Selection of a Traffic Control Strategy for Long-Range Travel Forecasting

This paper addresses the problem of determining the number and placement of signals on traffic networks for long-range urban travel forecasting. An algorithm for determining the signalization strategy was developed and given a large-scale test on the network for a metropolitan area of about 150,000 people. The algorithm attempts to mimic the actions of traffic engineers as they make adjustment to the traffic system over a long period of time. Tests indicate that the algorithm produces a network that closely approximates one that has been optimized for vehicle hours traveled, but one that still respects safety and fairness issues. The algorithm is highly computational, so limits needed to be arbitrarily placed on the precision of the traffic forecast, the precision of optimization steps, and on the number of intermediate forecast years. INTRODUCTION AND RESEARCH OBJECTIVES There has been a growing recognition that travel forecasts should be properly sensitive to traffic flow and a longstanding realization that all alternative transportation plans should include, at the very least, all lowcost traffic engineering improvements, sometimes referred to as TSM (transportation system management) strategies. A large and important subset of TSM strategies concerns upgraded or optimized traffic control systems. A very difficult question arises when doing long-range transportation planning: What should be the design of the signal system when traffic patterns in the forecast year are estimated to be substantially different from today? By even asking this question, it is assumed that the travel forecasting methodology can differentiate between forms of traffic control (signs and signals), understand the various ways in which signals can be operated, and calculate delay at traffic controls in a manner consistent with good traffic engineering principles. In addition, there is an expectation that travel demand is fully responsive to and reflective of the delays experienced by drivers on the network, at a minimum affecting both the distribution of trips across the network and route choice. This question was and continues to be asked in the Cedar Rapids metropolitan area, where the Linn County Regional Planning Commission (LCRPC) is currently preparing the transportation plan for the metropolitan area though 2030. LCRPC already had a travel forecasting network containing explicit traffic controls for its base year (1, 2), but no formal mechanism beyond trial and error for determining the nature of traffic controls in future years. The network contains explicit representation of all signalized and stop-controlled intersections on the major street network. The problem with long-term traffic control is similar in concept to what has been identified as the network design problem (NDP) from the transportation research literature (3, 4). The NDP attempts to find the best combination of links or link capacities to satisfy a given demand. Long-range planning in urban areas should also find the best network for the demands by specifically changing the nature of traffic control. Finding the exact combination of traffic controls and control strategies to optimize a network for a whole metropolitan area would be enormously difficult. Consequently, this paper explores an algorithm that mixes optimization techniques with elements of standard traffic engineering practice and travel forecasting theory to obtain a desirable network. Several practical issues must be faced when building an algorithm to produce such a long-range design. 1. There is a strong relationship between the implementation of a traffic control change and the distribution of travel in response to such changes (e.g., drivers discouraged from making left turns onto busy streets off a two-way stop control would not be so discouraged if the control were changed to a signal or all-way stop). Horowitz and Granato 2 2. The design for the forecast year is dependent upon the order in which traffic control changes are implemented in previous years. 3. The design should be reasonably consistent with good traffic engineering practice that not only tries to attain the best performance of the traffic system but maintains a locally acceptable level of safety and tries to achieve a reasonable degree of fairness toward all drivers. 4. A critical element in the design decision is determining which of the many unsignalized intersections in the metropolitan area should get signals. The most authoritative source of guidance on judging when to upgrade signed intersections to signals is the Manual on Uniform Traffic Control Devices (5). 5. It is reasonable to pursue lowest cost strategies first, e.g., get the best performance out of the existing signals before attempting to add signals. 6. In a growing city, signals that were warranted and installed at one time are almost never removed. 7. The algorithm must be computationally tractable. There are a number of different measures of effectiveness that could be tested. In this research the algorithm attempts to achieve a low value for vehicle hours traveled (VHT), which would normally occur when delays at intersections are low. The next section describes an algorithm that creates an approximately optimal network for long-range travel forecasting and address the issues cited above. THE SIGNAL SYSTEM DESIGN ALGORITHM The signal system design algorithm is illustrated in Figure 1. Figure 1 contains two flow diagrams; the flow diagram labeled “Traffic Forecast” on the right occurs twice in the main flow diagram on the left. The traffic forecast is embedded into two loops, one that optimizes cycle lengths and a second one that selects unsignalized intersections for signalization. Trip Generation Select Forecast Year Optimize Cycle Lengths