T-Top (tabletop socially assistive robot) · Open-source tabletop socially assistive robot for research: a Stewart platform carries a 7-inch touchscreen face on a base that turns without limit, with an NVIDIA Jetson AGX, a 16-microphone array and RGB‑D and wide-angle cameras in the torso
Open-source tabletop socially assistive robot for research: a Stewart platform carries a 7-inch touchscreen face on a base that turns without limit, with an NV…
- Maker
- IntRoLab (Intelligent / Interactive / Integrated / Interdisciplinary Robot Lab, Université de Sherbrooke)
- Form
- Open-source tabletop socially assistive robot for research: a Stewart platform carries a 7-inch touchscreen face on a base that turns without limit, with an NVIDIA Jetson AGX, a 16-microphone array and RGB‑D and wide-angle cameras in the torso
- For
- Human-robot interaction (HRI) researchers and labs: the README says it provides richer interaction modalities and develops higher cognitive abilities from interacting with people, and the team's paper calls it an advanced HRI experimentation platform
- Sensors
- Intel RealSense D435i RGB‑D camera and Arducam AR0230 wide-angle camera; 16 xSoundsMicrophones on a 16SoundsUSB sound card; 7-inch 1024×600 capacitive touchscreen; INA220 current/voltage sensing, 4 ALS‑PT19 light sensors and 4 buttons; output through 4 Dayton Audio DMA45‑8 speakers (2 MAX9744 amplifiers) and an LED strip (README)
- Connectivity and app
- Intel 8265 dual-band Wi‑Fi and 100 Mbps Ethernet; no phone app: developers reach ROS 2 Humble on the robot over SSH and VS Code and test each function in a control panel GUI. Cloud services are optional: speech to text uses Google Cloud or local Faster Whisper, speech synthesis defaults to local Piper (Google optional), and chat uses the OpenAI API or a local Ollama model; the telepresence demo uses OpenTera, stand-alone or through the Université de Sherbrooke's OpenTera server
- Subscription
- Self-built open source with no subscription; the user pays for OpenAI or Google Cloud services if chosen (the chat doc calls the OpenAI API a paid service)
- Openness
- The README's Licensing section: source code files are GPLv3 (LICENSE_SOURCE_CODE) and all other files are CC BY‑NC‑SA 4.0 (LICENSE_OTHER; GitHub reports NOASSERTION). On that split, the non-code files (SolidWorks CAD, STL/DXF, KiCad PCBs and Gerbers, docs) may be copied, shared and adapted for non-commercial purposes only, and adaptations must carry BY‑NC‑SA too. The repo holds the Teensy firmware, ROS 2 packages, three custom PCBs and 13 assembly steps; deep-learning weights ship as GitHub releases (DNN_Weights_v4.1.0, 2024‑05‑13)
- Privacy
- Two cameras and a 16-microphone array are on board. Perception (face and person identification, speaker identification, sound classification, Whisper speech recognition) can run locally on the Jetson; choosing Google Cloud speech to text sends audio to Google, and choosing OpenAI chat sends text to OpenAI (the doc says OPENAI_API_KEY is already set on T-Top's Jetson), with local Ollama as the alternative. Telepresence carries audio and video over OpenTera WebRTC.
- Not stated
- Overall size and weight, battery life, BOM total, build time, recommended age
Vendor claims
- An open-source tabletop socially assistive robot: telepresence, emotions, talking, greeting, face following, dancing and sound following
- A 16-microphone array plus RGB‑D and wide-angle cameras, with on-board deep-learning perception: SLAM, object detection, pose estimation, face recognition, speaker identification, speech to text and sound source localisation
- The team has used it in an exploratory evaluation with older adults (HRI 2024 paper)
- Source code under GPLv3; CAD, PCBs and docs under CC BY‑NC‑SA 4.0 (no commercial use)
Parts
- Teensy 4.0 · Controls Stewart platform motors and feedback
- 8265 NGW · Provides dual-band wireless network connectivity
- DMA45-8 · Plays robot speech and sound effects
- MAX9744 · Drives speakers and outputs audio
- AR0230 · Captures wide-angle visual images
- D435i · Captures color images and depth data
- ALS-PT19 · Measures ambient light intensity
- INA220 · Measures battery voltage and current
- XL430-W250 · Drives the Stewart platform and rotating base
- RRC-PMM240 · Charges the robot battery
- RRC2054-2 · Powers the robot
- 16SoundsUSB · Connects sixteen microphones to the onboard computer
- NF-A4x20 5V · Cools the onboard processors
Sources
- github.com/introlab/t-top
- github.com/introlab/t-top/blob/ros2/LICENSE_OTHER
- github.com/introlab/t-top/blob/ros2/LICENSE_SOURCE_CODE
- github.com/introlab/t-top/blob/ros2/documentation/assembly/README.md
- github.com/introlab/t-top/blob/ros2/documentation/assembly/00_ORDER_PRINT_LASER_CUTTING.md
- github.com/introlab/t-top/blob/ros2/documentation/getting_started/README.md
- github.com/introlab/t-top/blob/ros2/ros/behaviors/chat/README.md
- github.com/introlab/t-top/blob/ros2/ros/behaviors/talk/README.md
- github.com/introlab/t-top/blob/ros2/ros/perceptions/speech_to_text/README.md
- github.com/introlab/t-top/blob/ros2/ros/demos/t_top_opentera/README.md
- github.com/introlab/t-top/releases
- youtube.com/watch?v=2jpqTp6jozc