Visual Foresight With a Local Dynamics Model

Colin Kohler, Robert Platt

Model-free policy learning has been shown to be capable of learning manipulation policies which can solve long-time horizon tasks using single-step manipulation primitives. However…