Sensor fusion for vehicular networks

Abstract

Traffic congestion, CO2 emissions and road traffic fatalities are the three major concerns associated with current transport sector. It is identified that human factor is the cause of road accidents in 80 percent of the cases. There are many driver assistance systems but they are mostly autonomous. With these current autonomous systems, if some disturbance is created in one of the vehicles on the road, then there is an amplification of the disturbances (called as string instability or shockwaves) in the follower vehicles. Cooperative driving systems provide a promising solution to the string instability problem and help in increasing throughput, reducing traffic congestion, improving safety, lowering CO2 emissions and reducing fuel consumption. However, for large scale deployment of these cooperative systems, ensuring safe and reliable operation is a big challenge. Grand Cooperative Driving Challenge (GCDC) is a competition that aims to demonstrate that the cooperative systems work and achieve some specific objectives. The current master thesis is done as part of this GCDC competition. In the master thesis, a real time sensor fusion system is developed for the application of vehicle platooning (road trains). The task of the sensor fusion algorithm is to provide filtered signals to the controller and make sure the system is robust to sensor failures. A careful balance is done between the information from the wireless communications and in-vehicle sensors by analyzing the limitations of each sensor and complementing them with other sensors. To perform the task of fusing/combining the signals, an extended Kalman filter set up is used in the project. Algorithms that handle asynchronous sensor data, popularly called out-of-sequence measurement (OOSM), are studied and implemented. The sensor fusion system along with other blocks in the project were tested in real time for the GCDC competition and the results convey that the sensor fusion system is working.