Compliant Robot Actuation by Feedforward Controlled Emulated Spring Stiffness

Existing legged robots lack energy-efficiency, performance and adaptivity when confronted with situations that animals cope with on a routine basis. Bridging the gap between artificial and natural systems requires not only better sensorimotor and learning capabilities but also a corresponding motion apparatus and intelligent actuators. Current actuators with online adaptable compliance pose high requirements on software control algorithms and sensor systems. We present a novel feedforward trajectory shaping technique that allows for a virtual stiffness change of a deployed series elastic actuator with low energy requirements. The performance limits of the approach are assessed by comparing to an active and a passive compliant methodology in simulation. For this purpose we use a 2-degrees-of-freedom arm with and without periodic load representing a 2segmented leg with and without ground contact. The simulation results indicate that the approach is well suited for the use in legged robots.