Enhancing Healthcare Assistance with a Self-Learning Robotics System: A Deep Imitation Learning-Based Solution
Posted by Coco Li
Author：Yagna Jadeja , Mahmoud Shafik , Paul Wood and Aaisha Makkar
Research Area:
Healthcare Robotics · Imitation Learning
Summary:
A self-learning healthcare robotics system using Elephant Robotics myCobot 280 Jetson Nano to learn assistive tasks from human demonstrations and perform object picking and handover using deep imitation learning and vision-based perception.
Healthcare robots need to be more adaptable because traditional robots rely on fixed programming and struggle with changing patient needs and environments.
Key Challenges:
● Limited adaptability and dependence on predefined tasks.
● Difficulty recognizing human actions and environmental changes.
● Safety and reliability concerns in human-robot interaction.
Solution:
The study developed a Self-Learning Robotic System (SLRS) using the Elephant Robotics myCobot 280 Jetson Nano as the 6-DOF robotic arm. It combines Deep Imitation Learning, YOLOv8, MediaPipe, and LSTM to learn tasks from human demonstrations and perform assistive tasks such as object picking and handover. The system achieved an 87.3% task completion rate and over 85% grasp success rate.
