Avalable onlne at www.scencedrect.com ScenceDrect Proceda - Socal and Behavoral Scences 96 ( 2013 ) 2404 2413 13th COTA Internatonal Conference of Transportaton Professonals (CICTP 2013) A traffc emsson-savng sgnal tmng model for urban solated ntersectons Rao Qan, Zhang Lun*, Yang Wenchen, Zhang Meng The Key Laboratory of Road and Traffc Engneerng, Mnstry of Educaton, Tong Unversty, Shangha 201804, Chna Abstract gnorng traffc envronmental beneft, a traffc emsson-savng traffc sgnal tmng model for urban solated ntersectons s presented. Frstly, wth dfferent statuses of vehcles on the road, for example movng wth a constant speed, slowng down speed, dlng speed or an ncreasng speed, there are dfferent knds of degree of contamnaton. Based on whch the urban road pollutant emssons model, and the crtera pollutant emssons model are establshed. Secondly, n order to analyze the dependence of the traffc sgnal evaluaton ndexes, the qualtatve analyss and the quanttatve analyss based on the numercal statstcs are adapted. Also, based on the selectng prncple of evaluaton ndex, selected performance ndcators for the emsson factors, and taken them nto consderaton whle establshng the traffc sgnal tmng model based on relatve evaluaton ndex system. Then, an mproved real-coded genetc algorthm to solve the traffc sgnal tmng model s presented. Lastly, the three algorthms are proved by a great deal of numercal calculaton. The result shows that the presented algorthm has a hgh precson whle solvng the models, and has a very good effect on reducng emssons and the effcent of controllng the traffc roads. 2013 The Authors. Publshed by Elsever by Elsever Ltd. B.V. Open access under CC BY-NC-ND lcense. Selecton and/or peer-revew under under responsblty responsblty of Chnese of Chnese Overseas Overseas Transportaton Transportaton Assocaton Assocaton (COTA). (COTA). Keywords:Intellgent transportaton; Traffc sgnal; Emsson model; Tmng model; Genetc Algorthm 1. Introducton Wth the envronmental problems caused by transportaton beng more and more severe, urban traffc congeston, road safety and polluton have become the common ssues that large ctes n Chna confront. Ratonal * Correspondng author. Tel.: +86 13788908005. E-mal address: Lun_zhang@tong.edu.cn 1877-0428 2013 The Authors. Publshed by Elsever Ltd. Open access under CC BY-NC-ND lcense. Selecton and peer-revew under responsblty of Chnese Overseas Transportaton Assocaton (COTA). do:.16/.sbspro.2013.08.269
Rao Qan et al. / Proceda - Socal and Behavoral Scences 96 ( 2013 ) 2404 2413 2405 organzaton of sgnal control n solated ntersectons can mprove traffc effcency, reduce traffc congeston and exhaust emssons, whch s the key to solve urban traffc problems. Optmzaton of ntersecton sgnal tmng theory was ntated n the 1950s, Webster was the frst to ntroduce the method to optmze ntersecton sgnal tmng targeted at shortest delay, the F-B method. On the bass of the F- evaluate the effect of sgnal tmng optmzaton (1981). Robertson found the relevance between the delay, stops, and fuel consumpton. In the TRANSYT system bult by hm, fuel consumpton s regarded as a benchmark, and the drect operatng expenses are regarded as the obectve functon (1980). Based on the research above, besdes delay and stops, capacty was added nto the obectve functon as a performance ndcator whle consderng the actual stuaton of the urban road traffc n Chna (Gu, 1998). However, the traffc sgnal control researches lsted above are all based on the average delay, number of queung vehcles, stops, ntersecton saturaton degree and capacty, etc., emssons was not ncluded n the traffc sgnal control metrcs. Zhou establshed emssons-savng b-level programmng model targeted at emsson control, but the model lacks the mcro-estmaton ablty (2008, 2009). A set of emsson measurement system based on the real-tme operaton of the vehcle was developed to analyss the mpact of the speed and acceleraton parameters on the emsson factors and fuel consumpton, but the concept and sgnfcance of the model were not presented (Ren, 2003). Ma focused on the fundamental problem of cycle length optmzaton based on the vehcle movement n ntersecton area, and researched the applcaton of the optmzed cycle length by ntroducng the concept of "margnal beneft" and "margnal cost", however t was only the data lst wthout establshng a detaled model ncludng emssons of each stage (20). In summary, exstng researches on traffc sgnal control consderng emsson factors are n the prelmnary stage. They are lack of the systematc research on the standardzaton of the ntersecton emsson model. On the foundaton of the tradtonal prncple of traffc sgnal control based on the statstc model, ths paper focuses on the traffc sgnal control method consderng emsson factors n urban ntersectons, am of ntroduce traffc emsson factors, establsh the traffc sgnal tmng model consderng envronmental protecton ssues, and propose the mproved Real number codng genetc algorthm model. Verfy the utlty of the model and arthmetc through an example of a typcal solated ntersecton. The remander of ths paper s organzed as follows. Secton establshes the model of the exhaust emsson accordng to the vehcle travel process, as well as the crtera pollutant model through ntroducng equvalent value of pollutant emsson. Secton establshes the dmensonless traffc sgnal tmng model by analyss the performance ndex. Secton presents the mproved genetc algorthm. Secton evaluates the proposed emsson-savng sgnal tmng model wth actual ntersecton data. Secton conclude the work. 2. Standard pollutant dscharge modelng. 2.1. The model of the ntersecton vehcle emssons process crcumstances, vehcle moves n a constant speed, when there s an ntersecton, vehcles wll frst decelerate untl cease and then accelerate to the ntal speed. E, ncludng the drvng emsson E cl E I, accelerated emsson E a, reducton emsson E d, and non- E ca on the approach. Establsh the model respectvely as follows, therento, represents the phase, represents the speces of pollutants. In addton, emsson factor was ntroduced n natonal standard GB5181- emssons of a pollutant under the nfluence of varous factors.
2406 Rao Qan et al. / Proceda - Socal and Behavoral Scences 96 ( 2013 ) 2404 2413 V 6 5 4 N L 7 8 9 LE 3 2 1 t Constant speed decelerate accelerate Idle speed 1112 Fg. 1. at the ntersecton Fg. 2. at the ntersecton Drvng emssons on the road Accordng to the defnton of standard v PCU EF (g/(pcu km)), drvng PCU emsson on the road s the product of the road vehcle number V (pcu/h), drvng emsson factor and the road length L (km), n whch vehcles wth a constant speed. PCU PCU Ecl V EF L (1) Emssons on the approach A. dlng emsson PCU EFI (g/(pcu h)), the dlng emssons of the parkng vehcle on the approach, s the product of stops, dlng emsson factors and dle tme. Generally recognzng, the dle tme s the average delay d (s)of the approach. Stops s the product of the arrved h. 1 PCU PCU EI V h EFI d (2) 3600 B. Accelerate and decelerate emsson PCU PCU Accordng to the de e a and e d (g/pcu) accelerate and decelerate emsson factors( pcu/h). PCU PCU PCU PCU E V, a h ea Ed V h e (3) d C. Drvng emsson on the approach Drvng emsson on the approach s the product of the Non-parkng vehcle number, drvng emsson factor and the approach length L E(km): PCU PCU E V (1 ) ca h EF L E (4) Summng up the above, establsh the model of pollutant n phase as Formula 5: PCU PCU 1 PCU PCU PCU PCU E V ( EF L EFI d h ( ea ed ) h EF LE (1 h )) (5) 3600 PCU PCU e a + e d, equals to emsson of the dle speed n0s (Ma, 20). Therefore, the phase emsson could be smplfed as Formula 6. PCU PCU 1 PCU 1 PCU PCU E V ( EF L EFI d h EFI h EF LE (1 h )) (6) 3600 36 Here, E s the emsson of pollutant n phase n the ntersecton, g/h; The sum of the pollutants n all phases n the ntersecton s as follows. n E E (7) 1
Rao Qan et al. / Proceda - Socal and Behavoral Scences 96 ( 2013 ) 2404 2413 2407 Here, n s the number of phases n a cycle. 2.2. Emsson model of standard pollutant There are many knds of emssons of pollutants from the vehcles exhaust n the ntersecton. Ths paper manly research on three pollutants ncludng CO, CH, NO. Dfferent pollutants match dfferent order and cause dfferent damage degree. So the emsson of the pollutants n the ntersecton does not equal to the sum of these three pollutants. Ths paper ntroduces standard pollutant, E S (g), whch means degree of the sum of pollutants n the ntersecto emsson, shown as Formula 8. ES we 1 CO w2ehc w3e NO (8) Here, E CO, E HC, E NO represent the emsson of the pollutants CO,CH, NO n the ntersecton respectvely; w 1, w 2, w 3 represent the weght coeffcent of the three pollutants respectvely, and Weght determnaton The equvalent value of pollutant emsson E ev, 3 k 1 w 1, w [0,1]. Consderng the damage degree, bologcal toxcty and the cost of the pollutants, accordng to Measures for the admnstraton of natonal pollutant dscharge fee collecton standards (2003), the equvalent values of each pollutant are shown n Table 1. Where, the weght of HC s averaged by the equvalent value of the pollutants of the matter. Table 1 The equvalent value of the pollutants (kg) k k CO CH NO 16.7 5.1 0.95 The equvalent value of pollutants s standard reference data for collectng cost of pollutants. Accordng to state regulatons, the cost of every pollutant per equvalent s the same. So, n terms of economc benefts of every pollutant, the weghts are determned as the equvalent values of per 1kg pollutant each. And accordng to the defnton of the standard pollutant, the weght of the three knds of pollutants needs normalzaton. The equvalent weght of pollutant s calculated as Formula 9. Emsson of the pollutant The equvalent weght of a pollutant (9) Equvalent value of the pollutant Accordng to (9), the weghts of pollutants are calculated as Formula. () 1 Eev w 3 1 1 Eev By calculatng, the weghts of pollutants CO, CH, NO are 0.046, 0.15 and 0.804 respectvely. The emsson model of standard pollutant s shown as Formula 11. (11) ES 0.046ECO 0.15EHC 0.804ENO
2408 Rao Qan et al. / Proceda - Socal and Behavoral Scences 96 ( 2013 ) 2404 2413 3. The sgnal tmng model based on emsson 3.1. analyss of performance ndex Exstng research have found that the Capacty (Saturaton) has the most mportant relatonshp wth the Cycle, then the Queue length, Delays and Stops, based on a survey on the nterdependence coeffcents between the Cycle of typcal ntersecton and some assocated factors. Research results show that the Saturaton of the road s proportonal to the Capacty, the same wth the Queung length wth the Stops and the Delay wth the Stops. In possble together wth the modelng n consderaton of the envronment. In tradtonal traffc control, the target functons are usually bult up wth the Delay, Stops and Capacty. In order to buld the functon of tmng model takng the traffc envronment nto consderaton, average Delay, Capacty and the standard pollutant emsson are chosen as the performance evaluaton ndex of adapted traffc control stratagem. Accordng to applcablty of dfferent delay model n ntersectons wth dfferent Saturaton, the research and practce shows that the WEBSTER model ft the current stuaton of traffc n our country well. So the calculaton of the standard pollutant emsson s done by the model presented n ths paper, wth adapted WEBSTER model as the calculaton model of the delay performance, as Formula 12 2 2 ( C g) ( C g) qc d (12) 2 C(1 y) 2Cq 2 sg ( sg qc ) Here, d s the average delay n phase, g s the effectve green tme n phase, q s the arrved traffc flow n phase, s s the capacty of the ntersecton n phase, y s the traffc ntensty n phase, C s the perod. The total delay of the vehcles n the ntersecton of every perod s: dqc. Capacty s calculated by the model n HCM2000 as Formula 13 g Q S u; Q S C (13) 3.2. Sgnal tmng modelng Whle buldng the target functons wth the absolute value of the performance ndexes, because of the dsuntes of dfferent dmensons of dfferent ndexes, and f one of the ndex calculatons occupes the monopoly -obectve to a sngle obectve. Hence, n order to avod the dfferent magntude of performance ndexes, smple the weghted calculaton and make the target functon non-dmensonal, ths paper ntroduces the relatve ndex modelng of performance ndexes by comparng the absolute value wth a standard value of the evaluaton ndex. delay and capacty. So ths research based on the Webster tmng plan (TRRL) the correspondng evaluaton ndexes as crtera, and then establsh the evaluaton system of the relatve ndexes. The traffc emsson-savng sgnal tmng model as Formula 14: max CPI k1 (1 avgd avge Q ) k2 (1 ) k3 ( avgdtrrl avgetrrl QTRRL 1) st.. gmn g gmax g L C (14) Cmn C Cmax 0 kn k 1 1 n
Rao Qan et al. / Proceda - Socal and Behavoral Scences 96 ( 2013 ) 2404 2413 2409 Here, CPI s the comprehensve performance ndex of the traffc sgnal control based on the emssons. avgd, avge and Q are the average delay tme, average emsson of standard pollutant, traffc capacty respectvely. avgd TRRL, avge TRRL and Q TRRL are the standard values of parameters calculated by Webster tmng plan.. g s the effectve green tme n phase. C s the perod. g mn and g max are the mnmum and maxmum tme of the green lghts. C max s the maxmum perod. L s the total waste tme of the ntersecton. k 1 k 2 k 3 are the weght coeffcent of delay, stop rate and traffc throughput respectvely, wth adaptve adustment by traffc demand varaton to meet the optmzaton obectves of dfferent traffc condton at ntersectons; Three weghted coeffcent values are calculated by Formula 15. 1 Y X U1, U2 1 Y, U3 X 1 Y U 15 k 3 U 1 In Formula 15, U s the temporary weghts of ndex, Y s the total flow rate of ntersectons; X, the weght correcton factor, s the total saturaton of ntersectons. k 1 k 2 decrease wth the ncrease of flow rate rato, k 3 ncreases wth the ncrease of flow rate rato, whch makes the control obectves focus on reducng delays n dle state, and let the traffc smooth and safe; In smooth condton, the control obectves focus on reducng delays and stops, and makes the operaton at ntersectons effcent; In the busy and congeston state, the control obectves focus on mprovng the throughput, and maxmzes the ntersecton management effcency. Thus, the weghted models acheve the traffc control strateges optmzaton under dfferent traffc status. 4. Soluton for the model Traffc sgnal tmng model s a model of nonlnear, mult-obectve and optmzaton. As wth too many model parameters and constrants, the tradtonal genetc algorthm n solvng optmzaton s slow to reach the whle the algorthm s convergence. A GAbased heurstc (AARGA) s method has been developed to yeld approxmate solutons for each control nterval durng the entre optmzaton perod (Yang, 2009). The AARGA generates the ntal populaton and ensures the qualty of the ntal populaton by punshment mechansm based on the sort of classfcaton. Meanwhle, cross rate and mutaton rate s adaptve accordng to generaton and ftness value. The process of mproved AARGA are as follows: (1) Intalzaton: generate ntal populaton, whch meet the sgnal constrants and prncple of ndvduals dfferent from each other, and ntalze GA parameters, whle, populaton sze M=150, evolutonal generaton Gen=0, maxmum cross rate maxp c=0.9, mnmum cross rate mnp c=0.5, maxmum mutaton rate maxp m=0.1, mnmum mutaton rate mnp m=0.01. (2) Ftness Evaluaton: Ftness evaluaton adopted the comprehensve performance ndex CPI of emssonsavng sgnal control. (3) Selecton: roulette wheel selecton. (4) Crossng: non-unform arthmetc crossover operator. (5) Mutaton: non-unform mutaton operator, by whch the degree of mutaton s adaptvely adusted wth generaton and ftness value. (6) Elte retenton strategy: replace the worst ndvdual n current-generaton wth the best ndvdual n parent-generaton. (7) Judgment of termnaton prncple: f n<gen, go to Step2. Otherwse, output the best soluton and the value of evaluaton ndces.
24 Rao Qan et al. / Proceda - Socal and Behavoral Scences 96 ( 2013 ) 2404 2413 5. Case study 5.1. Case narraton Shenzhen Lanhua- Xnzhou sgnal control ntersecton s a key ntersecton n Shenzhen, where the traffc s busy and each flow dstrbuton s sgnfcantly dfferent and characterzed by uneven arrval. The control effect of the exstng tmng model s poor durng the rush hours, hence t s the traffc black spot n Shenzhen. Currently, restrcton of left turn n the traffc control method s adapted for the transportaton gude. Ths paper selects such ntersecton as research obect, to valdate the proposed tmng model and optmzaton algorthm. The plane geometry layout of the ntersecton, road traffc organzaton and phase structure are shown n fgure 3, wdenng the approach way of the ntersecton; channelzng the rght turn traffc, whch wll not controlled by the traffc sgnal; whle adaptng standard four phase structure. Fg. 3. (a)lanhua- Xnzhou Road organzaton dagram; (b) Lanhua- Xnzhou phase dagram Adaptng the mean value of the measured traffc flow of the ntersecton nducton col n dfferent status wthn one month for the research (2008.03.01-2008.03.31), the statstcal results are shown n the Table 2. Table 2 Intersecton statstcal average flow Phase name A B C D Included traffc ES*,WS EL,WL NS,SS NL,SL Idle traffc volume q (veh / h) 126,76 52,64 271,325 28,120 Smooth traffc q (veh / h) 604,494 404,317 17,1389 236,530 Busy traffc q (veh / h) 822,642 513,384 1619,1764 252,529 Congested traffc q (veh / h) 945,738 589,441 1861,2028 290,608 Phase Saturaton flow S*(veh / h) 6600,6600 30,30 6600,6600 30,30 Saturaton lmts 0.95 0.95 0.95 0.95
Rao Qan et al. / Proceda - Socal and Behavoral Scences 96 ( 2013 ) 2404 2413 2411 Note: S * represents the saturaton flow, by addng the correspondng lane saturated flow; ES: Go straght from east. EL: Turn left from east. Traffc sgnal control plan s affected by the parameters lke geometry, tmng and traffc flow, etc. Green lght nterval, as well as other statc tmng parameters value accordng to the traffc flow characterstcs n varous state, phase loss tme G and other parameters are set as shown n Table 3 (Yang, 2009): Table 3 Threshold values of tmng parameters Traffc state G(s) y O xp gmn(s) gmax(s) Cmn(s) Cmax(s) Idle 4 [0 0.42] 0.35 0.95 45 56 120 Smooth 5 (0.42 0.54] 0.35 0.95 12 60 60 180 Busy 6 (0.54 0.80] 0.35 0.95 15 75 84 250 Congeston 6 (0.80 0.95] 0.35 0.95 15 90 84 300 Due to the ntersecton n ths study s an urban ntersecton, based on the results of natonal natural scence fund proect of southeast unversty, Urban transport system energy consumpton and envronmental mpact PCU PCU analyss method (Wang, 2002). Take the data of EF and EFI when the speed of vehcle s 40km/h. CO HC NOx EF PCU (g/(veh h)) 44.27 5.12 2.01 EFI PCU (g/(veh h)) 640.76 72.07 7.34 Accordng to mult obectve weghts determnaton method based on fuzzy evaluaton, the weght values of delay, standard pollutant emsson, and traffc capacty n four states are as shown n Table 4: Table 4 Weght table of nstance performance ndcators n each state Traffc state D SE Q Idle 0.88 0.12 0 Smooth 0.71 0.29 0 Busy 0.42 0.35 0.23 Congested 0. 0.22 0.68 5.2. Results Analyss Performance Analyss Evolutonary process of AARGA and RGA algorthm are shown n Fgure 4. Compared to the RGA algorthm (Chen, 2008), n the course of evoluton, the ftness value of AARGA algorthm after runnng 50 generatons tends to be stable, wth less average teratons, fast convergence speed, and overall ftness value s better than RGA algorthm. In the 50 tmes ndependent numercal calculatons, a feasble soluton can be obtaned n AARGA algorthm each tme, and wth more tmes of the convergence, small ftness value fluctuatons, suggestng that n ths paper, the presented AARGA algorthm for ths research nstance has a good applcaton effect. Moreover, even n congested traffc condton, the sngle runnng tme of AARGA algorthm s usually no more than 180s, can meet the real tme demand of the control system.
2412 Rao Qan et al. / Proceda - Socal and Behavoral Scences 96 ( 2013 ) 2404 2413 Fg. 4. AARGA and RGA algorthm evolutonary process dagram and stablty Control effect Under dfferent traffc states, Contrast the best tmng scheme of the AARGA and RGA and classc Webster (TRRL) algorthm tmng scheme. Statstcal evaluaton numercal calculaton results are shown n Table 5: Table 5 Performance ndcators calculated results of AARGA, the RGA and TRRL algorthm Traffc state algorthm C(s) d(s) E S(g) Q(veh/h) g1(s) g2(s) g3(s) g4(s) Idle Smooth Busy Congested AARGA RGA TRRL AARGA RGA TRRL AARGA RGA TRRL AARGA RGA TRRL 56 56 63 118 122 7 148 158 156 223 239 250 19.63 19.63 20.71 43.52 46.87 53.48 64.55 67.78 71.37 85.24 88.59 90.58 6.77 6.77 9.74 13.73 17.83 19.87 43.92 51.90 58.88 68.91 70.90 73.89 Note: C s the cycle. g1-g4 s the effectve green tme of the phase 1-4. d s the average stop delay. E S s the average equvalent value of standard pollutant. Q s the traffc capacty. As can be seen from Table 5, the control effect of AARGA algorthm s better, whch can reduce control delays and emssons, and ncrease the ntersecton traffc capacty. Take the contrast of AARGA algorthm and TRRL algorthm as an example, AARGA algorthm mproves n the delays, emssons and capacty wth about 1% to 15% mprovement. In the congested state, AARGA algorthm can effectvely mprove the capacty, but stll lackng of emssons and delay mprovement, however ths depends on the control obectves, and the volatlty s low, therefore, the method n the paper s reasonable and effectve. 6. Concluson A traffc emsson-savng sgnal tmng model consderng the operaton characterstcs of vehcles n urban solated ntersectons s presented n ths paper. Frstly, the urban road ntersectons emssons model of vehcles s establshed. Secondly, the crtera pollutant emssons model s establshed. Thrdly, traffc sgnal tmng model 3473 3473 3702 4160 4185 3751 4327 4275 4199 4559 4491 45 16 17 14 24 27 23 34 36 39 19 19 18 27 23 29 45 48 51 15 39 40 28 46 48 48 74 80 83 12 24 26 23 27 29 32 46 51 53
Rao Qan et al. / Proceda - Socal and Behavoral Scences 96 ( 2013 ) 2404 2413 2413 consderng emssons factors, based on traffc sgnals wth relatve evaluaton ndex system s establshed. Then, real-coded genetc algorthm s mproved to solve tmng model whle desgnng. Fnally, a test s conducted n the typcal ntersecton, n the non-peak and peak traffc scenaros, usng three knds of control strategy for numercal calculaton. The expermental results are shown as below. Compared wth RGA and TRRL control methods, on the performance of the algorthm, AAGRA algorthm has mproved qualty and hgher calculaton effcency. On the control effect, AARGA algorthm can effectvely reduce the ntersecton delay and the stops and ncrease travel speed, whch s consstent wth the actual traffc management obectves. However, ths research only used a numercal calculaton method to evaluate the effectveness of the model and algorthm. The proposed method wll be verfed by Vssm mcroscopc smulaton, and the traffc sgnal control mechansm road whch under the crcumstance of collaboratve envronment for emsson wll be researched n the future. Acknowledgement The work of ths study s supported by Natonal Scence Foundaton of Chna (Proect No. 50408034) and Shangha Educatonal Foundaton for Innovaton (Proect No. 11ZZ27). Reference Transport and Road Research Laboratory. Berkshre, UK. GU, H.Z., WANG Journal of Southeast Unversty. 03:70-74. Fourth Internatonal Conference on Natural Computaton: 530 534. Doctor degree thess, Wuhan Unversty of Technology, Wuhan, Chna. REN Journal Of XngTa Vocatonal And Techncal College, 20(3): 57 58. Traffc and Transportaton Studes. ASCE. State Development Plannng Commsson of the people's Republc of Chna., Mnstry of fnance of the people's Republc of Chna., State Envronmental Protecton Admnstraton of the people's Republc of Chna., State Economc and Trade Commsson of the people's Republc Orders. Doctor degree thess, Jln Unversty, Jln, Chna. - IEEE trans. On Evolutonary Computaton, 12(1):64-79. http://www.partcleswarm.nfo/programs.html Scence Press. - The nnth nternatonal conference of Chnese transportaton professonals, Harbn, Chna: 19-1918. -tme traffc sgnal optmzed control and smulaton for urban Computer Engneerng and Applcatons, 46(33): 239-243. Yang, W.C., Zhang, L., He, Z.C., Zhuang -stage Fuzzy Control for Urban Traffc Sgnals at Isolated Intersecton and Paramcs Smulaton. //Proceedngs of The 15thInternatonal IEEE Conference on Intellgent Transportaton Systems (ITSC 2012) Anchorage, Alaska, USA, Sep16-19, 2012, pp: 391-396. Chen, X.F., Sh Z.K. (2008) Novel adaptve penalty strategy and ts applcaton n traffc sgnal tmngs optmzaton. Computer Engneer ng and Applcatons, 44 ( 26) : 5-7.