Optimal Pursuit with Rapidly Exploring Random Trees

The research addresses the problem of guiding a constant speed vehicle to a stationary target while enforcing a terminal heading constraint. A path that minimizes control effort and avoids obstacles scattered in the 2D environment is sought. A variation of the rapidly exploring random tree (RRT) algorithm that is probabilistic complete and is guaranteed to asymptotically converge to the optimal solution is utilized.
The talk will begin with an overview on incremental sampling-based RRT algorithms. Then, the problem at hand will be presented along with the solution approach consisting of two parts: with and without obstacles. The proposed connection methodology between two points sampled by the algorithm, using a linearization based closed-form guidance solution, will then be discussed. Presentation of methods to reduce computational effort using newly introduced lower bounds for the cost of feasible paths will follow. Analysis of simulation results and some discussion will conclude the talk.