Planification robuste pour systèmes robotiques
Srour, Ali · theses.fr (ABES) · 2024
A key challenge for autonomous systems is operating under real-world uncertainties. Robots rely on models of their environment and themselves for decision-making, but these models are inherently approximations. Consequently, uncertain parameters can lead to significant discrepancies between the intended and actual system behavior. This thesis addresses the issue of parametric uncertainties by developing trajectories that are intrinsically robust. Through optimizing these trajectories within the closed-loop system using novel concepts of state and input sensitivities introduced in this work, the approach enhances robot performance in uncertain conditions. The primary goal of this thesis is to extend and apply these sensitivity-based methods for robust trajectory planning. The validity of the proposed optimization framework is empirically assessed through extensive statistical campaigns, both in simulations and real-world experiments, on two widely used robotic platforms: a quadrotor drone and a robotic manipulator.