GE Robotics and Autonomous Systems
The project will investigate an artificial intelligence (AI), specifically reinforcement learning (RL), enabled tool and framework, to reduce the unnecessary time spent on manual programming, tweaking, and validating the robot tool path for manufacturing.
Specific project tasks are as follows:
- Conduct a literature study on state-of-the-art research, commercially available technologies and products on robotic manipulator tool path generation automation.
- Get started and follow tutorials with the following enabling technologies:
- ROS (robot operating system)
- MoveIt Motion Planning Framework for ROS
- Open Motion Planning Library
- Bullet Physics SDK (PyBullet)
- Gazebo – an open-source 3D robotic simulator
- Gymnasium - an open-source Python library for developing and comparing reinforcement learning (RL) algorithms
- Q-Learning: a simple RL algorithm
- Develop a simulation robotic environment with reinforcement learning for motion planning. The expected outcome of this task is to develop a software pipeline to explore RL for motion planning, with all or a subset of the above enabling technologies. With the pipeline, it should have the following functions: (1) robot motions can be generated on demand and used as training inputs to the reinforcement learning model; (2) the RL model/algorithm can be trained; and (3) the trained RL model can be used to take inputs to generate robot motions based on the predefined environment. Here are some example projects found online:
- Artificial Intelligence
- Background of ROS and machine learning would be plus.