An optimization based view on robot dynamics

Mathematical optimization has turned out to be a very powerful tool in robotics. The robotics community is increasingly subscribing to the paradigm of treating every robot task as a constrained optimization problem. This becomes especially important when we want to achieve dynamic control of robots and produce motions which look very natural to their biological counterparts. The aim of this talk is to provide a holistic optimization based view of robot dynamics which combines three important insights. First, taking inspiration from analytical mechanics (in particular Lagrangian formulation), it will be shown how our universe is a natural optimizer and how our laws of classical physics are a result of this optimization process. Second, building dynamic and energy efficient robots requires an optimal design where its natural dynamics is in harmony with the task's dynamics. Third, it will be argued that optimal design is not enough alone, we also need optimal control which exploits dynamics of the robot instead of cancelling it out.