Abstract
Our research primarily focuses on developing mechanisms to intelligently route autonomous vehicles in an urban area while managing to reduce traffic congestion. We propose an Intelligent Transportation System for Autonomous vehicles based on Multi-Agent paradigm. These agents observe traffic conditions over a period of time. Historical traffic information along with information of factors that affects traffic(like weather conditions) are used to develop a regression based learning algorithm that predicts time required to traverse a route for a given time-slot, which can be used to find out shortest path for a given source-destination pair, before beginning to travel. This estimated traverse time for every lane is communicated to the vehicles. Based on this information, the vehicles find out k shortest paths. Reservation based protocols are used to obtain a single path such that the overall congestion is minimized and the destination is reached in the shortest possible time. The shortest path obtained is then traversed with the help the vehicle control system with optimized fuel efficiency, ensuring vehicle stability and obstacle avoidance. To achieve this path trajectory, we used nonlinear Model Predictive control (MPC) approach based on bicycle model of vehicle dynamics. The main advantage of MPC is the fact that it allows the current time-slot to be optimized, while taking future time-slots into account. We further investigate the linear MPC strategies for Co-operative adaptive Cruise control which directly minimizes the fuel consumption.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call