Control of Contact Forces using Whole-Body Force and Tactile Sensors: Theory and Implementation on the iCub Humanoid Robot

Abstract

The new generation of service robots requires force control to safely cooperate with humans in unstructured dynamic environments. The objective of this thesis was to improve the abilities of humanoid robots to interact with their surrounding. In particular, the idea of integrating force and touch feedback was investigated, so as to control the robot in presence of external forces acting on any part of its body. To this aim, this work exploited the humanoid robot iCub as test platform, and it tackled three main issues: i) spatial calibration of tactile sensors, ii) estimation of contact forces using tactile sensors and force/torque sensors, and iii) prioritized position and force control. The main contribution of this thesis is a new framework for prioritized position and force control of floating-base robots. The framework was compared to other state-of-the-art similar frameworks, both analytically and in simulation, and it proved preferable in terms of optimality and computational efficiency — twice as efficient, while preserving the optimality of the solution. Moreover, a method for estimating the 3D positions of tactile sensors was proposed. The method relies on force/torque measurements and it was exploited to calibrate the 1500 tactile sensors mounted on the arms of the iCub robot, with an average error of approximately 7 mm. Another method was introduced, which makes use of the calibrated tactile sensors, together with the distributed force/torque sensors, to estimate an arbitrary number of contact forces acting on any part of the robot’s body. The method is based on the Recursive Newton-Euler Algorithm, and it was implemented as part of the open-source C++ library iDyn. Furthermore, a theoretical and empirical analysis investigated how incorrect estimation of contact points may affect the resulting contact forces and induce undesired joint accelerations. Tests on the iCub robot demonstrated a significant improvement in the performance of the force controller when the tactile system was used. All things considered, this work advanced the state of the art of force control of humanoid robots, providing estimation methods and control strategies that can be applied to make robots safely work side-by-side with humans.

Publication
Italian Institute of Technology and University of Genova
Date
Links
comments powered by Disqus