Underactuated System

Underactuated systems, characterized by having fewer actuators than degrees of freedom, present significant control challenges but offer advantages like reduced complexity and weight. Current research focuses on developing robust and efficient control strategies, often employing reinforcement learning algorithms (like average-reward maximum entropy RL) and model predictive control, alongside innovative model architectures such as port-Hamiltonian neural ODE networks for improved accuracy and generalization. These advancements are impacting diverse fields, from robotics (e.g., legged robots, soft grippers, and aerial vehicles) to aerospace (rocket landing) and even sustainable technologies (autonomous seaweed farming), by enabling more agile, efficient, and robust control of complex systems.