Soft Robotic Manipulator Control

4th year project developing control systems for a soft robotic manipulator using MATLAB and Simulink.

The Challenge

Soft robots present unique control challenges - their non-linear dynamics and continuous deformation make traditional control approaches ineffective. The project focused on a soft pendulum system.

Approach

• System Identification: Characterizing the system's dynamic behavior from experimental data
• Model-Based RL: Using learned models to train controllers in simulation
• Neural Networks: Implementing NN-based controllers for handling non-linearities

Tools

MATLAB, Simulink, reinforcement learning toolboxes