Robots that collaborate with humans have been around for a while and are gaining in popularity in the recent years. An example of such robots are exoskeletons, which encase the human body and provide assistance to human motion.

Most of the current control methods focus on rehabilitation or power-augmentation, be it for medical, industrial or military use. In rehabilitation, the goal of the exoskeleton is to simply perform repetitive motion, replacing the work of a physiotherapist. In case of power-augmentation the exoskeleton robot needs to either help with the motion of the impaired or augment the capabilities of the able-bodied humans. This power augmentation is usually uniform and it doesn’t take into account the configuration of the human limb.

We propose a control method based on human muscular force manipulability. In robotics, force manipulability describes how the end-effector position and orientation can change in a given configuration of the robot. This manipulability measure is presented by an ellipsoid, whose major axis represents the direction in which the robot can exert the highest force. We humans however don’t have electric motors in our limbs, they are instead driven by muscles which wrap around the joint and are attached to the bone. Contracting the muscles creates a torque on the joint and consequentially a movement of the limb. Taking into account the Hill’s muscle model, which treats the muscle as a spring-damper system we then expand the force manipulability measure. This muscular force manipulability measure now describes how arm muscle forces relate to forces exerted by the hand in a given configuration of the human arm. This is then incorporated in an exoskeleton control method so that the exoskeleton provides more assistance in low manipulability direction in which the human can’t exert high forces, compared to the high manipulability direction in which the human is naturally stronger. This method effectively reduces the effort when the human performs tasks in the directions of low manipulability.