RRT* Trajectory Scheduling Using Angles-Only Measurements for AUV Recovery

Abstract

Sensor trajectory optimisation involves extensive search over the sensor motion space against an optimisation criterion. The search under dynamic programming or fixed grid is often computationally nontrivial even for a myopic search scenario. In this paper, we study the problem of an autonomous underwater vehicle planning its return route to a moving recovery vessel. To complicate the issue, the AUV needs to localize the vessel using angle-only measurements. Accordingly, we propose a random sampling based trajectory planning algorithm that incorporates both a dynamic goal and the need to localize that goal. More precisely, we incorporate an information theoretic cost into a rapid-exploring random tree trajectory planning framework thus allowing the AUV to both localize and reach the recovery vessel. Our experimental results show that the proposed method may achieve the same trajectory optimisation performance as that under dynamic programming method but with greater computational efficiency.