MEMS 6-axis Sensor for Robotic Hand Calibration

As humanoid robots advance toward more humanlike capability, their designs must grow increasingly complex to replicate the intricacies of human movement and perception. Consider the human hand: its dexterity and sensitivity remain difficult to reproduce in a robotic counterpart. To approach human-level performance, a robotic hand requires more precise control and feedback.

Motors already provide actuation and limited feedback in robotic hands, but additional calibration and learning are necessary to move closer to the touch, responsiveness, and sensitivity of a human hand. In robotic grippers, force control is commonly achieved by regulating the electrical current to the motor, which is proportional to motor torque.

However, this method has limitations. Once a motor is installed into a robotic hand, factors such as mechanical friction, deformation, and other system inefficiencies introduce non-linearity and variability. Although the torque–current relationship for a standalone motor is known, it changes once integrated into the hand. For this reason, it is more effective to relearn or recalibrate the torque–current characteristics after the motor has been installed. This allows the system to compensate for real-world integration effects and improves overall performance.

A 6-axis force/torque sensor designed to measure forces within the range of a human touch is an ideal tool for this recalibration. For example, the sensor can be used to measure the force produced by a robot finger as it presses down onto the sensor surface. By repeating this process at different force levels, a characteristic curve between motor current and resulting fingertip force can be established. If the relationship is linear, two calibration points may be sufficient. If it is more complex, additional data points are required. This process can also be extended to other motions—pushing, pulling, sliding—as well as movements that involve multiple motors acting together.

Until force, torque, and friction sensors are fully integrated into robotic hands and fingers, recalibrating the motors after installation remains the most practical method for improving humanlike behavior. Once the system is properly calibrated, the robot’s range of touch and control can approach that of a sensor-integrated design. However, limitations still exist at the very low end of the torque range, where the current-to-torque relationship is less reliable. As a result, a hand that relies solely on learned motor behavior will not achieve the same dynamic range of tactile sensitivity as one equipped with embedded sensing—or, of course, the human hand itself.

Nevertheless, advancements in sensing, actuation, and calibration technologies continue to narrow this gap. One day, a robot may indeed be capable of playing guitar with the nuance and skill of Eddie Van Halen.