Activity: Participating in or organising an event › Participation in workshop, seminar
Novel exoskeletons are continuously growing in complexity to try to match the inherent complexities of human natural walking and improve physical human robot interaction. In order to improve transparency and performance, exoskeleton designers are are now moving towards machines with several degrees of freedom, compliant actuators, proximally located actuation systems. Although those new designs may serve for improving the overall performance, ergonomics and safety of the exoskeletons, most of them come at a cost with added challenges from the control engineering point of view. Non-linearities, flexibility, backlash and complex geometrical models make clasical controllers and modeling techniques fall short. In this workshop we want to share our experiences on how did we used machine learning to compensate for a system with complex dynamic behavior and how that strategy can be extended to be integrated into a generic robot control architecture.