Stochastic Optimization for Motion Planning of Helicopter and Underwater Vehicles Reposition Task

We present a stochastic optimization based motion planner for use in aerial and underwater vehicles. Our approach draws from state of the art research in path planning for robotic arms, STOMP, in order to develop a motion planner which generates smooth feasible trajectories for repositioning tasks. We refer to repositioning as helicopter air-taxi or hover-taxi operation which is a preferred way of flight when movements require within an airport while avoiding flight over other aircraft, vehicles, and personnel. Our method, which we coin STOMP-RP (RP for RePosition), can be used in realistic scenarios where the vehicle can only view a small subset of obstacles due to sensor range limitations and can be used in dynamically changing environments. STOMP-RP is a practical motion planner for repositioning because (a) it generates paths that do not need to be post processed by a smoother, (b) optimizes trade-offs between path smoothness and obstacle clearance, and (c) can optionally be given an initial global plan path to be locally optimized. Additionally, STOMP- RP iteratively returns increasingly optimized paths which makes it desirable for real-time use.