This technology project was funded through an ARM Institute Technology Project Call (ARM-TEC-22-01), which focused on the following special topic areas:
- Adaptive Real-Time Path Planning and Control
- Human Action & Intention Prediction
- Robot Learning
- Feature and Pose Recognition and Estimation
- Virtual Commissioning of Advanced Robotic Systems
- Artificial Intelligence (AI) and Data for Advanced Manufacturing Robotic Systems
- Multi-Agent Motion Planning and Tasking for Discreet and Continuous Manufacturing Processes
- Virtual Masking for Advanced Coating Applications.
Project Overview
This project seeks to employ a Reinforcement Learning (RL) algorithm to allow manufacturing robotic systems to perform new tasks with minimal or no manual intervention. The team will do this by building a hierarchical RL framework in which expert knowledge, historical data, and digital twins are used to effectively learn manipulation policies to perform unstructured tasks. These outputs are expected to reduce the time and cost of implementing flexible automation.