For stabilization of a robot manipulator upon collision with a stiff environment, a nonlinear bang-bang impact controller is developed. Under this control, a robot can successfully achieve contact tasks without changing control algorithm or controller gains throughout all three modes: free space, transition and constrained motion. It uses a robust hybrid impedance/time-delay control algorithm to absorb impact forces and stabilize the system. This control input alternates with zero when no environment force is sensed due to loss of contact. This alternation of control action repeats until the impact transient subsides and steady state is established. After impact transient, the hybrid impedance/time-delay control algorithm is utilized. This bang-bang control method provides stable interaction between a robot with severe nonlinear joint friction and a stiff environment and achieves rapid response while minimizing force overshoots. During contact transition, we employ one simple control algorithm with same gains that switches only to zero, while other controllers use more than one control algorithms or different control gains. It is shown via experiments that overall controlled performance is superior or comparable to existing impact force control techniques.

In this paper, we present a method for motion generation from automatically derived behaviors for a humanoid robot. Behaviors are derived automatically by using the underlying spatio-temporal structure in motion. The derived behaviors are stored in a robot's long-term (or procedural) memory. New motions are generated from the derived ones with a search mechanism. In our approach, vision, speech recognition, short-term memory and decision-making operate in parallel with long-term memory in a unique architecture. This organization is intended for autonomous robot control and learning.