Trajectory planning for autonomous nonholonomic vehicles for optimal monitoring of spatial phenomena

This paper considers optimal trajectory planning for autonomous nonholonomic vehicles used in investigating environmental phenomena. In particular, we present an algorithm that generates locally optimal trajectories to find the global maximum of the underlying environmental field. Our algorithm uses Gaussian process priors to estimate the unknown field and the notion of expected improvement to develop an objective function for optimal planning. Monte Carlo simulations focusing on two-dimensional spatial fields show the advantage of our algorithm at finding the global maximum over existing methods.