Limb Movement in Dynamic Situations Based on Generalized Cognitive Maps

The fundamental bases of how our brain solves different tasks of object manipulation remain largely unknown. Here we consider the problem of the limb movement in dynamic situations on an abstract cognitive level and propose a novel approach relying on: i) transformation of the problem from the limb workspace to the so-called hand-space, and ii) construction of a generalized cognitive map (GCM) in the hand-space. The GCM provides a trajectory that can be followed by the limb, which ensures an efficient collision-free movement and target catching in the workspace. Our numerical simulations confirm the approach feasibility but also reveal the problem complexity. We then validate the GCM-based solutions in real-life scenarios. We show that a GCM-equipped humanoid robot can catch a fly ball in a similar way as a human subject does. The static nature of the GCMs enables learning and automation of sophisticated cognitive behaviors exhibited by humans.