Robust Planning for Unmanned Underwater Vehicles

In this thesis, I design and implement a novel method of schedule and path selection between predetermined waypoints for unmanned underwater vehicles under uncertainty. The problem is first formulated as a mixed-integer optimization model and subsequently uncertainty is addressed using a robust optimization approach. Solutions were tested through simulation and computational results are presented which indicate that the robust approach handles larger problems than could previously be solved in a reasonable running time while preserving a high level of robustness. This thesis demonstrates that the robust methods presented can solve realistic-sized problems in reasonable runtimes – a median of ten minutes and a mean of thirty minutes for 32 tasks – and that the methods perform well both in terms of expected reward and robustness to disturbances in the environment. The latter two results are obtained by simulating solutions given by the deterministic method, a naive robust method, and finally the two restricted affine robust policies. The two restricted affine policies consistently show an expected reward of nearly 100%, while the deterministic and naive robust methods achieve approximately 50% of maximum reward possible.