Lessons Learned from 10-Years of Operating Transit Signal Priority on Howard Street, in Baltimore, Maryland: It s Not Just About the Technology

Transcription

1 Lessons Learned from 10-Years of Operating Transit Signal Priority on Howard Street, in Baltimore, Maryland: It s Not Just About the Technology Keith Riniker Director of Traffic Engineering & ITS Sabra, Wang & Associates, Inc Samuel Morse Drive Columbia, MD , KRiniker@Sabra-Wang.com ABSTRACT Transit Signal Priority (TSP) has been in operation for 10+ years at 16 intersections on Howard Street. Travel time is one of the most important performance measures for Light Rail Transit (LRT) in Baltimore, and the signals were initially thought to be the main source of travel time delays. Through our studies, we ve learned that the technology of TSP is important, but equally, or perhaps more important is Passive TSP in the form of traffic signal timing, active maintenance of the system components, and staff KSA (knowledge, skills and abilities). We ve also learned that non-traffic signal related delays (due to LRT operations) are equal to, or higher than traffic signal related delays for the LRTs. This paper summarizes lessons learned from operating TSP for 10+ years, and a summary from our latest investment in good old fashioned engineering, maintenance and operations on the subject signal system. KEY WORDS Transit Signal Priority Light Rail Transit Traffic Signal Timing Traffic Signal Maintenance and Operations Travel Time and Delays TRANSIT SIGNAL PRIORITY: A BRIEF OVERVIEW Active Transit Signal Priority (TSP) modifies the normal signal operations to provide preferential treatment to the vehicle making the request without disrupting the traffic signal cycle or coordinated operations of the system. TSP may be Passive or Active. Passive TSP does not require specialized equipment as is required for Active TSP. Specialized equipment for Active TSP may include: Priority Request Generators (typically on the transit vehicle), Priority Request Servers (typically at the traffic signals), supporting communications, and traffic signal controller firmware that supports TSP. An example of Passive TSP is designing the timing the traffic signals to include transit vehicle progression. Priority does not interrupt the general timing relationship between specific green indications at adjacent intersections (i.e. coordination). When requested, TSP may or may not be granted depending on a number of factors. Active TSP requires special components to request and grant TSP based on the presence of the LRT. Active TSP is the type of TSP in operation on MTA s Central Light Rail Line on Howard Street in Baltimore City. -1-

2 TRANSIT SIGNAL PRIORITY: A HISTORY OF SUCCESSES & FAILURES A lot of the literature on TSP focuses on benefits and successes of TSP projects. And, it should. TSP does provide benefits. It is also human nature to tout our successes and hide our failures, and so the TSP projects that get reported tend to be those that were successful. The other thing to remember is these benefits were often measured when the system was first implemented; before there were any maintenance or operational concerns. We tend to hear the soundbytes of the benefits and don t consider the cost. Examples of soundbyte benefits (I ve taken the context out of the quotes) include those listed in ITS America s 2005 Transit Signal Priority (TSP): A Planning and Implementation Handbook: TSP and signal optimization reduced transit signal delay about 40% TSP experienced a 10% improvement in travel time Average of 15% reduction (three minutes) in running time 25% reduction in bus travel times with TSP But, is not experience the best teacher? Of course! And so we should also describe our failures as well. Consider the following quotes, which are based on a combination of personal conversations and other TSP projects: no statistical difference in performance (i.e. getting back on schedule) between scenarios where the TSP was active versus when it was inactive. The [TSP] installation was discontinued because it didn t provide desired time travel benefits without resulting in excessive congestion on minor roads. No slack time on the side streets was available. TSP had no impact on decreasing bus travel times citing two main problems related to the green extension and that TSP was limited to one call every 10 minutes Downtown was originally running on priority but has been shut down by the City due to inconsistencies in the system. Most of the [TSP] hardware and configuration has gone to the weeds as they do not have the staff to maintain it. As may be observed from the above quotes, there are a multitude of reasons that TSP does not succeed. ITS America s 2005 Transit Signal Priority (TSP): A Planning and Implementation Handbook did an excellent job of addressing each of these issues. THESIS STATEMENT & CONCLUSION Our experience on Howard Street is that maintenance and operations play a very important role in transportation technology, and it s something that we as an industry often neglect. So, it s not that technologies such as TSP do not help; it s just that they are not the panacea that we hoped it would be. As we invest in new technology, we need to balance our desire for and spending on toys with old fashioned engineering, management of maintenance and operations, and investment in training our human resources. In the following sections, I hope to prove my point to you. Readers familiar with TSP may choose to skip the following section on Background. -2-

3 BACKGROUND ON THE HOWARD STREET LRT CORRIDOR The Howard Street Light Rail corridor, between the Camden Yards and Mt. Royal stations, in the Baltimore Central Business District (CBD) operates within Howard Street in a 1.9 mile exclusive guideway, through 16 signalized intersections operating with Active Transit Signal Priority. Figure 1 illustrates a map of the downtown area of Baltimore City with the Howard Street corridor circled. Figure 2 is a photo of the corridor south of Lombard Street. LRT operations are line of sight in the corridor - there is not active train control system in the downtown area, and LRT operations are controller by the traffic signals. Outside of the downtown area, there is an active train control system that provides vehicle location and provides travel time data. However, vehicle location is unknown and there is no travel time data in the subject 1.9 mile segment. There are LRT signal indications as per the MUTCD (Manual on Uniform Traffic Control Devices) located on the traffic signal mast arms. The LRT speed code is 25 miles per hour, but there are sections where operators typically drive much slower due to track alignment and frequent pedestrian activity. There are two tracks one for northbound and one for southbound trains. The TSP equipment consists of Model 1000 GTT Opticom (formerly a 3M product) GPS (Global Positioning System) Phase Selectors (the NTCIP term is priority request server) communicating via a 2.4 GHz Spread Spectrum radio system with proprietary communication protocol located within the card rack of the Baltimore City traffic signal cabinets. The radio antennas are mounted on the Baltimore City traffic Figure 1: Area Map of Howard St LRT signal poles. The 3M / GTT Opticom system Corridor in Downtown Baltimore also includes vehicle equipment (the NTCIP term is priority request generator) located in the Light Rail Vehicles with the radio antenna mounted on the top of the LRT. The priority request generator is tied into the doors closed switch so that when the LRT doors are open, the priority request generator (PRG) does not send requests to the priority request server (PRS) for TSP. Figure 3 illustrates photos of the TSP equipment. -3-

4 Figure 2: Photo of Howard Street and LRT Corridor Figure 3: Photos of TSP Equipment As part of the City s signal system upgrade in 2007, Siemens, Inc. deployed Siemens NEMA TS 2, Type 1 m50-series controllers with NextPhase software at the 16 signalized intersections on the Howard Street Light Rail corridor for Transit Signal Priority. These intersections communicate with the Siemens i2 signal system software using AB3418 protocols. The i2 system provides the City with the capability to monitor the status of all signals in the City in real-time though system graphics that display the active status of signals that are operating in coordination, free, transition, priority and/or preemption and special plans. i2 also stores controller data for split monitoring, time-space diagram analysis, and Synchro software data exchange. i2 is able to produce reports of Transit Signal Priority activity. Baltimore City TMC staff prepares TSP reports from i2 and the reports (see Figure 3 for a sample of the report) to the Maryland Transit Administration, Light Rail Operations staff. Note that Baltimore City TMC staff does not evaluate the TSP report data (note that the TSP report, as described above primarily contains LRT related information, not traffic signal performance information and this is the reason that the City does not evaluate the data). -4-

5 One of the unique features of the north-south oriented Howard Street LRT corridor is the east-west arterials carry considerably more traffic than Howard Street. Hourly traffic volumes are 400 to 600 vph on Howard Street, whereas cross street traffic is 600 to 3,000 vph. The conflict between vehicular volumes makes Passive TSP (i.e. signal timing for LRT progression) particularly challenging. Figure 4: Baltimore City i2 Signal System Software TSP Report Sample TSP strategies in use are limited to green extensions and red truncations within the coordinated traffic signal system. Traffic signal TSP split timings on Howard Street are generally 10 seconds different than the normal coordinated split timings (i.e. there is 10 seconds available for early green or green extension). In addition, there are some sections of passive priority on Howard Street, such that the signal timings are designed to provide for vehicle and LRT progression. TSP is enabled at each intersection, but not on each approach. There are 7 stations (14 locations, 1 per each direction) in this segment, and all 7 stations are immediately adjacent to a signalized intersection. There are 10 locations where the station is on the near side of the intersection and TSP is disabled on this approach. Various strategies were tried to enable TSP on the approach to near side stations, but the final solution was due to the variability in station dwell times (15 to 45 seconds) and false TSP calls. The subject Light Rail line serves 1 main line and 2 branch lines; in the Howard Street corridor there are a total of 3 overlapping routes. Due to the overlapping routes, the published LRT headway pattern is 10 minutes - 10 minutes - 5 minutes - 5 minutes. Headways tend to be erratic and at times we observe 20 second headways between LRTs in the same direction due to the overlapping routes. From the discussion above, it is shown that the system is generally transparent to the Maryland Transit Administration (MTA) and the Baltimore City Department of Transportation the operators, engineers and staff do not interact with the system on a daily basis to make it work. MTA and the City of Baltimore have a Memorandum of Understanding (MOU) for the operations and maintenance of the TSP equipment. MTA owns the PRS and PRG. The City performs maintenance on the PRS, and is reimbursed by the MTA for labor costs; MTA provides the City with all equipment. The MTA performs maintenance on the PRG. At the MTA s Central Light Rail Yard, the MTA installed a PRS -5-

6 with a traffic signal light so that as train operators leave the yard every morning, they can verify that the PRG is working when the traffic signal illuminates. HOWARD STREETTSP: OUR HISTORY OF SUCCESSES & FAILURES Initially, when the Howard Street Light Rail line opened in 1992, the system operated under what was then called priority control, a form of traffic signal preemption. LRT operators would manually request preemption of traffic signals between stations through a push button. The preemption system was overly disruptive to downtown traffic and was disabled. This was because Howard Street is generally the side street, and preemption interrupted coordination and progression on the mainline / cross streets. Preempt calls exacerbated congestion on most of the major cross streets that were already operating under congested conditions and experiencing long queues during normal peak hours. Additionally, it took up to 6 minutes (3 cycles) for the controller to transition from a preempt call and get back in step for normal signal coordination. The preemption system was disabled in the mid-1990s. In 2006, the MTA and City hired a consultant to retime the traffic signals with a focus on balancing the needs of LRT progression with vehicular delays, resulting in an average (over a typical weekday) of 1 ½ minutes of travel times savings for LRTs along the route. In 2007, as part of the City s signal system upgrade, the TSP equipment previously described was installed and made operational. This further reduced LRT travel time along the route by an average (over a typical weekday) of an additional 1 minute. Between 2007 and 2014, there was very little involvement from the MTA or the City in terms of maintenance and operations of the TSP system. As a result, LRT travel times increased. In 2014, the MTA hired our firm to address some of the maintenance and operations issues with TSP and evaluate and implement methods to reduce LRT travel times. The results of this study are presented later in this paper. In 2014, at the start of the project, we found that of the 16 intersections, TSP was only fully functioning at only 50% of signals. The specific maintenance items addressed included replacing broken equipment, and updating firmware for improved operations. Operations improvements included assessing and optimizing the TSP and signal timings parameters. At the start of the project TSP was not functioning due to a combination of reasons, including: Priority Request Server Antennas Malfunctioning Priority Request Server Manually Disabled (unplugged from the card rack; no one noticed and not plugged back in) Priority Request Server Malfunctioning Signal Timing Plans that did not accommodate TSP (i.e. no slack time) PRS Detection Zones in the wrong location (a Station was moved and the PRS zones were not updated) PRGs (on board the LRT vehicles) Malfunctioning The reason for the lack of maintenance is complex, but some of the reasons are: The TSP equipment is in no-man s land. It s owned by the transit agency but maintained by the signal agency. -6-

7 Some of the communication to the traffic signals in this corridor has been down for many years and the only way to diagnose malfunctioning equipment is to inspect the cabinet. Signal technicians are overwhelmed with other work and don t pay as much attention to transit equipment. The signal technicians don t have adequate training on how to operate the TSP equipment. So, when it breaks, because it appears complicated to fix it, it gets deferred. The TSP system operations are transparent to the transit agency (engineering, operators, and maintenance staff. So, if one component is broken, it is difficult to notice. It takes some amount of time before someone notices that LRT travel times have increased (which is what happened in 2014). The transit agency does not have easy access to their TSP equipment they must arrange for an inspection with the signal agency. Neither does the transit agency have any automatic notification if the TSP equipment fails. The transit agency maintenance staff, like the signal technical staff, is also overwhelmed with other work. As a result, the only human resource that has adequate time, knowledge and abilities are typically consultants. And so, when consultants get involved, LRT travel times improvement. Figure 5 presents and illustrated history of LRT travel times on the subject corridor. This is based on various data collected as part of before and after travel time studies. Figure 6 is an illustration based on anecdotal evidence that we have observed; it basically says that when we invent our time and resources, we meet our primary performance metric of reduced travel times. So, it s not that technologies such as TSP do not help; it s just that they are not the panacea that we hoped it would be. As we invest in new technology, we need to balance our desire for and spending on toys with old fashioned engineering, management of maintenance and operations, and investment in training our human resources. Figure 5: History of LRT Travel Times on Howard Street -7-

8 Figure 6: Illustrative Correlation between Travel Times and Maintenance & Operations As mentioned, in 2014, the MTA hired our firm to address some of the maintenance and operations issues with TSP and evaluate and implement methods to reduce LRT travel times by optimizing TSP parameter and signal timings. Between 2007 and the period where the TSP system was largely ignored, LRT travel times in the study section increased between 1 and 3 minutes. Recall that in 2007, the TSP system was brand new and fully functioning. In 2014 Before we began our optimization, we had repaired or replaced some of the malfunctioning equipment, but TSP was only functioning at about 75% of the intersection. After we completed our optimization, travel times decreased. Table 1 summarizes the data. Table 1: History of LRT Travel Times on Howard Street* (Minutes:Seconds) Direction & Peak Hour 2014 After 7 years Increase in Travel Increase in Time 2007 Initial TSP System 2014 Maintenance & Signal Timing 2014 Decrease in Travel Time (%) Time (%) (min:sec) Southbound AM 13:13 15:17 16% 2:04 14:21 6% Midday 12:23 16:17 31% 3:54 14:01 14% PM 14:03 15:16 9% 1:13 14:54 2% Northbound AM 14:16 14:26 1% 0:10 13:37 6% Midday 12:59 13:46 6% 0:47 14:11-3% PM 13:44 16:57 23% 3:13 13:34 20% * It is difficult to assess the exact impact of travel times due to differences in the start and stop points and the method of the travel time study between the 2007 and 2014 studies. The 2007 studies collected and then presented average travel time over a 4 hour period using time-stamp data from the GTT TSP equipment, whereas the 2014 studies manually collected data during the peak-of-the-peak period (i.e. 7 AM to 9 AM). In the table above, I have normalized the 2007 data for comparison here based on data we collected in 2014 using the 2007 methodology. -8-

9 TSP PARAMETERS AND SIGNAL TIMING OPTIMIZATION Data Collection The following data were collected in order to 1) determine the source of delays for LRTs, 2) evaluate the effectiveness of the current TSP and signal timings, and 3) serve as a performance measures for comparison between Before and After studies. LRT Travel Times and delays Station Dwell Time, including: o Boarding / Alighting Delays o Disabled passenger boarding/ alighting frequencies o LRT operator start-up / lost time (the time between the Go signal and when the LRT starts to move). See Table 2. Traffic Signal Delays Number of Stops and Locations LRT acceleration and deceleration rates and Table 2. Start-Up / Lost Time (seconds) Average Minimum Maximum speeds in each block), which provided the basis for plotting LRT progression on timespace diagrams. Traffic Signal Timing, Phasing & Geometric data Traffic volumes (LRT and intersection counts) TSP Frequencies, using data from the i2 split monitor reports. See Figure 7. Disabled passenger boarding/ alighting frequencies are important because the LRT dwells for a much longer period of time. At the first car in the train, the operator manually deploys a boarding ramp to assist passengers in boarding from the High-Block Platform. Figure 7: Traffic Signal Cycles where TSP was Granted Before Optimization -9-

10 The data illustrated in Figure 7 is interesting in that it shows that signal timing has the largest impact on TSP frequency. At signals where the signal timing pattern matches LRT progression, TSP is rarely or never granted (nor requested). This is of course intuitive, however, it is interesting to note that at signals that don t match the LRT progression, TSP is granted quite frequently (i.e. Franklin at Howard). Note that TSP is granted about 20-25% of the cycles at Howard and Lombard this is about 1 of 4 cycles and one would expect that it does not have impact to traffic operations. A Level of Service (LOS) analysis was conducted for the AM peak with the average (average over normal and TSP cycles is about a 2 second change) splits and revealed LOS C. However, due to train bunching that frequently occurs, we noticed that TSP was granted in some periods at 4 out of every 5 cycles! When the LOS analysis was reconducted using this average split data, the analysis revealed LOS E. Signal Timing Methodology Limitations to the signal timing changes included retaining the cycle length (we could not retime the entire downtown zone), retaining the offset pattern between two heavily traveled route with heavy left turns in a short block, and retaining progression along several high volume east-west routes (Frankilin, Mulberry, MLK signals). At several signals (Saratoga, Centre, etc.) the City allowed the MTA full-reign (opportunities) to optimize the timing solely based on LRT needs. A Synchro model was constructed and used as a basis for developing the optimized signal timings. An analysis of vehicular volumes revealed that signal timing splits and offsets did not need to change as there was not much change in traffic volumes downtown between 2007 and The goal, within the above constraints and opportunities were to not have the LRT stop between stations. The signal timings were planned to account for station dwell times. At this point, the Synchro model was abandoned and a time-space diagram model was used. Figure 8 illustrates the AM peak period Time- Space Diagram Model. The signal timing changes consisted of splits and offset changes at 10 of the 16 intersections, and cycle length changes at 3 intersections. Figure 8: Time-Space Diagram Model with LRT Trajectories -10-

11 TSP Parameter Methodology TSP changes including updating zone lengths, and TSP minimum and maximum travel times. Refer to Figure 9 for an illustration of the zone lengths. The zone lengths are based on the GPS zones set in the PRG. The goal was to maximize these lengths, but not make them so long as to extend into adjacent intersections. The longer the length translates into a longer time between when the LRT checks-in to a traffic signal s TSP zone. With longer times, there would be more opportunities for the traffic signal to grant TSP. The problem is that in short blocks, the zone cannot extend into adjacent intersections; otherwise an LRT could check-in to a downstream signal and request and be granted TSP, but be stopped at a red signal upstream; it is inefficient. A problem that exacerbates this short-block scenario is the 2-3 second lag time between when an LRT departs an upstream signal (Intx. 1) and check-in to the downstream signal s TSP zone (Intx 2). This is due to the GPS equipment in use. Practically, this means that if the normal LRT travel time from intersection to intersection in a 400 foot block is 10 seconds, the TSP equipment at the traffic signal does not know the LRT approaching until 6-8 seconds out. At this time, the LRT clearance (the yellow signal for the LRT; about 6-8 seconds) may have already started, denying an opportunity for TSP. In addition, the minimum and average travel time values (signal timing parameters for TSP in the signal controller that dictate the range split extensions or truncations). These were initially calculated using the following, and then field tested and updated based on field observations. MMMMMM TTTT (ssssss) = TTTTTTTTTTTT dddddddddddddddd VV mmmmmm LLLLLL tttttttt = LL 50 VV mmmmmm 3 MMTTmm TTTT (ssssss) = BTTBssB DMMssddTTMMssss PPP LLTTLL DMMssddTTMMssss 50 VV mmm Intx. 2 Check-in Intx. #1 Intx. #2 TSP Zone Intx. 2 Check-out Figure 9: TSP Zone and Min/Max LRT Travel Time Calculation Summary of Improvements The following tables summarize the improvements. Figure 10 illustrates the presentation of the results for the AM peak hour on a signal-by-signal basis. -11-

12 Table 3. Average Travel Time Table 4. Average Signal Delays Table 5. Sources of Delay Table 5. Arrivals on Green Figure 10: Sample of Illustration of Results, AM Peak -12-