Embodiment-First Learning - Interactive Dashboard
Dec 15, 2025
·
1 min read
Embodiment-First Learning: Tactile Sensing Research
This project presents an interactive research dashboard for exploring robust tactile sensing in adversarial environments. It addresses the challenge of how robots maintain accurate tactile perception when operating with protective gloves, in wet/oily conditions, or with sensor wear.
Key Features
- Interactive Signal Analysis: Visualize tactile sensor responses under different environmental conditions (clean, gloved, oily)
- Parameter Space Exploration: Understand how physical parameters (stiffness, friction) change with embodiment state
- Embodiment-First Learning Loop: Step through the proposed methodology for autonomous sensor state recognition and adaptation
- AI-Powered Analysis: Chat with an AI research assistant powered by Gemini to discuss the research concepts
- Real-time Scenario Analysis: Analyze how standard control approaches fail in different conditions
Research Context
The robot must autonomously learn its own “body state” (clean, gloved, damaged) before attempting manipulation tasks. The dashboard demonstrates:
- State Recognition - Comparing actual sensor responses against clean reference
- Physical Exploration - Micro-movements to estimate physical parameters
- Adaptation - Updating control gains based on learned embodiment state
Technologies Used
- Interactive data visualization (Chart.js)
- Simulated tactile sensor signals
- Gemini API for research discussion
- Responsive HTML/CSS interface
