04PUBLISHED · ARXIV

ADHD ROBOTICS

ROBOTIC ASSISTANT FOR ADHD THERAPY — CO-AUTHORED RESEARCH

ROLE

CO-AUTHOR & ENGINEER

TEAM

TUM RESEARCH GROUP

DURATION

FEB 2024 — APR 2024

TOOLS

PYTHON, RASPBERRY PI, SOLIDWORKS, OPENCV

OVERVIEW

Co-authored research on a robotic assistant designed for ADHD therapy sessions. The system combines physical robotics with LLM-powered conversation to guide therapy exercises, track attention patterns, and provide real-time adaptive responses.

SYSTEM DESIGN

The robot runs on a Raspberry Pi with a custom SolidWorks-designed chassis. Computer vision via OpenCV tracks the patient's attention and engagement. GPT-4 and Claude-3 power the conversational therapy module, adapting responses based on real-time behavioral signals.

COMBINING PHYSICAL ROBOTICS WITH LLM REASONING FOR THERAPEUTIC INTERVENTION.

TECHNICAL CONTRIBUTIONS

HARDWARE — Designed and assembled the robot chassis with SolidWorks, integrated Raspberry Pi compute

COMPUTER VISION — Built OpenCV pipeline for gaze tracking, attention scoring, and engagement detection

LLM INTEGRATION — Engineered prompts and conversation flows for GPT-4 and Claude-3 therapy modules

SPEECH PIPELINE — Real-time speech-to-speech using Azure OpenAI for natural conversational interaction

2

LLM MODELS INTEGRATED

1

ARXIV PUBLICATION

3

MONTHS BUILD TIME

4

TEAM MEMBERS

← PREVIOUS

PAIGE