GTO-MPC Based Target Chasing using a Quadrotor in Cluttered Environments

Abstract

This paper addresses the challenging problem of chasing an escaping target using a quadrotor in cluttered environments. To tackle these challenges, we propose a guided time-optimal model predictive control (GTO-MPC) based practical framework to generate chasing trajectories for the quadrotor. A jerk limited approach is first adopted to find a time-optimal jerk limited trajectory (JLT), an initial reference for the quadrotor to track, without taking into account surrounding obstacles and potential threats. An MPC based replanning framework is then applied to approximate the JLT together with the consideration of other issues such as flight safety, line-of-sight maintenance, and deadlock avoidance. Combined with a neural network, the proposed GTO-MPC framework can efficiently generate chasing trajectories that guarantee flight smoothness and kinodynamic feasibility. Our simulation and actual experimental results show that the proposed technique is highly effective.