REACH Project

Complete project overview, solution, technologies, and timeline.

The Problem

The overall problem that REACH is in charge of is that robotic arms aren't designed well enough to handle the complexity of a regular human limb

Some context to this problem is that robotic arms are too rigid: they use hardcoded movements that break down in dynamic environments and that there isnt an anything out there that addresses this issue in its real-world use. - Impact on the sponsor's business

Project Overview

The solution that REACH has come up with for this is problem is the creation of a simluation-based reinforced-learning framework, which is expected to allow a robotic arm to learn and later assist with daily task, while being smooth, responsive, and have energy-efficient assistance.

Original Project Proposal: View the initial concept provided by our sponsor

High-Level Requirements

for this project, a few requirements are needed:

  • A physical data simulator
  • An RL framework
  • A robotic arm for testing the framework

Technical Solution

Our techinal solution is the creation of a sim-base, reinforced-learning framework.

This framework allows for the robotic arm to learn from its actions and self-improve while giving assistance to those that have upper limb disabilities. This will also use YOLO, a camere, to detect and operate with visual data within its environment.

Technologies & Tools

Python

Why we chose it: The reason why this language was chosen was it was the recommended language we'd use by the clients in order to accomplished the framework for this project.

Role in project: Python's role in this project is to be used in making a framework with reinforced-learning in order to later help operate a robotic arm which is attached at the waist.

Development Tools
  • MuJoCo: We use MuJoCo so that we can get physical data based on the desired outcome of the robotic arm for the framework.
  • GitHub: GitHub is our repository management approach for this project.