An innovative dynamic bus lane system and its simulation-based performance investigation

Transcription

1 An innovative dynamic bs lane system and its simlation-based performance investigation The MIT Faclty has made this article openly available. Please share how this access benefits yo. Yor story matters. Citation As Pblished Pblisher Hong Yang, and Wei Wang. An Innovative Dynamic Bs Lane System and Its Simlation-based Performance Investigation. IEEE, Copyright 29 IEEE Institte of Electrical and Electronics Engineers (IEEE) Version Final pblished version Accessed Sat Mar 3 4:23:34 EDT 219 Citable Link Terms of Use Detailed Terms Article is made available in accordance with the pblisher's policy and may be sbject to US copyright law. Please refer to the pblisher's site for terms of se.

2 An Innovative Dynamic Bs Lane System and Its Simlation-based Performance Investigation Hong Yang Civil and Environmental Engineering Rtgers, The State University ofnew Jersey Piscataway, NJ Wei Wang Civil and Environmental Engineering Massachsetts Institte oftechnology Cambridge, MA Abstract-The strategy of exempting bs from other traffic throgh exclsive bs lanes (XBL) is prevalent. Rather than jst deploying the XBL system, in this stdy, a new innovative dynamic bs lane (DBL) operation system which is expected to be more prodctive is initially introdced. Unlike conventional stdies which are mainly focsed on operational performance of bses, this stdy attempts to examine the inflences on bses and adjacent traffic following the provision of both XBL and the new system. Impact measrements in terms of travel time and traffic conflicts changes are sed as assessment indicators. Comparisons with the exclsive bs lane system as well as the mixed traffic flow are analyzed and highlighted throgh the application of micro-simlation approach. Simlation reslts showed that both XBL and DBL have positive impact on bses and negative impact on adjacent traffic. Traffic conflicts freqency will increase if either bs lane system is applied. Also it is fond that DBL performs better than XBL in terms of achieving the goal of bs operation improvement while limited the performance deterioration to other vehicles and conflicts risks in a relative lower magnitde. Reslts also sggest the necessity of selecting the optimal strategies according to the tradeoff between the operational performance and safety performance nder different traffic conditions. I. INTRODUCTION Traffic congestion is a common phenomenon that experienced in many rban areas arond the world de to the increase in travel and nmber ofvehicles. This isse can be primarily attribted to the limited transportation system spply that does not satisfy the increasing traffic demand. Simply, the road space increasing is nable to keep p the rapid growth of vehiclar traffic. Land-se and travel behavior change may also encorage the travel mode shifting from pblic transport towards private traffic. As intensive agmentation and enhancement of transportation infrastrctres spplement sally will be neconomical, alternative soltions to the congestion isse is expected. Realizing the potentials of existing resorces can be a practical way to achieve the goal of efficiently delivering more passengers and goods. Among the measres that have been identified and implemented in field, innovative bs priority systems are regarded as desirable treatments to enhance bs attractiveness and promote the shift of travel mode from cars to bses, which in tm will finally reslt in relaxation or disappearance oftraffic jam. One of the most common priority systems so far is the application of exclsive bs lane (XBL). By allocating a reserved lane on arterials to exempt bses from other private traffic, bs service qality in terms of travel speed, travel time reliability, etc, are anticipated to be well secred by XBL. Operation performances of XBL have been widely reported based on the feedback from practices. However, the impact of provision of XBL on adjacent traffic is not always reported at the same level of detail. For an optimal system, it is necessary to establish the balance of the entire system based on a completely investigation of its performance and impact. Bt crrent major nderstanding ofdeploying XBL system still limited the scope within its operation performance. There is still no adeqate nderstanding of its performance other than the operational aspect. Since the system is primarily designed with a preference of bs operation, intitively, it navoidably has negative impact on the qality ofother adjacent traffic. To realize the fll potentials of XBL, it is necessary to display its advantages as well as disadvantages for better system design. In the meanwhile, identifying the distinctive featres of XBL is also beneficial for making improvement or seeking better system withot great effort. Therefore, the specific objectives concerned in this stdy are mainly focsed on twofold: (1) Other than operational performance of bs, the impact of provision of XBL on adjacent traffic as well as system safety performance will also be investigated sing simlation model; (2) An innovative bs priority system based on the extension of XBL concept is introdced and its performance both in operation and safety aspects will be highlighted by comparing with XBL. The following sections of paper are seqentially organized as follows: A comprehensive review of related stdies and practices deploying exclsive bs lane is presented in section 2. Section 3 describes the strctre of newly proposed bs priority system. Section 4 illstrates the stdy approach and measres on investigating bs as well as /9/$ IEEE 15

3 adjacent traffic performance. Based on the rnning of experiments, test reslts and analysis are addressed in the fifth section. Finally, the findings are smmarized and ftre potential work is discssed. II. LITERATURE REVIEW The impact of provision exclsive bs lanes are freqently discssed in the literatre. Several stdies condcted before-and-after comparisons of XBL to identify the performances of XBL on rban networks in the United States, India, China as well as Soth Korea, etc. Both sccessfl and nsccessfl reslts were obtained after the implement of XBL. For instance, Erdman et al. (1976) performed an investigation to stdy the impact ofxbl on a two-directional roadway with two lanes in each direction in Baltimore metropolitan area. It was fond that the average travel time for major commters in the morning peak period sing bs lane rather than a passenger car was determined to be 5 percent longer than before. The athors conclded that the XBL was detrimental to both atomobile and bs movements in terms of travel time. Similarly, Sarin et al. (1983) evalated the XBL system, which was fitst introdced in Delhi, India, in The stdy revealed that the system failed to save travel time as a conseqence of ineffective enforcement. The system was discontined in The fitst XBL was implemented in Knming, China, in Wei et al. (22) compared its performance before and after 2 years operation. It was fond that the average speed of bs increased 68%, from 9.6km/h to 15.2 km/h. The corresponding impact on other vehicles was not reported. Choi et al. (1995) condcted their stdy in Soth Korea and arged that the XBL system was sccessfl. Bs travel time was significantly redced, a mode shift from car to bs was estimated to be more than 12%, and accident rates were redced. Similar positive impact of XBL in Soth Korea was frther illstrated by the stdy of Kim (23). His srvey showed that the XBL were sccessfl in improving average bs performance compared with other adjacent traffic in terms of relative speed changes. Rather than focsing on bs operation performance, Karim (23) evalated the effect of XBL on travel time of other modes sing floating car techniqe. It was fond that the mean travel time for other vehicles are significantly increased after the execting bs lane dring morning and evening peak hors. De to the long cycle of condcting before-and-after comparison, simlation method also has been sed in recent years as an alternative for the prpose of operational analysis ofxbl. Shalaby (1999) applied the TRANSYT-7F simlator to investigate the impact of provision of the reserved bs lanes on throgh bses and other traffic in an rban arterial in downtown Toronto, Canada. Crrie et al. (24) proposed a balanced framework for roadway space reallocation associated with transit priority. Simlation model was deployed to clarify the operational performance of transit priority measres. In their latest stdy [9], the methodology to assess trade-offs in the se of the limited road space in Melborne, Astralia for new bs and tram priority projects is introdced. Again, micro-simlation model was sed to examine road space reallocation impacts. Mori et al. (26) sing simlator NETSTREAM evalated the se of exclsive bs lanes on Nagoya-Seto Expressway which is the main expressway link from the Tomei Expressway in Japan. EBL rnning strategy dring the Expo 25 was sggested according to their findings. Arasan et al. (28) investigated the impact following the introdction of XBL on the highly heterogeneos traffic flow on rban roads sing the micro simlation model HETEROSIM. It is clear from the review of previos stdies that most of the efforts have been focsed on the operational performance ofbses either by before-and-after comparison or simlation test. Impacts of XBL on adjacent traffic operation as well as the safety performance to date have been limited, which is important to assist in nderstanding the essential benefits ofxbl. Since bs does not interact or conflict with other vehicles, its operational improvement is easily to be expected. However, the space taken away from mixed traffic wold create congestion or more conflicts de to lane availability constraints on other vehicles, which might lead to an inefficient and nsafe operation of the traffic. To jstify the efficiency and safety performance of bs priority system, it is necessary to make more efforts on comprehensively investigating their potential impact. Then improvements can be made accordingly. III. THE STRUCTURE OF DYNAMIC BUS LANE In this stdy, we propose a dynamic bs lane (DBL) operation system. The new concept of or dynamically assigned bs lane is designed as shown in the following Figre 1 where the rightmost lane is reserved for bses (yellow color) nder certain traffic sitations. When there is no bs coming, the crbside lane is open to all the traffic as shown in Figre 1(a). When a bs is detected, the downstream crbside lane (between intersections 11 and 12) is then temporarily restricted for bses only. Vehicles ahead of the bs which are already travelling on the reserved bs lane can keep flowing within the lane or it cold change lanes towards the other lanes on the left side. However, vehicles on the other lanes are restricted from accessing to the bs lane ahead of the bs. This is ensred by activating the warning signs along the midline to remind private vehicles on the left lane. The Figre l(b) demonstrates sch sitation as the other vehicles following bses have to change lane once the bs is detected approaching. One specific occasion is that for those right-tm vehicles, they are allowed to travel on the reserved bs lanes with bses. To implement the dynamic bs lane, the following techniqe components are reqired, sch as Atomatic Vehicle Location (AVL) system or loop detectors (to detect bs); vertical variable message signs (VMS) to disseminate rightmost lane sage information (to other vehicles); and the horizontal in-pavement lighting system as dynamic roadway markings to remind the drivers on adjacent lanes. The VMS cold be set at the pstream intersection approach lane. According to the link length, mltiple VMSs mst be installed along the link to ensre drives get the message and 16

4 have sfficient time to make their decision. It is assmed that all drivers wold obey the lane-changing rle once they see the posted VMS informationof lane restriction. As shown in Figre 1 (b), in this stdy, bs arrival is assmed to be detected by the loop detector at pstream link. Also it shold be noted that all the traffic signal control strategies are not modified to provide priority for bses. Compared to the XBL, the DBL is expected to achieve similar objective of improving bs service in a more efficient way. Deploying DBL system, roadway resorces are expectedto be allocated in a prodctive manner ~ ~,;;",o[;,,; ~ i : :::::ignal _VMS:lna:tr,E o Figre 1 Strctre ofthe dynamic bs lane IV. PERFORMANCE ASSESSMENT APPROACH Previos practices can only provide s some experience of XBL's operational performance. Bt there are still no specific stdies reporting some safety evalation hints abot the strategies. Also for the new proposed DBL system, it is important to determine its effect compared to the mixed traffic flow as well as XBL. To obtain a better nderstanding of the two strategies, this stdy applies a simlation-based assessment approach to highlight their performances. Paramics is selected as the micro-simlation tool to spport the stdy. The major reasons encorage s to apply this test platform are: First, cstomizable models can be obtained throgh the Application programming Interface (API), which is a significant advantage over most other simlators. Second, its potentials to be applied for safety evalation based on srrogate safety measres analysis [12]. Cstomized API which activates and deactivates the tilization of DBL is programmed to make the bs lanes behave in a dynamic way as stated in previos section. With the principal objective of qantifying the impacts of different strategies on traffic, two measrements are sed for the prpose: average travel times and traffic conflicts. The former one is freqently sed as conventional indicator to determine the operation performance, and the latter is applied to highlight the safety performance. The traffic conflicts in this stdy are identified by the modified time-tocollision (MTTC), which was introdced by Ozbay et al. (28). Detail description of the simlation-based safety analysis approach canbe referred in the stdy. To make the reslts comparable, different networks inclding mixed-traffic, exclsive bs lane and dynamic bs lane design are modeled for test. All the inflential parameters sch as driver behaviors, speed limits, and signal timing are assmed to be the same among the networks to avoid mixing their inflences with the operation strategies. Experiments ofdifferent test scenarios are listed in Table 1. Table 1 Design ofexperiments Variable Options Type ofsystem Mixed, XBL, DBL Link Length (m) 1m with 3 lanes Bs Volme (veh/h) 3,6,9,12,15,18 Non-Bs Volme 5,7,9,11,13,15, (veh/h) 17,19,21,23,25, 27,29,31,33 Considering the stochastic natre of the simlation model, mltiple rns mst be condcted in order to get statistically reliable reslts from the simlation experiments. The simlation rns of each scenario with different random seeds are identified sing the seqential approach. This statistical procedre aims at obtaining the mean J1 = E(X) of the selected measres of effectiveness (MOE) X, within a specified precision. If we estimated X sch that Ix -JlI/IJlI = r, then r is called the relative error ofx. The specific objective of this approach is to obtain an estimated /1 with a relative error of r and a confidence level of 1(1 - a) percent. Denote the half-length of the confidence interval by ~(n, a). Frther details abot the approach are presented as follows [13]: (1) Make an initial nmber of no replications of the simlation and set n = no then calclate initial (crde) estimates X(n) and S2 (n) from X 1'X2',xn, (2) Decide the size of allowable relative error r = Ix -pi/ipl., (3) Calclated the adjsted relative error r' = r/(i- r); (4) Decide the level ofsignificance a ; (5) Calclate the half-length of the confidence interval ~s2(n) 8(n,a) =1 -- n - 1,1 - a / 2 n,. (6) If o(n,a)/ix(n)i:o; i se X(n) as the point estimate for f1 and stop, else make one more replication and set n = n + I, then go back to step 2. This approach assmes identical, independent (lid) otcomes, bt they need not be normally distribted. Ths - 2 the estimates of X(n) and S (n) for the mean and variance, 17

5 as well as the estimation qality improved with the incremental iteration. A relative error of r =.5 and a confidence level of 95% are sed to examine the effectiveness ofreplications. In or case, 9 random seeds are fond to satisfy the reqirement, and the average reslts of each scenario are then sed for frther comparison analysis. v. RESULTS AND ANALYSIS The simlation otpts that measre the inflence of different bs lane operation strategies on operation performance and safety ofthe vehicles are analyzed for each simlated scenario. The analysis of variance (ANaYA) is sed to statistically test the differences among the strategies. The reslts are reported based on a significant level of p vale=o.5. Yehicles that traversed the looom stdy section are sed in the data aggregation. Figre 2 demonstrates an example of bs and car travel time changes nder different operation strategies. It is clearly to see that DBL and XBL almost eqivalently decrease the bs travel time compared to mixed traffic flow. And DBL performs better in terms of the impact on other vehicles' travel time increment. U) co E i= r- ID > ro < ~ (.J) :::J CO l.d -6- DSl --- XSl ~- MIX o (a) BUS=9 veh/h (b) BUS=9 veh/h travel time. In addition, when bs volme is lower than l2veh/h, DBL performs better as its impact is less serios than that of XBL. Once the bs volme is heavy, DBL is almost eqivalent to XBL. These reslts might be attribted to the fact that the adjacent vehicles will have few chances to tilize the crbside lane ifthe bs lane is freqently sed by bs. It is worth noting that the travel time ofbs is more stable regardless of increase in traffic flow. This is attribted to the exemption of bses from other traffic distrbance. And the travel time of adjacent traffic will increase with increase of traffic flow de to redction of accessible lanes. All these findings provides the hint that DBL cold be a better alternative to efficiently allocation of the roadway resorces. Table 2 Differences ofaverage travel time Comparis DDS Changes Scenario 1 Scenario 2 Scenario 3 Scenario 4 Scenario 5 Scenario 6 Bs=3 Bs=6 Bs=9 Bs=12 Bs=15 Bs=18 DBLvs.1\IDC Deaease -5.45% -5.8% -6.4% -5.96% -6.12% -6.29% (Bs) P-Vale 2.2e e e e e e-16 XBLvs.1\IDC Deaease -5.42% -6.18% -6.22% -6.12% -6.18% -6.14% (Bs) P-Vale 2.2e e e e e e-16 DBLvs.XlI. Increase -.3%.4%.19%.17%.7% -.16% (Bs) P-Vale DBLvs.1\IDC Increase.52% 1.4% 1.76% 2.1% 2.33% 2.38% (Car) P-Vale e e e e e-l XBLvs.1\IDC Increase 3.44% 3.24% 3.6% 2.86% 2.8% 2.68% (Car) P-Vale 7.516e e e-U 9.867e-U 7.773e-ll 3.3e-ll DBLvs.XlI. Deaease -2.81% -1.77% -1.25% -.82% -.46% -.29% (Car) P-Vale 3.585e-ll 3.515e e The average traffic conflicts rate for each vehicle is sed as a srrogate measre of safety. Figre 3 shows an example of demonstrating the different impact of the strategies. Briefly, both DBL and XBL will increase the conflict freqency significantly. (a)bus=9 veh/h U) co E i= r- ID > ro < ~ ro l.d...e.- DSl --- XSl --+- MIX o Figre 2 Travel time changes nder different operation strategies U -6- DSl :E >- --- XSl MIX (Y') a.. (/) ("\I U (b)bus=18 veh/h Table 2 smmarizes the average travel time of bs and other vehicles for all six simlated scenarios. The reslts indicated that both XBL and DBL are relatively beneficial since it reslts in shorter travel time ofbs compared to the mixed traffic operation. Also these two strategies almost achieve the similar level of bs travel time improvement as the p-vales show no significant differences thogh DBL yields slightly shorter travel time. Regarding the impact on adjacent traffic, the reslts in Table 2 indicate that either XBL or DBL will have negative inflence by increasing the U...e.- DSl :E >- -.:::t --- XSl MIX a.. (/) ("\I Figre 3 Traffic conflicts changes nder different operation strategies 18

6 Table 3 presents the freqency differences of traffic conflicts among vehicles. A closer look on the conflict difference shows that DBL performs better as it redcing vehicles' exposre to conflicts compared to XBL nder the case of lower bs volme. However, when bs volme increased, the differences in conflicts rate among vehicles on roadways approach to the same magnitde regardless of the type ofbs lane system imposed. This may be attribted to the fact that DBL act as XBL given the high freqency of bs arrival. In all scenarios the nmber of conflict rate increases with the increase of traffic flow. De to the redction of lane availability and increasing density, adjacent traffic will have fewer opportnities to change lanes to avoid some conflicts compared to all lanes accessible case. These reslts sggest that the probability of conflict risk is deteriorated when both XBL and DBL are deployed as an operation system on the street, bt DBL is more likely to yield fewer conflicts which may reslt in traffic crashes. Table 3 Average traffic conflicts rate difference Compari5om Chanp5 Scenario 1 Scenario 2 Scenario 3 Scenario 4 Scenario 5 ~nario6 B5=3 B5=6 B5=9 BU5=12 B5=15 B5=18 Increase 19.61~D 35.3% 4.17% 44.21% 47.3% 49.64% DBLv5.M1X P-Vale e-l 2.23e-l 8.26e-ll 433e-ll Increase 54.2% 53.89% 52.4% 52.29% 51.1% 5.7% XBLv5.M1X P-Vale 5.995e e e-ll 1.21e-ll 3.55e-ll 1.822e-ll DBL \'5. XlI. VI. Decrease % % -8.5% -5.27% -2.69% -33% P-Vale 1.251e-7 o.f.14q CONCLUSIONS AND FUTURE WORK This stdy sed a simlation approach to investigate changes in performance measres of bses and adjacent traffic following the deployment of different bs lane strategies. The following are the important contribtions and findings ofthis stdy: Previos stdies and practices provide some experience to nderstand the operational performance of execting exclsive bs lane. Bt not too many efforts have been done to qantify the impact of the specifically reserved bs lane system on other vehicles of adjacent lanes. Safety performance associated with the implementation ofthe system is still far beyond to be clarified. A new bs lane system named dynamic bs lane (DBL) is proposed and the simlation model is developed to test and compare the impact ofexclsive bs lane (XBL) and the new system. It has been fond throgh the stdy that both XBL and DBL is beneficial to significantly redce the bs travel time. However, they all navoidably have negative impact on adjacent traffic in terms of travel time increment. The magnitde of inflence increase with the increase of traffic flow. The analysis of changes in operational performance shows that DBL system will perform better given a relative lower bs arrival freqency. This wold be attribted to more efficient lane allocation system sing DBL. Safety performance of providing XBL and DBL is evalated sing traffic conflicts as a srrogate safety measre. This is the fitst attempt to highlight the potential impact of different bs priority strategies from safety aspect, which was sally neglected or ignored. The conflicts freqency is fond to get detonated with the increase of traffic flow. Also deploying DBL will yield less risks compared to the XBL system ifthere are still great opportnities for adjacent traffic sing the bs lanes. Optimal tradeoff among the operational performance of bses and cars as well as their safety performance need to be considered regardless of which type of bs lane system is to be implemented. A balanced decision cold be sggested given the traffic flow conditions (Le. bs volme, othertraffic flow) as discssed in the stdy. This stdy is attempted to determine the impact ofa new proposed bs lane system compared to conventional exclsive bs lane. Since it is a new concept and no practice experience is available, simplified simlation model is applied to demonstrate the capability of the sggested design. More calibration efforts ofthe model are expected to be discssed given the availability of field data in ftre. The impact of the system configration sch as nmber of lanes, bs stop, traffic signal, VMS and detector location, etc, deserves frther investigation. The athors are crrently involved in the stdy ofsch kind isses to frther highlight the featres ofthe bs priority systems. REFERENCES [1] Erdman, J. and Panska, E., "Exclsive Bs Lane Experiment," Traffic Engineering, Vol. 46, No.7, pp.28-33, [2] Sarin, S., Sarna, A., Sharfdin, and Sharme, B., "Experience with Bs Lanes nder Mixed Traffic Conditions," presented at Institte of Transportation Engineers 53rd Annal Meeting, Agst, [3] Lin Wei, and Tang Chong, "Theory and Practice ofbs Lane Operation in Knming", DISP, No.151, pp.68-72, 22. [4] Choi, D. and Choi, W., "Effects of an Exclsive Bs Lane for the Oversatrated Freeway in Korea," presented at Institte of Transportation Engineers 65th Annal Meeting, Agst [5] Kim, H. J., "Performance of Bs Lanes in Seol: Some Impacts and Sggestions", IATSS Research, Vo1.27, No.2, pp.36-45, 23. [6] Siti Nrbaya AB. Karim, "The Effect of Bs Lane on Travel Time of Other Modes Using Floating Car Method", Proceedings ofthe Eastern Asia Society for Transportation Stdies, Vo1.4, pp , October, 23. [7] Shalaby, S.A., "Simlating performance impacts ofbs lanes and spporting measres", Jornal of Transportation Engineering, Vol. 125, No.5, pp ,1999. [8] Crrie, G., Sarvi, M. and Yong, W., "A Comparative Approach to Balanced Road Space Allocation in Relation to Transit Priority", TRB Annal Meeting CD ROM, Paper Nmber ,

7 [9] Crrie, G., Sarvi, M. and Yong, W., "A New Approach to Evalating on-road Pblic Transport Priority Projects: Balancing The Demand for Limited Road-Space", Transportation, Vol. 34, No.4, pp ,27. [1] Mori, H., Kitaoka, H., and Teramoto, E., "Traffic Simlation for Predicting Traffic Sitations at Expo 25", R&D Review of Toyota CRDL, Vo1.41, No.4, pp.45-51, 26. [11] V. Thamizh Arasan and P. Vedagiri, "Bs Priority on Roads Carrying Heterogeneos Traffic: a Stdy Using Compter Simlation", Eropean Jornal oftransport and Infrastrctre Research, Vol. 8, Isse 1, pp.45-63, March, 28. [12]Ozbay, K., Yang, H., Bartin, B., and Mdigonda, S., "Derivation and Validation of New Simlation-Based Srrogate Safety Measre", Transportation Research Record (in Press), 28. [13] Law, A.M. and W.D. Kelton, Simlation Modeling and Analysis, 3rd edition. New York: McGraw-Hill, 2. 11