Robust on-line routing in intelligent transportation systems

This work provides a robust and efficient real-time route planning (RTRP) framework which produces dynamic path strategies based on both historical and real-time travel conditions. Such a framework is especially useful in the context of Intelligent Transportation Systems (ITS), where Information Service Providers (ISPs) collect and process real-time traffic data from transportation systems and broadcast general travel information (e.g. travel speed, congestion and incidents) to a population of motorists or provide personal routing instruction (e.g. optimal route) to individual travelers. The RTRP framework is intended for use by the ISPs to generate such personal routing instruction based on traffic data (historical and real-time). In this work, both the conceptual RTRP framework and its specific algorithmic components are provided.; The contributions of this work are summarized as follows. (1) A RTRP framework is presented, which will produce robust dynamic en route path strategies. (2) The RTRP framework permits the use of only a subset of network arcs for collecting real-time information and updating future travel time PDFs, offering tremendous flexibility in implementing the framework. (3) The essential properties of ICAs are discussed and a heuristic algorithm is presented for determining the ICAs. (4) In the RTRP framework, any appropriate forecasting and updating procedures can be embedded, making this framework practicable in various environments. (5) A mechanism based on the information discounting strategy is presented for employing real-time information to update future travel time PDFs. (6) Specific computational steps are given for producing routing suggestions in signalized networks, where signal timings are known with certainty or only probabilistically. (7) A reoptimization approach and corresponding modified algorithms are presented for efficiently deriving solution paths from previous solutions and current travel time predictions in both unsignalized and signalized networks. The reoptimization-based RTRP framework is faster in average computational performance and more suitable for on-line application. This reoptimization technique can also be applied to networks where travel times are deterministic, time-invariant, or both.