Information Gathering Planning with Hermite Spline Motion Primitives for Aerial Vehicles with Limited Time of Flight

Abstract

This paper focuses on motion planning for information gathering by Unmanned Aerial Vehicle (UAV) solved as Orienteering Problem (OP). The considered OP stands to find a path over subset of given target locations, each with associated reward, such that the collected reward is maximized within a limited time of flight. To fully utilize the motion range of the UAV, Hermite splines motion primitives are used to generate smooth trajectories. The minimal time of flight estimate for a given Hermite spline is calculated using known motion model of the UAV with limited maximum velocity and acceleration. The proposed Orienteering Problem with Hermite splines is introduced as Hermite Orienteering Problem (HOP) and its solution is based on Random Variable Neighborhood Search algorithm (RVNS). The proposed RVNS for HOP combines random combinatorial state space exploration and local continuous optimization for maximizing the collected reward. This approach was compared with state of the art solutions to the OP motivated by UAV applications and showed to be superior as the resulting trajectories reached better final rewards in all testing cases. The proposed method has been also successfully verified on a real UAV in information gathering task.