Transit Signal Priority: Current State of the Practice (2020)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

11 C H A P T E R 2 This chapter is organized in four sections: • Rationale for TSP Deployments: Describes the context and operational problems that motivate the development of TSP and discusses how TSP fits with other transit-supportive treatments. • Current TSP Prevalence and Practices: Reviews reports and research papers that document the current state of the practice in North America and elsewhere. • Research on Transit Signal Priority: Discusses developments in the academic and professional literature on TSP. • Summary of the Literature Review: Summarizes key findings from the literature review. Rationale for TSP Deployments Two of the main challenges facing bus transit services that operate in mixed traffic are traffic congestion, which slows down buses, and travel time variability, which reduces buses’ reliability. Although not the only source, traffic signals are a key cause of delay and travel time variability for buses. Delays at traffic signals reduce the speed of buses and makes them less appealing as a travel mode. Travel time variability is a problem for transit agencies, because it increases the difficulty of maintaining a schedule or maintaining regular headways. There are operational and signal-based interventions available to transit agencies and their local partners to help improve speed, reliability, or both. Operational Approaches to Improving Speed and Reliability There are different operational approaches to improving bus reliability or on-time perfor- mance, and many transit agencies use a combination of operational and scheduling techniques. Transit agencies can actively monitor and manage the bus fleet using AVL systems to improve bus reliability. Similarly, bus schedules need to be updated on a regular basis to adapt them to changing traffic and road conditions. Schedulers can use on-time performance reports, operators’ feedback, and field observations to optimize bus schedules. Additional scheduling techniques for improving bus reliability are adding slack time and layover time to schedules. Adding slack time is an approach that gives buses more time to travel along a corridor than is typically required. The additional time allows delayed buses to main- tain their schedules. If the additional time is not needed, buses can pause at designated points along the route to get back on schedule and avoid being early. Too much slack time unduly delays passengers and increases the cost of the route; too little slack time results in little or no benefit to reliability. As Newell (1977) pointed out, the amount of slack time required to smooth out all travel time variability is impractically large, though it is useful at smoothing out small Literature Review

12 Transit Signal Priority: Current State of the Practice variations. Adding layover time, which is usually time at the terminus of a route, can allow the route to recover from being late and gives the bus operator a chance for a small break. Too much layover time increases the cost of the route; too little layover time results in little or no benefit to reliability. Slack time and layover time help improve reliability but are difficult to optimize. There is an inherent tradeoff between the ability of a bus’s schedule to absorb all amounts of travel time variability and the resulting scheduled speed and service cost. As slack time and layover time increase, the effective speed of the bus decreases and the cost of service increases. Slack time also has a customer service implication—riders on board buses that are pausing to expend available slack time may be frustrated by the seemingly wasted time. However, buses can become late enough that they cannot recover through slack time alone. In this case, transit agencies can implement short-term operational interventions to help buses get back on schedule, including boarding limits, stop skipping, and short turns. A boarding limit seeks to cap dwell time at bus stops (Delgado et al., 2009; Delgado et al., 2012). Once the maximum dwell time has been reached, the bus will close its doors and depart, even if there are still passengers who want to board. This practice is common in rail operations and has been used on some high-frequency bus routes. Stop skipping is the practice of skipping a single bus stop (Sun and Hickman, 2005; Liu et al., 2013) or sometimes several bus stops in a row (Nesheli and Ceder, 2014, 2015) to make up time, assuming no passengers want to get off. A short turn is when a late bus is directed to go out of service before the end of the route and turn around to begin its return trip. The intent is to get the bus back on schedule even though part of the route is not served by doing so. Each of these interventions has customer service implications. Customers may be denied service, be delayed, or be forced to get off a bus in an undesired location. These interventions are also short-term solutions and do not correct the underlying causes of unreliability. So, how can transit agencies and their local partners work to implement sustainable solutions to improve reliability while avoiding the negative customer impacts associated with the previously discussed operational interventions? One solution is through implementing signal priority, which may be passive or active. Signal-Based Interventions to Improve Speed and Reliability Passive priority interventions (i.e., signal priority measures that are not activated by transit vehicles but instead seek to improve transit vehicle movement along a corridor) can be used to address some sources of transit delay and unreliability. Urbanik and Holder (1977) mention several passive priority strategies, including adjusting cycle length, developing areawide signal timing plans, and metering vehicles. Shorter cycle lengths benefit buses because they reduce signal delay (Garrow and Machemehl, 1999). An areawide signal timing plan considers several signals at once. Designing signal progression for buses is challenging, because dwell times at bus stops are highly variable (Lin et al., 2015). Many urban arterial streets have signal progression, in which the timing and offsets of signals are designed to form a green band, which allows cars to pass through several consecu- tive intersections without having to stop for a red light. Buses may not be able to benefit from signal progression, because they stop for passengers and fall out of the green band (Duerr, 2000; Skabardonis, 2000). This phenomenon can cause buses to experience more signal delay than other vehicles. Another approach to signal progression is to time signals in accordance with average bus speed instead of general traffic speed. This type of signal progression may benefit buses, but could have some detrimental impacts on general traffic flow.

Literature Review 13 Some studies have used the presence of buses to justify additional green time for bus approaches, which increases the probability of buses arriving on green and reduces their waiting time when they arrive on red (Skabardonis, 2000; Ma and Yang, 2007). Metering (also called perimeter control) restricts the flow of traffic on certain links (i.e., road segments) to prevent a corridor or area from becoming overly congested (Geroliminis and Daganzo, 2008; Ortigosa et al., 2014). Buses benefit from more reliable travel times in the controlled area and are often exempted from the metering through use of a dedicated bus lane, queue jump, or pre-signal (Wu and Hounsell, 1998; Guler et al., 2016). Passive priority measures may perform better than TSP at high frequencies (e.g., 60 buses per hour) when the number of TSP requests would be highly disruptive (Hounsell and Wu, 1995). Passive priority measures can also be used together with TSP—for example, by allowing buses to request priority when they fall outside of a corridor’s signal progression–based green band. TSP can provide improved bus speed and reliability without the unintended negative cus- tomer experiences that result from adding slack time or enacting other short-term operational interventions. Both passive priority and TSP improve bus speed; however, the on-demand nature of TSP may help balance the competing demands at intersections with lower bus frequencies, particularly when priority is conditional and not requested by every bus. TSP can also improve bus speed and reliability without increasing the service’s scheduled operating cost (in terms of revenue hours and operator requirements). Improving bus speed is appealing to transit agen- cies, because it allows them to run the same bus frequency with fewer vehicles and less environ- mental impact (Smith et al., 2005; Bunch, 2018). Improving travel time reliability allows transit agencies to reduce slack in the schedule, further improving speed, and is appealing to passengers (de Palma and Picard, 2005; Smith et al., 2005). Although TSP may have significant benefits for transit agencies and their riders, there are many different approaches to designing and implementing TSP systems, and these approaches are occasionally documented in publicly available literature and reports. Current TSP Prevalence and Practices This section reviews TSP reports and research papers that document the current state of the practice in North America and elsewhere. The review is intended to be a sampling of reports and papers that describe where and how TSP is implemented and is not an exhaustive list or a meta-analysis. • Several previous reports have produced lists of agencies known to be using TSP (e.g., Casey, 1999; Radin, 2005; Smith et al., 2005; Kittelson & Associates, 2007; Danaher, 2010; Tindale- Oliver & Associates, 2014). These reports, and other sources, were used to develop the list of survey recipients for this synthesis. • Collura et al. (2001) reviewed bus detection technologies available in 2001. Purdie (2002) described the deployment of a TSP system using GPS in Glasgow, Scotland. Smith et al. (2005) and Danaher (2010) mention that infrared, inductive loops, sound waves, radio frequency tags, GPS systems, visual recognition, and wireless communication have all been used to enable TSP. • Nash (2003) described the TSP deployment in Zürich, Switzerland, which is integrated with other transit priority and traffic control strategies (e.g., traffic calming and metering). Tindale-Oliver & Associates (2014) conducted interviews with a number of transit agencies that have successfully implemented TSP and developed a set of guidelines describing the implementation process and best practices. • Sabra, Wang & Associates (2016) described the process of identifying corridors and inter- sections for TSP deployment in Montgomery County, Maryland. Corridors were considered

14 Transit Signal Priority: Current State of the Practice according to the presence of significant transit service and performance issues. Key factors included average bus speeds during the p.m. peak period, productivity (ridership per vehicle mile), functional class of cross streets, and intersection level of service. • Bunch (2018) described a planned deployment of TSP within a new bus rapid transit (BRT) corridor in Montgomery County, Maryland. This deployment uses a distributed architec- ture and headway-based conditional TSP. (Priority requests will be issued by the trailing vehicle when the gap between vehicles is greater than 1.5 times the planned headway and will continue until the gap equals the planned headway.) Priority request conflicts will be resolved on a first-come, first-served basis, and intersections will have a three-cycle lockout period. Intersections were selected primarily on the basis of available slack time within the signal timing phases and volume-to-capacity (V/C) ratio. Other factors considered included cross street functional class, bus stop location, the presence of other bus priority treatments, other (non-TSP-using) bus routes, daily ridership, and average peak-hour speed. • Lozner et al. (2018) described a large deployment in 2016 of TSP at 195 intersections in Washington, D.C. This deployment uses a distributed architecture and rules limiting priority requests to late buses (with a threshold of 7 minutes) and peak periods/directions (inbound from 6:00 a.m. to 9:00 a.m. and outbound from 4:00 p.m. to 6:00 p.m.). Intersections were selected according to stop location, flexibility in signal timing, and V/C ratio. Although TSP is implemented in many regions, research actively continues to better under- stand optimal system architectures, business rules and parameters, and benefits (and the determinants of benefits). Research on Transit Signal Priority This section describes developments in the academic and professional literature, including work on TSP technology, methods for resolving conflicting priority requests, TSP benefits for schedule-based operation and implications for schedule design, TSP benefits for headway- based operation, and impacts on other road users and mitigation strategies. Again, this review is intended to provide a sampling of TSP research and should not be considered an exhaustive list or meta-analysis. TSP technology has evolved over time, and two recent works highlight how these changes may lead to new priority strategies: • Hounsell et al. (2008) described London’s transition from an AVL system using roadside bea- cons to one using GPS, and predicted that the lower location accuracy of GPS would reduce bus delay savings from TSP by 2% to 5% but also enable new priority strategies. • Hu et al. (2014) has proposed implementing TSP with connected vehicle technology, which permits two-way communication between buses and signals and would allow for new types of TSP strategies. Several studies have looked at the effects of various parameter values on the benefits of TSP: • In early work, Jacobson and Sheffi (1981) modeled traffic impacts of bus priority at a sim- ple isolated intersection and recommended parameter adjustments (e.g., cycle length, phase length) to increase the benefits of TSP. • Rakha and Zhang (2004) looked at the sensitivity of benefits to signal timing and bus location parameters. They simulated a single intersection within a coordinated corridor and reported higher benefits to buses as congestion increases, higher benefits to buses when there are more signal phases, and a decrease in systemwide benefits with increased near-side dwell time. • Anderson and Daganzo (2020) showed that a lateness threshold parameter could be tuned to eliminate mean schedule deviation.

Literature Review 15 When TSP is implemented on intersecting bus routes, conflicting priority requests must be addressed at the signals where the bus routes cross. Different methods can be used to resolve conflicting requests, ranging from simple rules (e.g., first come, first served) to optimization techniques that try to serve the “best” request. Recent work on this topic includes • Christofa and Skabardonis (2011), • Zlatkovic et al. (2012), • Ma et al. (2013), and • He et al. (2014). Several case studies (both real world and simulation based) examined the travel time benefits of TSP: • An early field trial of unconditional TSP in Louisville, Kentucky, reported a time savings of 9% to 17% compared with express buses that did not use TSP (Capelle et al., 1976). Other traffic on the corridor was also found to benefit, which the study attributed to poor signal timing (i.e., not enough green time for the major road) in the base case. • Bunch (2018) reported an 8% to 13% improvement in travel time for all traffic after TSP for the Maryland Route 355 BRT was launched in 2017. • An early field trial in Miami comparing bus priority strategies found that bus travel time was reduced by 19% to 26% with unconditional TSP (Wattleworth et al., 1977) but that sched- ule reliability was slightly worse. This finding is consistent with later research arguing that conditional signal priority is needed to improve schedule reliability and that unconditional TSP, unless accompanied by schedule control, does not improve reliability, because all buses are treated the same (Janos and Furth, 2002; Anderson and Daganzo, 2020). • Hounsell et al. (1996) reported a 20% to 30% bus delay reduction with TSP. • A simulation study of unconditional TSP on Columbia Pike in Virginia (Dion et al., 2004) reported a 2.3% to 2.5% travel time savings for express buses, a 4.8% travel time savings for local buses, and an 18% increase in average travel time for all traffic when all buses receive priority. • A simulation study of the US-1 corridor in Virginia (Kamdar, 2004) found that a TSP strategy using only 10-second green extensions could reduce bus travel times by 4%, with an increase in maximum queue length on side streets of 1.23%. Research has also discussed how bus schedules should be adjusted to take advantage of TSP (Janos and Furth, 2002; Altun and Furth, 2009; Albright and Figliozzi, 2012; Anderson and Daganzo, 2020). Essentially, TSP allows buses to go faster, so the schedule should also be faster when TSP is used. If buses must be late to request priority, then the rate of priority requests will vary with the speed of the schedule. Anderson and Daganzo (2020) showed that condi- tional TSP alone can keep buses on schedule if an appropriate schedule speed is used, and that schedules slower than unconditional TSP can be maintained if both TSP and schedule control are used. Gordon (1978) described the problem of headway instability (also known as bus bunching) and suggested that traffic signals could be used to speed up or delay buses as necessary. Head- way control strategies based on holding buses at stops were developed in several recent papers (Daganzo, 2009; Daganzo and Pilachowski, 2011; Xuan et al., 2011; Bartholdi and Eisenstein, 2012; Daganzo, 2017) and demonstrated in field trials (Argote-Cabanero et al., 2015). The idea of using traffic signals for headway control resurfaced in Ma et al. (2010), which proposed using signals both to speed up buses (with green extension/early green) and to delay buses (with red extension/early red). Chow and Li (2017) used the same set of priority and delay actions to implement the headway control strategy of Daganzo (2009). However, strategies that use signals to delay buses may be difficult to implement in practice because of the number of requests from

16 Transit Signal Priority: Current State of the Practice buses and the impact on other traffic. To address this shortcoming, Anderson and Daganzo (2020) proposed a headway control strategy extending Daganzo (2017) using signals to speed up buses and holding at stops to delay them. Some papers in the academic literature have considered how to reduce the impact of TSP on other road users: • Bowen et al. (1994) recommended varying TSP parameters on an intersection level, according to spare capacity. • Sunkari et al. (1995) conducted a field study in College Station, Texas, to obtain accurate estimates of the delay to other vehicles when TSP is used. • Al-Sahili and Taylor (1996) conducted a simulation study of a corridor in Ann Arbor, Michigan, and reported increased delays to car traffic with TSP attributable to high volumes and interruption of signal progression. The ratio of arterial to cross street volumes was identified as a key factor. • Hounsell et al. (1996) recommended using only green extensions to reduce the impact on other traffic. • Balke et al. (2000) conducted a simulation study with hardware in the loop and recommended that TSP can be used at V/C ratios up to 0.9 without significantly delaying cross street traffic. • Hu et al. (2015) proposed using connected vehicle technology to monitor the traffic state and grant TSP when doing so reduces signal delay per person. • Chow et al. (2017) used both priority (green extension/early green) and delay (red extension/ early red) actions in an optimal control formulation to minimize schedule deviation and traffic delay. Summary of the Literature Review TSP can be a valuable tool in the toolbox for transit agencies that want to reduce bus travel times, reduce travel time variability, and improve headway reliability. However, there are many different possible supporting technologies, business rules, and parameters, all of which inter- act with local operating and traffic environments to produce different results. The literature review suggests that research on TSP has been, and continues to be, very active; and this synthesis will help contribute to the corpus of knowledge by helping to update the industry’s understanding of what is currently in place in North America and what technologies, busi- ness rules, and parameters appear to have the most promise for maximizing benefit for transit agencies and their customers.