Abstract:

Perception systems for acquiring environment information and control systems for commanding a system operating in the environment are usually separately designed, but their performance is often interdependent: control decisions affect perception, e.g., the distance from an object affects its sensing, and perception affects control decisions, e.g., ar- eas may be traversed or avoided based on the amount of available information. Perception-aware control accounts for such an interdependence to optimize the overall performance. We recently proposed a perception-aware chance constrained MPC (PAC-MPC) that considers the impact of control on the evolution of the environment uncertainty, then used within chance constraints. In this paper we obtain a stabilizing design the PAC-MPC, by first determining stability conditions in a general nonlinear setting, and then deriving specify design rules for the linear-Gaussian case, which results in a specific choice of the cost function parameters and in design conditions for the estimation algorithms that determine uncertainty propagation. The results are illustrated by means of a numerical example.