Exploration of RL-based agents in the context of space robotic systems

This research will explore machine learning methods in order to devise a control scheme for robotic manipulators(Candarm3) in the context of space exploration. The objective is to develop an early prototype for an autonomous learning agent which can carry out standard control tasks without any operator supervision.
The primary machine learning methods that will be studied will revolve around deep-reinforcement learning methods, in which an agent iteratively improves its performance in a given task. This is done through simulating training exercises, where the agent is rewarded for performing well. The agent modifies its behaviour in order to maximize its expected reward in future training exercises.

Faculty Supervisor:

Chi-Guhn Lee

Student:

Partner:

MacDonald, Dettwiler and Associates Inc (Brampton, ON)

Discipline:

Computer science

Sector:

Manufacturing; Professional, scientific and technical services

University:

University of Toronto

Program: