Model Predictive Control for Target Tracking in 3D with a Downward Facing Camera Equipped Fixed Wing Aerial Vehicle

In this paper, we consider the problem of tracking a ground vehicle with a fixed-wing aerial vehicle (FWV) equipped with a downward-facing camera. The complexity of the problem stems from the highly nonlinear kinematics of the FWV and the stall speed constraint. We propose a Model Predictive Control (MPC) approach for this problem that has two main contributions. Firstly, we model the tracking requirement through a novel constraint function that relates FWV’s position and orientation to the field of view of the camera. Secondly, we make a case for reformulating the underlying optimization of the MPC as an unconstrained problem and solving it through the state of the art gradient descent variants like ADAM and RMSProp. Specifically, we show the real-time performance of this optimizer while achieving good tracking performance under various kinematic constraints. We validate our MPC through extensive simulations, specifically highlighting the 3D spiral-like trajectories obtained for the FWV when tracking a slow-moving ground vehicle. We also present a quantitative analysis of the efficacy of the different gradient descent variants.