Compliant physical human-robot interactions by combining nonlinear model-based control with learning-based principles

The project researches the development of a new control framework for soft robots that can guarantee safe operation with humans in the loop. This is achieved through the utilization of a modular
control architecture and the subsequent design, testing and validation of three modules:
• A nonlinear model-based adaptive control that ensures stability of the closed-loop.
• A machine-learning-based approach (e.g., neural network) that ensures learning and adaptability
for the soft-robot during interactions with the environment.
• A biomimetic control loop that encapsulates the model-based and learning-based control schemes
to ensure high-accuracy end-effector control and high compliance for the soft robot.