Measuring Control Delay Using Second-by-second Gps Speed Data

High-resolution vehicle speed profiles obtained from sophisticated devices such as global positioning system (GPS) receivers provide an opportunity to accurately measure intersection delay, composed of deceleration delay, stopped delay, and acceleration delay. Although the delay components can be measured by manually examining the speed profiles or derived timespace diagrams, identifying when vehicles begin to decelerate or stop accelerating is not always a straightforward task. In addition, a manual identification process may be laborious and timeconsuming when handling a large network or numerous runs. More importantly, the results from a manual process may not be consistent between analysts or even for a single analyst over time. This paper proposes a new approach to identifying control delay components based on secondby-second vehicle speed profiles obtained from GPS devices. The proposed approach utilizes both de-noised speed and acceleration profiles for capturing critical points associated with each delay component. Speed profiles are used for the identification of stopped time periods, and acceleration profiles are used for detecting deceleration onset points and acceleration ending points. The authors applied this methodology to sampled runs collected from GPS-equipped instrumented vehicles and concluded that it satisfactorily computed delay components under normal traffic conditions. TRB 2007 Annual Meeting CD-ROM Original paper submittal not revised by author. Ko, J., Hunter, M., and Guensler, R. 2 INTRODUCTION As delay is closely linked with driver discomfort, frustration, fuel consumption, and lost travel time, it is a key performance measure for signalized intersections and used as a criterion for evaluating and designing traffic control systems. In fact, the Highway Capacity Manual (HCM) has adopted delay as the prime measure for determining the level of service (LOS) at an intersection. The HCM [1, 2] defines intersection LOS based on control delay, which includes initial deceleration delay, queue move-up time, stopped delay, and final acceleration delay. Thus, the identification of acceleration and deceleration delays as well as stopped delay (which for this effort is taken to include queue move-up time) is critical to the analysis of the performance of signalized intersections or corridors containing traffic signals. However, deceleration and acceleration delays are not easy to capture without the help of sophisticated devices providing high resolution vehicle speed profiles, such as second-by-second. In contrast, stopped delay is relatively easy to measure in the field, which may explain why stopped delay has long been the primary field measured intersection delay and why control delay has been estimated based on the measured stopped delay. However, stopped delay does not reflect every aspect of intersection performance affected by traffic signals. The relationship between stopped delay and other delay components may not be established in a single all encompassing function, due to site-specific factors affecting the relationship such as signal timing and driver characteristics. Indeed, three different sources reported three significantly different relationships between control delay and stopped delay as follows. Stopped delay = 0.76 × Control delay TRB [3] Stopped delay = 0.959 × Control delay 19.3 Quiroga and Bullock [4] Stopped delay = 0.58 × Control delay 2.31 Mousa [5] For example, if measured stopped delay is 10 seconds, resulting control delays based on these three different equations are 13, 31, and 21 seconds for TRB [3], Quiroga and Bullock [4], and Mousa [5], respectively. This significant discrepancy implies that each delay component should be measured in the field rather than estimated based on stopped delay. One tool allowing for sufficient data collection for delay determination is a global positioning system (GPS) device [6]. A GPS device can provide data describing detailed vehicle speed trajectories, enabling researchers to detect when a vehicle begins to decelerate, stops accelerating, stops, or starts moving. However, these critical points are not always easy to identify as speed profiles sometimes have irregular patterns requiring researcher’s subjective judgment, resulting in potential inconsistencies in delays calculated by different analysts and even for a single analysts. In addition, individual review of every vehicle trajectory is laborious and time-consuming work when handling a large network or numerous runs, further requiring the development of efficient automatic processes. DEFINITION OF CONTROL DELAY COMPONENTS Control delay at a signalized intersection is generally defined as the delay attributed to the traffic signal operation. Control delay is a portion of the total delay which includes all the delay components, including control delay, geometric delay, volume delay, and incident delay [7]. TRB 2007 Annual Meeting CD-ROM Original paper submittal not revised by author. Ko, J., Hunter, M., and Guensler, R. 3 However, in a practical sense, the distinctions between these delay components are considerably difficult to measure, particularly when using remotely monitored data. For simplicity, this paper defines intersection control delay as the sum of deceleration delay, stopped delay including queue move-up time, and acceleration delay, as illustrated in Figure 1. It is seen in Figure 1 that control delay may be calculated as the difference between the travel time when a vehicle’s trajectory is affected by traffic control and the travel time when the vehicle’s trajectory is unaffected by the traffic control. The control-affected region ranges from d1 to d3, and correspondingly from t1 to t4 in Figure 1. The initial part of the control delay is the deceleration delay, which is due to a slow-down from a normal speed. Next, the vehicle is stopped (from t2 to t3), and this time period represents the stopped delay. Finally, acceleration delay occurs while the vehicle is returning to a normal speed. Thus, the computation of control delay requires the identification of the critical delay points (i.e., t1 to t4) when a vehicle begins to decelerate, stops, or starts moving, and reaches its normal speed. An examination of speed profiles reveals the two points associated with stopped delay. Even though, strictly speaking, zero speed is the criteria for a vehicle to be considered stopped, a higher speed threshold may be applied to allow for the identification of vehicles crawling forward in a queue as stopped. For example, Mousa [5] used 1 1.5m/s (2.237 – 3.356 mph), and Colyar and Rouphail [8] applied 3 mph as the stopped criteria when measuring stopped delay. Also, analyses using GPS data require a threshold greater than zero speed as the values of speed are rarely represented as exactly zero within a GPS data stream [9]. Therefore, in this research effort, stopped delay is computed as the amount of time during which the velocity of a vehicle is below a given threshold.