Incidents on the freeway: Detection and management

Incidents on the freeway disrupt traffic flow, cause significant delay to motorists and present serious safety hazards while they are being cleared. These factors create the need to quickly and efficiently manage and contain incidents on the freeway. In this dissertation, we explore two common ways to manage incidents: automatic incident detection systems and Freeway Service Patrol tow trucks.; Quick and efficient automatic incident detection has been an elusive goal of the transportation research community for many years. But it is the linchpin of future transportation management systems. In this dissertation we review the field of incident detection and we present and explore a novel probe vehicle-based incident detection algorithm. We present ideas for incident detection via sensor fusion and we explore a framework for evaluating incident detection systems in operation.; The second incident management technique we study, Freeway Service Patrol (FSP) systems, consist of a collection of rotating tow trucks on the freeway that assist motorists involved in accidents or breakdowns. Many claims have been made about the benefits of FSP systems on freeway systems and we examine these claims in detail. We present a method for determining incident delay from loop detector and/or probe vehicle data. We apply this method to evaluate the Bay Area FSP system and conclude that many of the claims regarding benefit-to-cost ratios of FSP systems in the literature are grossly overstated. The error is due to statistical misspecification. Finally, we present an empirical methodology for determining where to place the FSP tow trucks on a multi-beat urban freeway system.; This dissertation is essentially empirical and is based on the data from the I-880 FSP Evaluation Study. A main theme of this dissertation is that the proper collection, filtering and processing of data is essential to the study of incident management performance on the freeway. We illustrate this theme in many sections of this dissertation. We conclude by examining a few research areas which we feel need further exploration.