Co-operative Driving at Intersections using Vehicular Networks and Vehicle-Resident Sensing
In order to distinguish essays and pre-prints from academic theses, we have a separate category. These are often much longer text based documents than a paper.
Road intersections are considered to be serious bottlenecks of urban transportation. According to the U.S., European, and global statistics, intersection-related crashes, with fatal outcome, represent approximately 20 percent of all traffic fatalities. More than 44% of all reported crashes in the U.S. occur within intersection areas, which, in turn, lead to 8,500 fatalities and approximately 1 million injuries every year . Furthermore, the impact of road intersections on traffic delays leads to enormous waste of human and natural resources. Statistics collected by FHWA in 2011 urban mobility report states that the average intersection delay endured by a commuter is 34 hours every year. The cost of this wasted time and related fuel consumption at intersection congestions is over $101 billion a year . Therefore, it is critical to address these safety and throughput concerns as one of the main challenges for manual as well as autonomous driving through intersections. This dissertation studies the problem of managing traffic through intersections, and develops new decentralized, reliable and efficient active safety methods to provide safe and efficient passage through intersections and roundabouts. Our cyber-physical framework called STIP (Spatio-Temporal Intersection Protocols) incorporates a fusion of vehicle-to-vehicle (V2V) communications, vehicle-to-infrastructure (V2I) communications and vehicle-resident sensing to enable co-operative driving of autonomous and manually-driven vehicles at intersections. The proposed system allows vehicles to traverse safely by avoiding vehicle collisions at intersections, and significantly increase traffic throughput. The STIP framework includes a family of distributed protocols and covers the following two main traffic environments, categorized based on the market penetration of autonomous vehicles: (1) Homogeneous Traffic: autonomous vehicles only, (2) Heterogeneous Traffic: mix of human-driven and autonomous vehicles. For the homogeneous traffic category, we introduce two sets of intersection protocols: (1) V2V-Intersection Protocols, which rely on V2V communications and localization to avoid vehicle collisions at intersections by controlling and navigating them within the intersection area. Autonomous vehicles approaching an intersection use DSRC to periodically broadcast information such as position, heading and intersection crossing intentions to other vehicles. The vehicles then decide among themselves regarding who crosses first, who goes next and who waits. (2) Synchronized movement-intersection protocols, which are designed to increase the parallelism at intersections by allowing the concurrent crossing of vehicles arriving from all directions. This method enforces synchronized and staggered arrival of vehicles at intersections. This method allows vehicles to cross the intersection without stopping or slowing down, and maximizes the capacity utilization of the intersection space. In case of the heterogeneous traffic category, in order to enable the safe co-existence of manually-driven and autonomous vehicles at intersections, we propose a set of communication-based and perception-based protocols, leveraging a fusion of V2V, V2I and on-board sensor systems such as cameras, radars and lidars. In this dissertation, we formally prove the deadlock-freedom property of our family of intersection protocols, and study the effects of packet loss on our V2V intersection protocols and measure the reliability of these protocols in the presence of channel impairments. We also measure the impact of position inaccuracy of commonly-used GPS devices on our V2V-intersection protocols and incorporate required modifications to guarantee their safety and efficiency despite these impairments. Additionally, we study sensor inaccuracy and occlusion’s impact on our perception-based intersection protocols, and propose simple solutions to deal with these shortcomings. The functionality of our methods is evaluated using our vehicular networks emulator-simulator, called AutoSim. Our results indicate that our proposed STIPs provide both safe passage through the intersection and significantly decrease the delay at the intersection by increasing the concurrency. Specifically, one of our V2V-intersection protocols yields over 87% overall performance improvement over the common traffic light signalized intersections. Throughput increases are even more significant in the case of our synchronized movement intersection protocols, as intersection delays are reduced up to 96% compared to the common traffic light signalized intersections, and the optimal intersection capacity utilization of 100% is achieved in certain cases.