Optimization Based Control for Target Estimation and Tracking via Highly Observable Trajectories An Application to Motion Control of Autonomous Robotic Vehicles

This paper proposes a Model Predictive Control (MPC) scheme to solve the target estimation and tracking problem. The objective is to derive a feedback law that drives an autonomous robotic vehicle to follow a target vehicle using an on-line estimate of the target's state. In this scenario, when the target is observed through a nonlinear observation model, e.g., bearing only or range only sensors, it is possible to show that solving the tracking problem independently from the estimation problem can lead to an unsatisfactory result where the follower-target system is driven by the controller through unobservable or weakly observable trajectories and, as result, the state of the target vehicle cannot be recovered or cannot be recovered with high accuracy leading to the failure of the control strategy. In this paper, we propose an optimization based scheme that embeds, in a seamless way, an index of observability in the design of the target tracking controller resulting in a closed loop behavior that balances the objective of target tracking with the competing objective of maintaining a good estimate of the state of the target. Numerical results are presented that illustrate this type of behavior.