Speaker: 易圣伦，国防科技大学

Title: A Game-Theoretic Approach to Robust State Estimation

Language: Chinese

Time & Venue: 2026.11.05 15:00-16:00 Zoom会议：64135181073

Abstract: Accurate state estimation is crucial in control, signal processing, and machine learning, with applications in navigation, robotics, finance, and environmental monitoring. Classical filters, such as the Kalman filter, assume precise knowledge of the underlying state-space model. In practice, however, models are often uncertain or only approximately known, which can render standard filters unreliable or unstable. In this talk, I will present a robust state estimation framework based on a minimax, game-theoretic approach. This perspective treats model uncertainty as an adversary and seeks estimators that minimize the worst-case error. I will first introduce the methodology in the linear setting, then discuss extensions to nonlinear systems. Finally, I will outline strategies to learn the level of uncertainty from data, bridging worst-case robustness with data-driven adaptation, and illustrate their application to UAV navigation in complex environments.