Shawn Ge's Talk on Equivariant Observer Design for Discrete-time Systems

Title: Equivariant Observer Design for Discrete-time Systems

Time, Date, Location: 11:00, Friday, 10 February, 2023, Brian Anderson Building Seminar Room

Abstract: State observer design for nonlinear systems is a key field in control theories. Various techniques have been proposed to solve this state estimation problem, such as Luenberger-like observers and extended Kalman filter, which are also widely used in real world systems. In the last 20 years, driving by the unmanned aerial vehicle industry, there has been a burst of works in novel geometric observer design for attitude and pose estimation by exploiting the symmetry of the systems. This leads to my research topic of equivariant observer designs. In this Thesis Proposal Review, two papers are presented. One proposed a general filter design methodology of discrete-time systems on homogeneous spaces, while the other one presented how this filter can be applied to inertial navigation systems. Additionally, plan for future work that extends the current filter architecture are detailed with short-term plans, both coming with derivations. The future direction addresses how I would like to generalize the current work onto a broader class of systems, with some early-stage insights of the ideas. The resources and logistics required for the future plans are provided.