Response surface methodology applied to toll plaza design for the transition to electronic toll collection

Electronic toll collection (ETC) holds the promise of greatly reduced congestion at toll roads, bridges, and tunnels. Using automatic vehicle identification (AVI), vehicles are charged a toll as they drive non-stop through a tollbooth. Since a transition to the AVI-dominated toll plaza cannot occur overnight, a plan for the increasing use of AVI tollbooths that does not increase congestion for non-AVI vehicles is needed. Experiments using a simulation model of a typical toll plaza with varying mixes of vehicle types and tollbooth allocations provided data for four output measures: vehicle volume, average queue length, average waiting time, and tollbooth utilization. Multiple linear regression analysis was then applied to fit response surfaces for each measure. The response surfaces provided the optimum tollbooth allocation for a given mix of vehicles. Two decision rules were evaluated based on their ability to specify near-optimal tollbooth allocations for varying mixes of vehicle types using vehicle volume as the output measure.