• adaptive control
• iterative learning control
• model learning
• articulated soft robots

Thanks to their compliance structure, soft robots are effective in tasks involving cooperation with humans and interactions with the environment. However, controlling such robots is challenging due to their highly nonlinear and hard-to-model dynamics. Additionally, the use of feedback action destroys their compliant behaviour and thus negates the benefits. Iterative Learning Control approaches are an effective tool to obtain feed-forward control action to effectively track a desired trajectory without requiring a complete and accurate model of the system. However, the required learning process is strictly related to the trajectory to be executed. In this work, I propose an iterative method to learn simultaneously both the dynamic model of the robot and the control action in order to obtain an improvement in the error correction and to achieve better results already from the first execution. Finally, I validate the method through simulations and experiments on a two degree of freedom robot.