Deployment and Extension of UltraBot with Tactile Feedback
Description
This project aims to deploy a state-of-the-art learning based autonomous ultrasound system (UltraBot, see this publication) on an existing robotic platform and extend its capabilities with tactile sensing for safer and more robust human-robot-interaction during ultrasound examination. Ultrabot leverages large-scale imitation learning and expert demonstration data to achieve human-level performance in carotid ultrasound scanning.
The student will focus on integrating the perception and control pipeline into a real system, adapting it to hardware constraints and implementing tactile sensing to improve contact safety, compliance, and scan reliability under anatomical variability. The project involves robot control, multimodal sensor integration and learning-based policy development.
Tasks involved are:
- Integration of UltraBot model into a real robotic system;
- Calibration and synchronization of ultrasound, vision, and tactile sensors;
- Extend the pipeline with tactile-based strategies;
- Validation and test.
Requirements
- Background in robotics and machine learning;
- Experience with ROS 2, C++ and Python;
- Experience with Docker is beneficial.