On simulation-based adaptive UAS behavior during jamming

Abstract

We address the issue of autonomously planning a flight path for a remotely controlled surveillance aircraft when control is lost due to jamming. An optimization problem arises where we want the aircraft to continue surveying in the jammed area so that a possible attack does not go unnoticed, but we want the aircraft to leave the jammed area (and report any collected information) while there is still time to respond and take defensive measures. We formulate this as a stochastic approximation problem, involving state parameters and a discrete set of path options (specifying candidate Bezier curves), and train on simulated data from realistic scenarios. The result is a discussion how to acquire a policy which considers both realistic tactical scenarios with varying initial values and simulated sensor characteristic.