OPTIMISED VEHICLE NAVIGATION IN CARLA

Abstract

The development of autonomous ground vehicles is a protractedly studied problem. One of the difficult scenarios is navigation in urban environments which are densely populated. A big hindrance in the research area of autonomous driving is cost of infrastructure required. Also, we incorporate the logistical issues for training and testing systems of physical world makes it even challenging. One way to resolve this issue is via simulation. We use CARLA (Car Learning to Act) simulator here. The simulator will give us an idea about the movement of our vehicle in various traffic scenarios. However, the approach of our vehicle looks optimal in choosing the path to the destination, but may consume more time. The challenge is to minimise the time with keeping the distance shorter as well. We proposed one such approach to Dynamic Lane Switching, where we make our vehicle took decisions in view of real world scenarios. We then look to compare our proposed approach with existing one in the CARLA Simulator.