Automatic motion generation exploiting the global structure of nonlinear dynamics based on finite-time reachability

In this paper, we propose a novel approach for motion generation in robotic systems based on the notion of finite-time reachability. The key idea is to segment the state space in terms of the reachability of particular target regions in finite time. The advantage of this approach is that precisely controlled joint trajectories are not necessary (as is often the case in traditional robotics). Complex movements can be generated by defining a hierarchy of controllers responsible for reaching multiple transitional regions toward the target region. Our method can simultaneously generate movements and the control for nonlinear robot systems. This method generates efficient motions exploiting the intrinsic characteristics of the mechanism and the environment. We present experimental results for two tasks: a pendulum swing-up and balancing problem and a bipedal walking problem.