A Limb Compliant Sensing Strategy for Robot Collision Reaction

This paper introduces a compliant limb sensor (CLS) concept for collision detection during robot-human contact. The CLS consists of an external rigid shell compliantly connected to the robot link with collision inferred from measured shell displacements. Measuring displacement of a rigid shell allows customizable compliance and high sampling rates due to the small number of required sensors. The proposed sensor is prototyped for the planar case using LED/light-to-voltage (LTV) sensors for shell pose measurement and foam as the compliant link between the shell and base. A physically motivated model for the output of LED/LTV sensor pairs is formulated for the estimation of the shell pose. Voltage measurements of redundant LTVs and a calibrated shell model are used with an iterative optimization routine to estimate the shell pose at high frequencies. Sensor performance is tested using five trajectories: rest, compression, shear, rotation, and arbitrary motion. Experiments confirmed that the CLS can sense the presence, direction, and intensity of impact. The potential application of the proposed sensor to safety in physical human-robot interaction is discussed. The novel sensing methodology also enables a new method of 3-D human-computer interaction due to the ability to modify the compliance and operating range of the CLS.