Motor-Current-Based Estimation of Cartesian Contact Forces and Torques for Robotic Manipulators and Its Application to Force Control

We present a Kalman filter-based approach for estimating external forces and torques relying on a dynamic model of a serial-chain robotic manipulator where only motor signals (currents, joint angles, and joint speeds) are measurable. The method does not require any additional sensing compared to standard robot control systems. The approach exploits redundancy in 7DOF arms, but also applies to traditional 6DOF manipulators. Automatic filter calibration routines are presented minimizing the number of parameters that must be tuned in order to successfully apply the proposed scheme and to optimize estimation quality. The approach is verified by measurement data gathered from an ABB YuMi, a dual-arm collaborative robot with 7DOF each arm. Furthermore, measurement results are presented employing force and torque estimates in a compliance control scheme, verifying that the estimation quality achieved is improved compared to existing approaches and is sufficient to employ the estimates in force-controlled applications.

Note to Practitioners—More and more robotic applications involve contact with at least partially unknown environments. As a consequence, they require control approaches that go beyond the traditional position control. In particular, information about contact forces and torques has to be taken into account. However, integrating additional sensing equipment to obtain the required force/torque information is often technically challenging and expensive. Cartesian contact force and torque estimation allows obtaining force/torque information solely from available sensors. The estimation technique can be regarded as a virtual sensor, and hence this brief deals with a key technology enabling force controlled robotic applications such as assembly, grinding, and deburring without the need for expensive additional sensing.