Trajectory Planning with non-holonomic constraints and obstacle avoidance

Autonomous vehicles operating in the real world will not always know about all obstacles in the environment. Thus, the vehicle must have the ability to plan and re-plan trajectories around known obstacles and emerging obstacles in the environment. Currently, there are many trajectory search algorithms for autonomous driving vehicles. In this paper, we explore the most prominent trajectory search algorithms like RRT, A*, and R*. Our goal is to implement these algorithms to compare them with one another and determine their strengths and weaknesses. We also implement a minor extension to efficiently update the RRT in case of new obstacle information without re-planning from scratch.

Demo running live on the server can be accessed here (might take a minute to spin up the server)