The Foundation for Tactile Robots: Approaching the Holistic Analysis of a Robot's Force Sensing Capabilities

Abstract

Contact estimation and force sensing are fundamental requirements for sensitive manipulation and safe physical human-robot interaction. The robot controllers that enable these functions rely on accurate and precise sensing. The performance of external force estimation is influenced by the design of the robot’s sensory system. And similar to how humans prefer specific arm configurations for performing precise and delicate tasks, e.g., drawing a thin, straight line, robots also have "sweet spots" that allow for the most accurate performance of tasks based on their sensing capabilities. To fully exploit a robot’s proprioceptive force sensing, it is essential to provide robot integrators, designers, and simulations with knowledge about these optimal settings including factors such as joint configurations, temperatures, and many more. This paper first investigates which of these factors are most relevant and how they can be best measured and based on that introduces force sensing error maps as a tool for structured research on robot force sensing performance and future developments of tactile robot applications. We first investigate the factors influencing the force sensing performance of 7-degree-of-freedom robots on the example of a Kinova Gen3 and then derive 2-dimensional Cartesian force sensing error maps for this robot, an LWR iiwa 14, and a Franka Emika robot. These maps enable comparison of robot sensing capabilities, revealing patterns and weak spots to guide application design toward more tactile areas.