Project 66: Motion Control of Micro-Soft Robots based on Reinforcement Learning
Contact Information：
Assoc. Prof. Zhuo-Chen Ma
Email: zcma@sjtu.edu.cn
Project Description and Objectives：
As a rapidly developing field that intersects with manufacturing, intelligent control and robotics, micro-soft robots are showing good application prospects in many areas of concern and challenge. However, due to the complicated dynamic characteristics of the soft structure, it is hard to establish accurate dynamic models for motion control. As a result, model-based control methods cannot work effectively in motion control tasks of micro-soft robots. As an alternative, reinforcement learning can be leveraged as an adaptive model to fit complex and nonlinear dynamic characteristics of soft deforming action. The objective of this project is to establish the input-output model of micro-soft robots’ motion with reinforcement learning. After establishing the dynamic model, a model-based control algorithm will be integrated to achieve accurate and robust control.
Eligibility Requirements
Familiarity with basic programming language (mainly Python and C++);
Basic knowledge of machine learning and reinforcement learning;
Basic knowledge of control theory.
Main Tasks：
Review relevant literature to acquire the basic knowledge of micro-soft robots
Construct a network and algorithm structure to deploy reinforcement learning in the control of micro-soft robots
Website：
Lab: https://robotics.sjtu.edu.cn/
School: http://english.seiee.sjtu.edu.cn/