The original, sensorless fingertips of the i-limb were removed and the new fingertips with LMSs (Figure 3 h) were mounted onto the i-limb using the same connection points (Figure 4 ). The prosthetic hand with the LMSs were next mounted onto a six DOF robotic arm (UR-10 (Universal Robots, Odense, Denmark)) to slide the hand along different surfaces and textures at different speeds (Figure 1 ). The hand was attached to the arm via a 3D printed mechanical adaptor with a coaxial electrical connector and a prosthetic hand lamination collar (Ossur, Reykjavik, Iceland) for stability and electrical connectivity. The robotic arm was programmed using the teach pendant to repeatedly perform the sliding contact against different surfaces and textures.
LM has high conductivity so the resistance of each LMS was approximately 1 Ω. A five-channel printed circuit board was designed to amplify the LMS signals using Wheatstone bridges (Figure S2 ). The amplification board was powered by the Teensy 3.6 microcontroller (PJRC, Portland, OR, USA) with 3.3 V. The LMSs were connected to the amplifier board and then sampled through the Teensy as a ROS node. The ROS master was initiated in Simulink (The Mathworks, Natick, MA, USA) and the Teensy 3.6 was a slave node that sampled and published the LMS signals to the master in Simulink. Data were recorded in Simulink with a 1 kHz sample rate (Figure 4 ).
LM has high conductivity so the resistance of each LMS was approximately 1 Ω. A five-channel printed circuit board was designed to amplify the LMS signals using Wheatstone bridges (