Smooth Trajectory Optimization in Wind: First Results on a Full-Scale Helicopter

Abstract

A significant challenge for unmanned aerial vehicles is flying long distances in the presence of wind. The presence of wind, which acts like a forcing function on the system dynamics, significantly affects control authority and flight times. While there is a large body of work on the individual topics of planning long missions and path planning in wind fields, these methods do not scale to solve the combined problem under real-time constraints. In this paper, we address the problem of planning long, dynamically feasible, time-optimal trajectories in the presence of wind for a full-scale helicopter. We build on our existing algorithm, κITE , which accounts for wind in a principled and elegant way, and produces dynamically-feasible trajectories that are guaranteed to be safe in near real-time. It uses a novel framework to decouple path optimization in a fixed ground frame from velocity optimization in a moving air frame. We present extensive experimental evaluation of κITE on an autonomous helicopter platform (with a human safety pilot in the loop) with data from over 23 missions in winds up to 20m/s and airspeeds up to 50m/s. Our results not only shows the efficacy of the algorithm and its implementation, but also provide insights into failure cases that we encountered. This paves the way forward for autonomous systems to exhibit pilot-like behavior when flying missions in winds aloft.