Cooperative target tracking for heterogeneous robots

This work deals with the motion planning of multiple robotic agents that, equipped with sensors with limited field of views, have the task of tracking an unknown target. A Nonlinear Model Predictive Control strategy is used to minimize the entropy of the estimation of the target’s state provided by an intermittent measurement Kalman filter. The scheme is validated by first implementing the controller on a team of robots comprised of an aerial and a ground vehicle, modelled with double-integrator dynamics in Matlab-Simulink. Then, by using the same NMPC scheme as a trajectory generator, the aerial vehicle is modelled as a quadrotor UAV following with a low-level controller the high-level planner’s trajectories, while tracking the target, another quadrotor UAV. To do so, the Matlab-Simulink simulations are coupled with Gazebo, a 3D dynamic simulator, and Genom, a tool for real-time software architectures.