes f or the 7 h ours and 15 minutes was reco rded for e ac h separate vehi cl e passage thr ough lane #5. For the t hree day, collec tive 7 hou r 15 m inu t e comparison, ldris

Perf()f7TI(lTIU va[ uaticn: Pee(( 'Traffics zaris perio d were 1,761, which, accord ing to Peek Traffic Inc.'s est ima tion of tota l 72hour vo lume, represent s about 26% of all veh icles passing thro ugh l ane #5. There were a total of f ive (5) ldris errors and a t ota l of eleven (11) U lt rasonic erro r s duri n g th e three-day eval uation p e r iod. Based on the init i a l results achi e v ed, FDOT O TO autho r ized P hase II, a s tatisticall y s i g n i f i cant exte n ded fi e l d t r i a l at Leesburg Pla za. The resu lts of the peak-hour comparative ev alu a t ion for each day a r e shown in T able 2 Table 2 Comparative Evaluation of Phase I Results Traff i c Volume From Time Video Ultrason ic ldris Remarks E valuation Results from August 1997 1 0:30 10 :44am 35 35 35 ldris'b did not reco g nize t wo vehicles ( 2 a x l e car 10:4 5 -10: 59am 30 30 28 immed i a t e l y f ollowe d by 2-axle jeep). 11:00-11 :14am 3 2 32 32 1 t : 1511:29am 33 33 33 T otal /30 /30 128 5:00 5 : 14pm 9 7 1 96 9 7 U l t rasonit did not retOJtniz e a 2ax l e car. 5 :15 5:29pm 1 26 126 1 2 5 ld r i s0did not re c ogni z e a 2-axl e car 5 : 30 5:44pm 111 111 111 5:45 5 : 59pm 107 107 1 07 6:006:14pm 100 100 100 Total 541 5 40 5 40 Total f or DaY 671 6 7 0 668 Evalu a tion ResultS from Au ust 7, 1997 11:00-11:14am 2 3 22 2 3 Ultrasonic did not recognize a 2-.axl e car ldris0 did n o t r e cogniz e a v ehicl e ( two, 3-ax l e lfuc k s in succession, wer e recogni zed a s ol e 1 1:15-11 :29am 30 30 29 v ehicl e ) 11:30 11 : 44 a m 2 1 21 2 1 1 1 :45 11 :S9 a m 3 1 31 31 Total 105 104 104 10 ------------------------


:FieU Performanu 'Evaluaticn.: Pee( 'I'roffic s I w Table 2 Comparative Evaluation of P hase I Results [' -, --. .,. ( ... ,..-. r' ,. .... __ ....... ,.,. ... .-;;;;. l; ' ' 1"" -' - < \.-'< _,:,_ ;< d "' j' '( ', : ,--. ,: : ';. ', ',.<;,;, <.-. [ ... --r L ,_., ' '<-.;:f' :t : ,"-'' . '' .c;. { Ltil;;l/:J: : :: U l trasonic did not recogniz e two veh i cles (2-ax l e 5:005: 1 4 pm 1 01 99 1 01 car and 2 a x le moto r cycle at no n-s u ccessive t i mes). 5: 1 5 5:29pm 117 1 117 117 5:30 5:44pm 1051 104 105 Ultra son i c d i d not recogni z e a 2axle car. d e tected a vehicle t hat was not p resent (de te c ted a veh icle between a 2-ax l e ca r f ollowed 5:45 5:59pm 103 103 104 in about 30 seconds by a 2-ax l e van). Total 426 423 427 Total fo r Dav 531 527 531 Evaluation Resu/1.$. /rom August 8,.. 1997 33 33 32 ldfisiSidi d not recognize a J..axle pick-up truc k w ith 10:45 1 0:59am a tra i ler. 11:00 -11: 14am 371 37 3 7 11:15 11:29am 371 37 37 1 1 :30 11:44am 37 3 7 37 Total 1 '14 144 143 5:15 5:29pm2 105 104 1 05 Ultrasonic d i d not recognize a 2-ax l e car. 5:305:44pm Ultrasonic did not recognize tv.o, 2-axle cars a t 116 1 1 4 116 non-svccessive times. 5:45 5:59pm U l t rasonic d i d not recognize tv'O, 2-ax le cars at 1 09 107 '109 non-successive t imes. 6:00 6:14pm 85 8 4 85 U ltrasonic did not recognize a 2 -axl e car. Total 415 409 415 Total for Day 559 553 558 Grand Total 1761 1750 1757 ldris"' had a total of 5 errors. Notes: 1 L a n e #5 video was no t cont inu ously avai labl e i n this t i me i ncrement due to secur i ty system switches t o monitor dat a room or tunnel upon entry. Comparisons were made onl y (or the vehicles that c oul d be manually observed f rom t he v ideot<>pe. 2. I t was raining during this time period. 11


Perfomumu 'Evaluation: Peel( 'Traffic 5 I

;JUUPerfomumce 'Evaluation: Peek_'Iraffic'sltfris Table 3 Summary of Phase II Observations 13 ----------------------.


!{ieU Peifcrmaru:.e 'Evaluation: Put(, 'Traffics Uris Table 3 Summary of Phase II Observ ations Total Daily Cumulative Veh icle Number o f ldris Errors Date (1998) Vehicles Couni O bserved ... May 1 846 29,456 1 May 18 590 30,046 1 May19 506" .. 30,552 1 May 20 566 3 1 ,118 0 May21 084 31,802 0 May25 178 3 1 980 0 May2.7 116 32,096 0 May 29 7 1 0 32,806 0 "' june 4 435 33, 241 0 j u n e 5 8 6 5 34,106 3 )tine 6 860 3 4,966 0 j u n e 7 1,067 36,033 1 .. )l!ne 8 765 36,798 2 ._. j u n e 9 658 3 7,456 1 June 10 700 38,156 0 June 1 1 9 5 38, 251 0 June 13 716 38,967 1 j un e 1 4 1 ,067 40,034 1 june 15 698 I 4 0,732 0 Jun e 16 333 41,065 0 J u n e 1Z 828 41,893'. 0 Jun e 18 347 4 2 ,240 0 jun e 19 957 43,197 0 J< n e 20 973 44,170 0 June 21 1,112 45,282 1 45,282 49 N otes: 1 From F e brual)' 18 t hro u gh F e br uary 26 ev al ua tion \'vas postponed. 2. Compl e t e data was n o t ava.ilab le f o r Ma r c h 1 3 P laza data was not ava i lable for March 2 4. f rom Ma r c h 5 t h r o u g h Apr i l 1 1 evalua t ion was post poned. 5. O n Apr i l 1 6 April 30, May 22, May 24, May 26, May 28 M a y 3 0, Jun e 3 and J une 1 2 v ideo t ime was not available. 14


'FteU Perfonnance 'Eva{uatlon: Peel(_ 'Traffic 5 Irlris Tab le 4 Summary of Phase II Errors by Period 27 -21 45 Config u ra t i on cha n ge was made to the ldris"'s a l gorithms to imp rove discrimination o f towed vs. tall-gating 99.90% vehicles. Corr ection was made to t he ldris"'s du r i ng ho urs o f greater vo l u me and mix of vehicles. 5.0 OBSERVATIONS OF IDRJS PROBLEM AREAS 5.1 Time Shifts T he ldris0 unit opera t ed as an ind epende n t unit from plaza equipm ent (except receiving NC power (rom plaza) during Phase I and II tests. The vendor technicians synchronized the ldris0time to the plaza real t im e on a regular basis. Despite these efforts, ldris t i me (based on PC clock) and the plaza real t i me were observed to deviate within 24-hours of operation. Once CUTR received the ldris output from the vendor, the ldris time was synchronized to the p l aza real time that was appea r ing in the plaza video. The applied t im e modification ranged from few seconds to th ree minutes during both Phase I and II tests. The modification was done merely for evaluation purposes. 5.2 Vehicl e Configurations Jdris had separation problems w i th transporter trucks ini t i ally. Upon being notified by CUTR and given the opportunity to adjust their algor i thms these changes resulted in improved accuracy each time. An analys i s of the ldris errors du r ing February 17 throug h March 4, 1998 showed that i n 4 of the 16 errors ldris split a pick-up truck w i th a tra il er into two vehicles (See T able 5 for error details), while i n the other 12 cases, vehicle transporter trucks were missed completely. A total of 5,691 vehicle passages were verified during th i s period resulting i n 99.72 percent vehicle separation accuracy before the ldris a l gorithms were modif ied. 15 ---------------------..


:FieU Perfol'lnlliUe 'Evaluation: Peek_ 'Traffics Itfris ., .... Day ?J,c "'' ., "-; .. lhe weel<. Time Feb. 1 7 Tues. 20:09:4 7 20:10: 4 9 2 0 :23:31 21:42:52 Feb. 27 F ri. 1 :41 : 59 4 : 2 3 : 4 1 8:56:03 9 :57:14 20:22: 3 4 20:5 0 : 3 4 Feb. 28 S a t 1 :52:03 tvlar 3 Tue. 0:04:04 1 : 4 0 :11 1 : 54:27 7:13 : 2 1 rvtar. 4 Wed. 8 :57:43 Apr. 1 2 Su n 13 : 57:54 1 6:33:02 17:26:1 9 A pr. 14 8:05:01 Apr 24 Fri. 7:03: 39 Apr 2 5 S at. 14: 36:37 Apr. 29 Wed. 17:10:14 20:1 1:17 2 0:14: 5 6 May9 Sat. 16:30:1 5 May 1 0 Sun. 18: 4 6 :06 18: 4 9 :38 Table 5 Detection Error s by fdris Ot$c:ription of &ror M issed a vehicle tra n sporter S-epara ted a picku p w ith a t railer i n t o two veh i cles Separ a ted a p i c k u p w ith a t raile r i nto tv-o vehi cles Separated a p i c k u p w i t h a t r aile r i n t o tv.o v e hicles M i ssed a vehicle uansportet M issed a semi-tractor ttailer M i ssed a veh i cl e tran S f)Or t er Missed a v e h i cle transporter Separated a pickup wit h a U-Haul i n t o t>A o vehicles M i ssed a veh icle tr a n s porter M i ssed a ve h i cle transportet M i ssed a v e h i cl e trans port et M issed a doub le-t racto r tr ailer M i ssed a v e h i cl e trans porter M issed a ve h i cle ttansport et M issed a v e h i cle transporte1' Separ ated a jeep with a trailer i nto t w o vehicles S epar ated a with a trai l e r i n t o two v e h i cl e s M issed a spo utility vehicle (SUV ) Separated a p ickup with a tta iler i n t o two v e h icles M issed a ca1 l d tis* counted o n ly two motOtcyclcs of t he ( ou t cam e together. Plaza c ounted on l y thr e e o f them. S eparated a S U V wit h a trail e r into two veh icles Mis.sOO consecuti vely : vehi cl e trans pone r SUV, aOO t wo cars Missed two car s consecutivel y Mi s sed a pickup and t hree cars came consecut i v e l y Mi s se-d one o i the t wo motorcy cles came together Mi s sed a car 16 Vehicle Preceding Sutceeding Jeep Pickup Pickup Pickup 2 -tr a ctor 2-tr ac tortrailer uai l e r Car Car C a r M i niVan Car Semi-t r a i l e r C a r Semit railer Min i Van Car Car car C a r M i n iVan P ickup Van S e mi-tra i ler Car S e mi-trail e r Car Car U Haol t ruck Mo tor cycle Car Car Car Car Car s u v M ini Van Car Car Car Semi t rail e r Car Car Car Car Car Car Car Car Car P i c k up Car Motorcy cl e P i c kup Car


!fieU Performance 'Evaf:uation: Pee( 'Traffics Itlris T able 5 De t ection Errors by ldris -.. --m-=r .. . ,--.,,. ---"--:,c : ", '' ,. '-J, ,r'\" c, ,. . I I , :'. 'i&tc: !. ., . <'' : >. .,, -<;_ 1 .. ,.1 ,.. ". -...... " .;,_ .. :._ ,,( o:-,_v . :. :;;i .. .-":;; ."!:' '' ,---:':'. ::...--/--'"v.,= ,..., .,., i\ <:. '' '. -<.','"r .. :ll .. ;; .. ": :-:-_ t > _____ May 1 7 Sun. 22;40:10 Separa t ed a truck with a traile r i n to lwo vehi cles Car Car May 1 8 Mon. 1:47:49 Missed a semi-trailer Van Semr-trailer May 1 9 Tues. 9:04:54 Missed a cat Car Car Jun 5 Fri. 10:36:48 Missed gas tanke r pi<;kup, and minivan f ollowing M i n ivan Car each o t her. Jun. 7 Sun 1 9 :3 5:25 Separated a SUV with a U-1-laul into two vehicles Car Van Jun. 6 Mo n 8:49:A4 Sepa r a ted a sem ilrail cr Car Car 16:46:46 Separated a vehicle transporter into Mo vehicles Car Van jun 9 T u es 7:09:10 Counted a gas tan k er as t wo veh i cles Car Semi-trailer Jun. 13 Sat. 18: 0 1:53 Separated a SUV p\llling a car Minivar1 Car w/ trailer J un. 1 4 Sun. 21:14 :31 M i ssed a gas t anker P ickup suv J un. 2 1 Sun. 19:22 :52 Separat ed a SUV with a trailer into two vehi cles M ini Van Car 6.0 SUMMARY OF FINDINGS The Phase I fie ld resul t s showed 99.72 percent vehicle separation accu racy of ldris with 1 ,761 veh i cle passes and 5/driserrors. I n Phase II a total of45,282 vehicles passes were verified with 49 ldris errors (99.89 percent overall accuracy). A l gorithm modifications by Peek Traffic, Inc. occurred o n February 27 and agai n on April 12, 1998 leadi ng to improved results after each modification For examp l e, for the evaluation p er i od aft er April 12 ldris0 separat i on accuracy was 99.92 percent. Both tests indicated that ldris had m i nor problems separating trucks with veh i cle t ranspor t i ng trailers, however, the h i g hest separat i on acc u racy was att a i ned during sampli n g periods tha t included greater volume and mix of ve h icles. On 40 out of 62 days, no e rrors were reported, and fu rt her, on two 8-day periods no errors wer e reco rded. Throughout the entire Phase II evaluatio n period, errors occurred on a verage only once every 1.55 days. 17 ______________________ _.


A PPENDIX A ldris0 Product Information


Table A 1 ldris 0 Product Information ... --.. .. ..,:. _,: . r,L>. f: ::-. : ... k '' .' :. ',_, __ ';A', Qgr,.'; \:'",.' ;, ,' t r,.... :.> l: ---:-._:___ Separation Accuracy 99.89/o Un i t Cost $3,500/L.ane (Separat ion Only) I nstalla t ion Cost Approx. $1-2000/Lane Installation Time ll Day/Lane Mai ntenance Cos t < $500/year Mean Time Between Failures > 25,000 H ours N otes: 1 SourceP eek Sarasota 2. Phase II tests con duc ted at Leesbu r g P l aza by CUTR. A total o f 4 5 ,282 veh i cles were ver ified w ith 49 ldrisf!.) errors. ldris0 System Flow Chart D D Loo p Opt ional Discrete Outp uts from I r-. Inductive RS2 l oop .,; detector I 3'' w ide Card The ldris Smart Loops system consists of: I

Capacity: 512K by tes of ROM 1 GB Hard Dr i ve -4M DRAM 4 One 3" card rack Note: Power supply and UPS to be supplied by integrator Interfaces The ldris da t a i nterface is an RS232 seri al communications link at 57600 ba ud to th e lan e controller using standard ASCII fixed format messages. Hand shaking i s by hard ware control l ines. The ldrisSmart Loops message in terface i n cludes : Trigge r -Speed only Per-Vehicle Record Speed, t ime o f day, length, class, pos ition in lane (mu l tilane), loop occupancy, direction. Class message -Time of day, classf or de l ayed message due to link fai lu re. Remote Boot Reboo t ldris Smart Loops Restart ldris -Restart algorithm Configure outstationConfiguration message An optional digital trigger output is ava il able fo r VES systems, which requ i r e a discrete trigger only3 l Source Peek Traffic


/dris81nt egration on a Lane Controller The JdrisSmart Loop a lg o rithm is written in ANSI-c and runs on QNX whi c h is a POSIX compliant op e rating s ystem. ldris can therefore be ported on to a ny s u itabl e lane co n t r o ll e r e l i minat ing the need for an extra mi cro-controll e r, s ignificant l y reducin g h a rdware requirements and maintenance costs while improving overall system r e liability. The integration costs are dependent upon the platfo r m runn ing on the lane controller. Optional Discrete Output$ from De t ec tors available D /\ Q Inductive loop RS-232 D '""" Output(() lane DO I RS232 > ConttoUer Loop array 3" wide run ning Cord R1lck ... ..:.I'IJ\ 1 + I l 8a&tt'l)' Baekup UPS (Supplied by integrator) ....... r --1 l... ......... ................. . .... . j


Related ldris Projects' E-470, Denver, CO. ldris was deployed as a vehicle separator, and VES trigger on 20 remote ramps on the E-470 toll road i n Denver, Colorado, i n l ate 1998 The system wa s deployed as an in tegr at ed system, with the a lgor ithms run ni ng on the same hardware as the lan e control p rocess. Delaware DOT Subsequent to the CUTR tests, Jdris"' technology has been installed by Delaware DOT to prov ide vehicle classification at Bidd ies Plaza in N ovember of 1999 T he insta ll ation is in a fourlane exp ress configura t i on (two lane s nor thbound, two l anes south bou nd with sho u l ders). Follow ing system commissioning, accuracy l e v els in exce s s of 99.5% fo r axlebased vehicle classification are being achieved. Treasure Island, Florida ldrishas also been in stalled as part o f a pre-c l assification system at Treasure I s l a nd in Florida. T h i s insta llation r equires the ldris system to pre-classify the ve hicles prior to arriva l at the payment point and provide a t readle emulation to th e ex isti ng l ane con t rol equipment in order to minimize s y s tem integration cos t s Separation and cl assificat i on accuracy in excess of 99 p e rc e n t has b een achieved in this challenging pre-classification envi r onment'. Othe r T ests ldris Smart loop Tri a l s were conducted at th e New J ersey Turnpike by t h e vendor to test veh icle classification and d ual tire detection in a lan e confi guration simil a r to the lane at Leesburg P l aza. A total of 1 6 h ours of video was analyzed, con taini ng 1807 veh i cles (4958 ax les). Vehicles with more than 2 axl e s comprised 27% of the vehicl e sampl e. The Jdris system achieve d 99 .7% classification accuracy, a nd 99 9% a ccu racy on axle detectio n Dua l Tire detection was greate r tha n 84%. Test S i tes (Exist ing and/or Planned)' ldris will also b e subject t o exte nd ed tria l s i n express -l ane applicat i ons in Illinois and Orlando in earl y 2000. 'Source-Peek Trame Facsimil e Message to CUTR dated Deeember2 1 1999.


Center for Urban Transportation Research College of Engineer ing Univers ity of South Florida 4202 E. Fowler Aven u e, CUT 100 Tampa, F lorida 33620-5375 (813) 974-3120, fax (813) 974-5168 http: //www .cU1r. eng .usf. edu