Rosie - Autonomous Home Assistance Robot
Project Overview
After completing my undergraduate studies, I developed Rosie, an AI-powered robot designed to assist elderly and disabled individuals with daily tasks and provide companionship. Motivated by the challenges faced by seniors living alone, my goal was to create a robot capable of autonomous navigation, object retrieval, and social engagement.
Technologies Used
- Programming Languages: Python, C++
- Frameworks & Libraries: ROS2, PyTorch, OpenCV, Micro-ROS
- Hardware: LiDAR, IMU, Depth Cameras, NVIDIA Jetson Nano, Raspberry Pi, ESP32
- Algorithms: SLAM, Reinforcement Learning (RL), Sensor Fusion, Extended Kalman Filter (EKF)
Key Achievements
- Integrated SLAM with reinforcement learning for navigation and adaptive manipulation.
- Applied sensor fusion techniques with EKF for improved navigation accuracy using data from LiDAR, IMU, and depth cameras.
- Synchronized real-time data from multiple sensors using a multi-threaded ROS2 architecture.
- Trained RL models for object retrieval, involving hyperparameter tuning and curriculum learning.
Results
- Developed a fully functional home assistance robot capable of interacting with household objects autonomously.
- Significantly improved navigation accuracy and grasping performance through sensor fusion and reinforcement learning.
- Enhanced object retrieval capabilities, making Rosie a reliable assistant for elderly individuals.
Future Work
- Expand Rosie’s capabilities to include advanced voice recognition and natural language processing for better interaction.
- Optimize power consumption for extended operation time in home environments.
- Integrate facial recognition for personalized interactions and emotional engagement.