Location of vehicle identification sensors to monitor travel-time performance

Traffic agencies are increasingly using automatic toll collection technologies to collect fees on toll roads. Associated with such technologies are toll tags on vehicles that identify the vehicle and its account for charging the appropriate toll, and roadside readers (or overhead readers) that read the toll tags. A by-product of such technologies is the data collected on traffic measures such as travel times from the point when the vehicle is first tagged to the point it leaves the system. That is, the toll reader becomes a traffic sensor that measures travel times of some tagged vehicles and volumes of such vehicles that go from intermediate origins to intermediate destinations. Two problems have been studied on the optimal locations of such automatic vehicle identification (AVI) readers on traffic network using criteria: (i) to maximise the total vehicle-miles monitored, and (ii) to minimise the variance of predicted travel times. Optimisation models are formulated through which it is shown that these problems are NP-hard. Subsequently, greedy heuristics are proposed. Exact optimal solutions, using CPLEX, and approximate solutions, using the heuristics are compared for several small hypothetical networks for a case study dealing with a real application as well as in Harris County, Texas, USA.