Minimizing redundant communication in vehicle networks for city mapping

Abstract

Modern cars are increasingly equipped with numerous sensors, as well as the ability to send this sensor data over the network to a remote server. This presents the opportunity for a variety of applications that make use of real-time vehicle sensor data. Combined with advances in real-time object detection, this enables real-time updates from vehicle networks that can monitor just about anything on or near the roads. These updates can cover time-sensitive situations as mundane as identifying traffic light outages to events as critical as amber alerts, where real-time updates on a vehicle's location could save lives. With many vehicles frequently passing by the same objects, these vehicles are bound to accumulate large amounts of duplicate data. While storage is often considered an area of interest for high-volume vehicle data, network transmission cost is often even more expensive, meaning that minimizing redundancy at the network communication level presents an opportunity to significantly mitigate the cost of duplicate data, allowing such applications to be economically feasible. Therefore, this thesis addresses the problem of determining how to minimize redundant object reports from vehicles to a common coordinator. The thesis presents and evaluates several protocols for this purpose, and demonstrates that with location-based protocols it is possible to realize substantial data overhead reductions when compared to simplistic protocols, thereby showing that optimizing network communication protocols to reduce data overhead presents the opportunity for significant financial savings.