Henrik Rehbinder
Division of Optimization and Systems Theory
Department of Mathematics
KTH
E-mail: henrikr@math.kth.se
The seminar will focus on different aspects of nonlinear observer design for the specific orientation estimation problems that the thesis is concerned with. Particular attention will be paid to how the observer design problem is tied to the rigid body kinematics representation. It will be demonstrated how rate gyros and accelerometers can be fused in a particularly simple way by using the geometric structure of the underlying system. This is compared to a more standard high-gain observer that does not use this structure. The geometric way of viewing these problems is pushed further in a study of computer vision where it is shown how observability of linear systems is naturally generalized to an implicit output system evolving on $SO(3)$. Finally, some implementational issues regarding the computational constraints of computer vision is considered.