1 Bullock Hardware in the Loop Simulation Darcy Bullock 1 , Brian Johnson 2 , Richard B. Wells 3 , Michael Kyte 4 , and Zhen Li 5 ABSTRACT The current generation of macroscopic and microscopic simulation packages do not realistically model the operation of modern traffic signal control equipment. Consequently, there is a need to develop cost effective procedures for evaluating state of the practice traffic signal control equipment so that informed deployment and design decisions can be made. In order to achieve that objective, this paper presents the motivation for using hardware-in-the-loop simulation procedures. Hardware-in-the-loop simulation presents a new set of challenges for traffic engineering model developers as the “correctness” of a real- time model not only depends upon the numerical computation, but the timeliness with which the simulation model interacts with external control equipment. This paper reviews the state of practice, summarizes the fundamental technologies necessary for implementing such a system, and uses a simple statistical test for assessing the real-time errors introduced into a simulation model. MOTIVATION Over the past half century, new traffic control procedures have been evaluated using macroscopic models, microscopic simulation models, or structured observation of field deployments. Macroscopic models, such as Transyst and Passer have been enormously useful for quickly evaluating benefits and designing fixed time signal plans [Roberston 69, Haenel 74]. More recently some packages offer optimization of basic actuated-coordinated controllers [Husch 00]. However, these models do not consider complex detector logic, shared right of way with light rail, or a myriad of control parameters available on modern traffic signal controllers. This discontinuity between macroscopic models and actual field equipment often leads to performance discrepancies when timing plans developed with macroscopic models are deployed. This discontinuity is particularly evident with emerging systems implementing concepts such as transit priority or adaptive control. Alternatively, some agencies do not use any models, electing instead to directly modify signal timings in the field and directly observe the results of these changes. Such procedures can be quite effective, particularly for tuning splits and offsets to achieve local optima. These procedures are much less effective for evaluating alternative cycle lengths and the quality of timings plans developed is largely due to the diligence of the technician doing the work. Furthermore, because of the risks of making a big mistake that could lead to gridlock, agencies rarely try creative or innovative timing plans and only find “local optima” with these procedures. In an attempt to provide a more realistic and uniform modeling procedure for engineers to evaluate alternative timing plans, microscopic simulation models have evolved over the last three decades [Farr 78]. In comparison to direct field observations, the current generation of microscopic models provide engineers with intersection level animation that is as good, if not better. However, because of the competitive market for traffic signal controllers, each vendor uses different procedures and parameters for configuring their traffic control equipment. Consequently, the current generation of microscopic simulation models do not cover the full range of features available in modern traffic signal controllers. The only way to evaluate many of the emerging real time adaptive control algorithms is to deploy them on the street, and observe their performance. For obvious reasons, it is extremely difficult, and often impossible, to obtain statistically sound before and after comparisons using field observations. Furthermore, direct experimentation with the motoring public requires extreme caution, which precludes trying a variety of innovative control models. The following section further details this problem. 1 Associate Professor, School of Civil Engineering, Purdue Univesity, West Lafayette, IN 47907; 765/494-2226; FAX: 765/496-1105; darcy@purdue.edu. 2 Associate Professor, Department of Electrical Engineering, University of Idaho, Moscow, ID 3 Associate Professor, Department of Electrical Engineering, University of Idaho, Moscow, ID 4 Professor, Department of Civil Engineering, University of Idaho, Moscow, ID 5 Graduate Assistant, Department of Civil Engineering, University of Idaho, Moscow, ID