Multi-UAV Velocity and Trajectory Scheduling Strategies for Target Classification by a Single Human Operator

Abstract

This work addresses the problem of enabling a single human operator to individually inspect targets for a fixed amount of time in a reconnaissance mission. The task of the operator is to classify the targets as friends or foes in real time, as they appear in video feeds from multiple UAVs. In order to account for cognitive limitations, the human is modeled as a single processing unit that can only execute one task at a time. A task is defined as a target inside the field of view of a given UAV, that needs to be inspected. Under the assumptions of this model, a linear program (LP) formulation is used to optimally find each task’s arrival time and latency in the system such that the human operator can inspect each target individually for some time Δt. Previous work by the authors investigated the idea of using UAV velocity modifications to meet the timing schedule specified by the LP solution. In this paper, the idea of UAV trajectory changes is introduced by modeling the UAVs as Dubins vehicles. Modifications to the bounds on the LP constraints are derived based on Dubins trajectories. The new bounds ensure that the LP solution returns a timing schedule achievable via maneuvers that combine velocity and trajectory changes to the UAVs’ flight plans. An on-line algorithm is developed that constructs and commands these velocity and trajectory changes in real time when conflicts arise. Correctness properties of this algorithm are analyzed and discussed for mission scenarios where the location of the targets is unknown and targets are discovered by the UAVs in real time.