Menendez PROVIDING BUS PRIORITY USING ADAPTIVE PRE - SIGNALS

1 Bus priority is important in cities to encourage people to use public transport. However, providing a dedicated 2 bus lane is not always feasible. It might not be the most efficient solution either, especially if bottlenecks such as 3 traffic signals exist. This paper examines an alternative strategy to use an additional signal upstream of the main 4 signal (a pre-signal) to provide bus priority at signalized intersections. Its primary purpose is to allow buses to 5 jump the car queues before the intersection while cars can still use all the lanes at the main signal to fully utilize 6 the capacity of the intersection. 7 This paper formulates an online control algorithm to operate an intersection with such pre-signal infrastructure. 8 The control algorithm is developed by determining the best operating strategy for the intersection at each 9 demand level using a micro-simulation model in VISSIM. In a simulation case study, it is observed that 10 compared to using a dedicated bus lane, the congestion upstream of the intersection during peak hours is 11 mitigated. Meanwhile, compared to using mixed-use lanes, the average person delay during the off-peak hours is 12 lower and bus priority is provided. Therefore, it is concluded that implementing a pre-signal with our proposed 13 online control algorithm provides a good balance between providing bus priority and sustaining car throughput. 14 3 He, Guler and Menendez INTRODUCTION 1 Providing priority to public transportation (e.g., buses) in cities could reduce journey times of buses. Often, this 2 kind of priority is provided with a dedicated bus lane. However, this might not always be feasible due to space 3 restrictions or political issues. It might not be the most efficient solution either, especially if the bus flow is low, 4 private vehicle (e.g., car) demand is high, or bottlenecks such as traffic signals exist. Under such situations, 5 alternative strategies could be proposed to provide bus priority while minimizing negative impacts on cars. 6 One of these alternative strategies is to use an additional signal (hereafter referred to as a pre-signal) to provide 7 bus priority at signalized intersections (1, 2, 3), see Figure 1. Its primary purpose is to allow buses to jump the 8 car queues upstream of the intersection while cars can still use all the lanes at the main signal to fully utilize the 9 capacity of the intersection. In this way, bus delays are reduced, while the capacity lost at the intersection is 10 minimized. 11 A static operating strategy for pre-signals has previously been proposed (4). In this strategy, the pre-signal turns 12 red in advance of a red main signal and also when a bus arrives to the pre-signal irrespective of the status of the 13 main signal. The goal is to ensure that the space between the pre-signal and the main signal is kept free of cars 14 for as long as possible, allowing any arriving bus to move in front of the car queue and discharge immediately 15 when the main signal turns green. The pre-signal turns green in advance of the main signal when no buses are 16 present such that cars do not experience additional delays, or after the bus has left the pre-signal when buses are 17 present. This operating rule is adopted as the basis to develop the online control algorithm in this paper. 18 19 FIGURE 1 Layout of a pre-signal (4). 20 However, a static operation might not be the optimal strategy for traffic scenarios varying throughout the day. 21 Different strategies have different effects on performance metrics of the road section. One possible strategy is to 22 extend the dedicated bus lane up to the main signal with the pre-signal switched off (hereafter termed the bus 23 lane strategy). This could further reduce the bus delay, and hence the average person delays if the bus 24 occupancies or the bus frequencies are high. Another possible strategy is to use both lanes as mixed lanes 25 (hereafter termed the mixed lane strategy). This could further increase the traffic throughput and reduce the 26 upstream queue length when the traffic demand is high. As one would expect, different strategies may perform 27 best with different car and bus demands. Therefore, there is a need to develop an online control algorithm to 28 dynamically switch between different strategies to best achieve traffic management and control goals. 29 To develop an online control algorithm, this research builds, validates, and uses a micro-simulation model. This 30 micro-simulation model is built in VISSIM and validated for a case study location in Zurich, Switzerland, using 31 empirical data collected at this site. Next, the micro-simulation model is used to analyze delays and other 32 performance metrics with different static operating strategies under different traffic demands. With a better 33 understanding of the real-time effects of the static strategies, an online control algorithm is formulated and tested 34 for an intersection with pre-signal infrastructure. 35 The remainder of the paper is organized as follows. In the next section, a review of existing literature on the 36 operation of pre-signals is discussed. Then, the calibration and validation of a micro-simulation model for a case 37 study site are discussed. Using this calibrated model, in the following section the performance metrics for the 38 three different static operating strategies (i.e. the mixed lane strategy, the bus lane strategy, and the pre-signal 39 4 He, Guler and Menendez strategy) under different car demand values are investigated. Finally, an online control algorithm for an 1 intersection with pre-signal infrastructure is developed and presented. 2 LITERATURE REVIEW 3 The idea of setting up a signal in front of the main signal (i.e. a pre-signal) is not new and has been analytically 4 studied before. Pre-signals have been proposed for several purposes such as to recover the lost time at the 5 intersection due to bounded acceleration (5) or to separate left-turning vehicles and through-moving vehicles to 6 maximize the discharging capacity of the intersection for both directions (6). However, here the pre-signal is 7 used to provide priority to buses. 8 The use of pre-signals to provide bus priority was first proposed in (1). That work provided guidelines to 9 determine the pre-signal timing for minimizing the system vehicle delay. The operating strategy proposed was 10 different from the one described in the introduction. Most notably, that strategy was proposed for intersections 11 without bus detection infrastructure. To the best of the authors’ knowledge, implementations of pre-signals are 12 rare; with only a few cases in U.K. cities, for example in London (7) with the operating strategy from (1). A 13 similar idea for pre-signals is mentioned in the German Manual for Transit Priority (8), but no implementation of 14 this strategy has been reported in Germany yet. In fact, to the authors’ best knowledge, the only pre-signal with 15 this prescribed operating strategy is located in Langstrasse, Zurich, Switzerland. 16 The static operating strategy that we will use in this paper was studied analytically using queuing theory for 17 isolated intersections assuming a fixed demand pattern (2, 3). For under-saturated intersections, the results show 18 that implementing a pre-signal could lower the average person delay if bus occupancy is in the range of 10-70 19 passengers. For over-saturated intersections, the pre-signal strategy always resulted in lower average person 20 delay compared to the bus lane strategy. Moreover, the pre-signal strategy resulted in lower average person delay 21 compared to the mixed lane strategy if the peak demand was greater than 105% of the signal capacity. 22 There are numerous detailed studies of online control algorithms at signalized intersections, for example SCATS 23 (9), and SCOOT (10). These online control algorithms apply to the main signal. However, the subject of this 24 research is an online control algorithm for a new kind of infrastructure, namely the pre-signal. To the authors’ 25 best knowledge, there is not yet research on this subject. 26 CASE STUDY LOCATION AND MICRO-SIMULATION MODEL 27 To develop and evaluate an online control algorithm for pre-signals, a micro-simulation model is used. However, 28 to validate this approach, first the driver behavior near a pre-signal should be accurately modeled. To do so, the 29 case study location and the corresponding empirical data, the micro-simulation model, and the calibration and 30 validation results are presented below in three consecutive subsections, respectively. 31 DESCRIPTION OF THE SITE AND EMPIRICAL DATA 32 The intersection on Langstrasse in Zurich has a pre-signal operation as described in the introduction and is used 33 as a case study location. The layout of the streets is shown in Figure 2. On this section of the road, traffic comes 34 from both upstream of the main signal and from the four side streets. The road upstream of the pre-signal 35 consists of one bus lane and one car lane. Between the main signal and the pre-signal, these two lanes merge into 36 a single shared lane. 37 Note that this is a special case because the road capacity before the main signal and the road capacity for cars 38 before the pre-signal are the same, so the red signal length for the pre-signal and main signal are the same based 39 on Guler and Menendez, 2013 (4). 40 5 He, Guler and Menendez 1 FIGURE 2 Layout of the pre-signal on Langstrasse, Zurich (4) and its simulation model. 2 Two data sets are used in this research. The traffic flow rates and traffic composition for both data sets are 3 calculated and summarized in Table 1. Note that the category ‘trucks’ includes vans and lorries of various sizes. 4 They are summarized in one category to simplify the calibration. Buses arrive with a uniformly scheduled 5 frequency of 8 buses/hour based on data set 1. 6 The main signal timing is complicated since the signals are adaptive with an unknown algorithm, and also transit 7 signal priority is implemented. Therefore, in reality each cycle might have a different cycle length, green-time 8 and red-time. However, during each measurement period, these values remained nearly constant, and hence the 9 signal is modeled as fixed-timed. Their average values are also summarized in Table 1. Note that as stated 10 previously, the length of the red signal and green signal for both the main signal and the pre-signal are the same, 11 with only a pre-signal offset. The ‘duration of the red signal with bus’ refers to the minimum duration of the red 12 pre-signal before the bus passes the pre-signal when a bus is detected. 13 TABLE 1 Summary of empirical data. 14 Data set 1 Data set 2 Date and time 14 November 2013 24 September 2012 Main road flow 288.1veh/h 346.8veh/h Side street 1 flow 28.9veh/h 39.5veh/h Side street 2 flow 10.6veh/h 2.6veh/h Side street 3 flow 29.7veh/h 7.2veh/h 6 He, Guler and Menendez Side street 4 flow 13.6veh/h 3.3veh/h Percentage of trucks 29% 15% Cycle length 85s 52s Duration of the red signal 57s 28s Pre-signal offset 7s 7s Duration of the red signal with bus 12s 12s DESCRIPTION OF THE SIMULATION MODEL 1 The simulation model is built in VISSIM according to the exact geometry specifications of the site as in Figure 2, 2 and the traffic flows and signal settings as in Table 1. 3 The desired speed in the simulation model (i.e. free-flow speed) is assumed to be the speed limit, which is 50 4 km/h in urban streets in Zurich. This is the same as in the calculation of the empirical data so the delay results 5 are comparable. Travel times are measured for 6 sections: from the main road upstream to the pre-signal, from 6 side streets 2-4 to the pre-signal, from side street 1 to the main signal, and from the pre-signal to the main signal. 7 The average vehicle delay is calculated by taking the difference between the measured average travel time and 8 the theoretical travel time at free flow speed assuming no signals. 9 MODEL CALIBRATION AND VALIDATION 10 This model is calibrated with data set 1 and validated with data set 2. The benchmark against which the model is 11 calibrated and validated is the average car delay measured with the two data sets. The total delay is an output 12 from VISSIM which measures the difference between the actual travel time and the free-flow travel time at the 13 desired speed. The total delay is divided into two parts, one part due to the pre-signal (delay PS) and the other 14 part due to the main signal (delay MS). The measured average car delays are summarized and compared to the 15 calibration and validation results in Table 2. 16 Each simulation run is 3700 seconds long including a warm-up time of 100 seconds. A bus arrives every 450 17 seconds. The Wiedemann 74 car following model is used and all the three parameters (average standstill distance, 18 additive part of safety distance, and multiplicative part of safety distance) are calibrated (11). For the lane 19 changing behavior, only parameters for compulsory lane changes namely minimum headway and safety distance 20 reduction factor, are calibrated,. The average values over ten simulation runs, each with a different random seed, 21 are used for calibration and validation. Table 2 summarizes the calibration and validation results. 22 From Table 2 we see that the measured values and the simulation results are close. We believe the micro23 simulation correctly models the operation of a pre-signal and the driving behavior near the intersection. 24 Therefore, it can be used as a basis to develop the online control algorithm. 25 TABLE 2 Summary of: a) calibration and validation results, and b) values for calibrated parameters. 26 (a) 27 Average delay PS Average delay MS Total delay Data set 1 measured 35.1s 6.3s 41.3s Data set 1 simulation (calibration) 33.2s 6.5s 39.7s Data set 2 measured 15.4s 6.0s 21.4s Data set 2 simulation (validation) 17.3s 7.2s 24.5s