This video showcases generic off-roads autonomous driving, with obstacles avoidance in very tight regions. Our off-roads autonomous driving research and technology takes a giant leap with this demo, showcasing bi-directional negotiation on single lane roads -- currently via novel decision making and motion planning algorithmic frameworks we have developed.
Our autonomous vehicle had to deal with stochastic traffic, and the typical driving behaviour customary to the Indian context. While doing so, the vehicle also also to deal with rough and bumpy terrain, and often had to deal with a few 10s of centimetre margin while avoiding obstacles.
This is one of a kind of a demo. Such capabilities have not been demonstrated off-roads before, by any company or startup anywhere in the world.
This research will further be scaled with unsupervised and reinforcement learning, to make autonomous off-roads autonomous driving map-less, without requiring any terrain map, and also eliminating explicit perception algorithms, like delimiters detection, road and navigable terrain detection, and obstacles detection.