GRADIENT LIKE OBSERVERS FOR INVARIANT SYSTEMS
12 septembre 2008, Salle L105, à l'Ecole des Mines, Paris.
14h00 : Robert MAHONY, Department of Engineering, Australian National University.
This talk considers the design of non-linear observers for invariant kinematic systems posed on finite-dimensional connected Lie groups. Both the case of full-state measurements and partial-state measurements are considered with the measurement process modelled by a transitive group action acting on a manifold termed the output space. Concepts of synchrony (between observer and the plant) and invariant innovations are used to provide a structure theory for observer design. A gradient like construction is proposed to design the innovation terms, leading to strong (almost) global convergence properties for observation error. Such a construction requires a cost function with certain invariance properties and I give a short discussion of how such functions can be generated. The proposed approach has been successfully applied to the design of attitude filters on aerial robotic vehicles.
14h00 : Robert MAHONY, Department of Engineering, Australian National University.
This talk considers the design of non-linear observers for invariant kinematic systems posed on finite-dimensional connected Lie groups. Both the case of full-state measurements and partial-state measurements are considered with the measurement process modelled by a transitive group action acting on a manifold termed the output space. Concepts of synchrony (between observer and the plant) and invariant innovations are used to provide a structure theory for observer design. A gradient like construction is proposed to design the innovation terms, leading to strong (almost) global convergence properties for observation error. Such a construction requires a cost function with certain invariance properties and I give a short discussion of how such functions can be generated. The proposed approach has been successfully applied to the design of attitude filters on aerial robotic vehicles.