Force estimation in electro-mechanical systems - theory and experiments.

In this paper, a purely data-driven approach for force estimation in electro-mechanical actuators is proposed. First, the contact with the surrounding environment is detected by means of an unsupervised learning procedure applied to the current measurements, then the clamping force is estimated in real-time by using a data-driven model of the interaction between the actuator and the system. The effectiveness of the proposed approach is illustrated on an experimental braking system setup, also showing to outperform the state of the art methodology.