UTS Unleashed! are the world champions in the @Home Social Standard Platform League at RoboCup 2019.
We programmed our Pepper robots to solve challenging problems in the home: working as a housekeeper, tour guide, party host and waiter.
This is the continuation of several years of success. We placed second in 2018. In 2017 (our first year), we placed second and also won the human-robot interface award.
This was a team effort that brought together many UTS students with staff.
Some of our developments include:
- A powerful speech recognition and natural language processing system
- Real time visual localization, mapping and navigation for low-end CPUs
- A custom build of Gentoo Linux to ease deployment
- A reliable speech production system
- An attractive tablet-based interface for user-friendly visual feedback
- Protocols and infrastructure for streaming audio and video over low bandwidth connections
- Pose and gesture detection
- Face detection and face recognition
- Fast object recognition
- Chair, door and obstacle detection
- Person attribute description (e.g., gender, outfits, glasses, hair)
- Grasping on robots with limited agility
This project incorporated work at all ends of the software development stack:
- Optimizing code to perform faster convolutions on Intel Atom processors
- Developing C++ interfaces to audio and video drivers and high-efficiency compression libraries
- Creation of a custom Linux distribution
- Design, training and deployment of deep convolutional neural networks for image and text processing
- Adaptation and optimization of pre-trained deep neural networks
- Development of novel stream-based architectures
- Distributed systems development with ROS
- Higher level application development in Python
- Web, websocket and canvas development
- User experience design