“…The most challenging aspect of interaction control lies in selecting an appropriate model for the environment, whose inevitable inaccuracies constitute the aspect the aforementioned works seek to compensate: in the literature, the most common choice is adopting a linear spring model [8], [12], [17], [18], [20], [21], [22], [23], [24], [26], [27], [28] but, as specified in [8], [29], [30], it constitutes, in general, an approximation. To cope with this inherent difficulty, recent research trends lean towards leveraging AI-based and data-driven methods to devise control laws from data, rather than relying on modelbased strategies.…”