A parsimonious friction model for efficient identification and compensation of hysteresis with non-local memory

A novel dynamic friction model, which allows to capture friction hysteresis with non-local memory, is presented in this paper. The model is conceived in order to find a trade-off between accuracy of the model prediction and difficulty of implementation in motion control systems with model-based friction compensation. The hysteresis function introduced into the model accounts for non-local memory, i.e., the property for which the friction output depends not only on the initial conditions but also on past extremum values of the input or the output. In comparison with other models incorporating a hysteresis function with non-local memory, the proposed model is demonstrated to reduce the number of parameters necessary to reproduce the hysteresis loops observed experimentally. Moreover, parameter identification can benefit from the availability of a closed form of the model solution.