Deliberative optimization of reactive agents for navigation

A simple reactive steering behavior for agents navigating through a field that influences their motion is presented along with a deliberative method to optimize the parameters of this behavior. The behavior is described in the context of its application to unmanned maritime vehicle navigation, and the results of three simulation-based experiments are discussed. These simulated tests include two within a slow, uniform, steady-state velocity field, and one within a fast, non-uniform, time-variant velocity field. The steering behavior is compared with simple homing and tracking behaviors, and outperforms both in the case of a high-speed, variable current field. In the other current fields, the optimized steering behavior performs better on average than the homing behavior but not as well as the tracking behavior. A navigation system based on the optimized steering behavior was implemented aboard a small-scale autonomous surface vehicle (ASV), which was deployed on a series of field trials in the Eau Gallie River and the Indian River Lagoon. These field tests consisted of simple point-to-point navigation tasks, and in all cases the ASV reached its objectives.