Back to market
- Arm Count
- 4.0
- LeRobot Support
- Yes
- Payload Capacity
- 1.5 kg
- Teleoperation method (reported)
- Leader-follower teleoperation
- Mobile Base
- No
Overview
The Stationary AI is Trossen Robotics' benchtop bimanual manipulation platform for machine-learning research and imitation learning. It pairs two WidowX AI leader arms with two WidowX AI follower arms so operators can teleoperate demonstrations and record synchronized data. Four Intel RealSense D405 depth cameras provide multi-view RGB-D observation, and the stack integrates with LeRobot, Hugging Face, ROS2, NVIDIA Isaac, MuJoCo, Gazebo, and OpenPI (pi0 / pi0.5). An optional Linux ML workstation bundle (NVIDIA RTX 5090, Intel Core Ultra 9) can be added for onboard compute.
Highlights
- Bill of materials: 4x WidowX AI arms (configured as 2x leader-follower pairs)
- Bill of materials: 4x Intel RealSense D405 RGB-D cameras
- Bill of materials: 1x 10-inch touchscreen
- Leader-follower teleoperation for demonstration/data collection
- Software stack: LeRobot, Hugging Face, ROS2, NVIDIA Isaac, MuJoCo, Gazebo, OpenPI/pi0
- Current successor to ALOHA Stationary
- Optional Linux ML workstation bundle: NVIDIA RTX 5090, Intel Core Ultra 9
Specifications
Configuration
- Arm Count
- 4.0
- Teleoperation method (reported)
- Leader-follower teleoperation
- Mobile Base
- No
- Leader Arms Included
- Yes
Software & Ecosystem
- LeRobot Support
- Yes
- Onboard Compute
- Optional Linux ML workstation bundle: NVIDIA RTX 5090, Intel Core Ultra 9
- ROS/ROS 2 Integration
- Yes
- Open-Source Assets Included
- LeRobot, Hugging Face, ROS2, NVIDIA Isaac, MuJoCo, Gazebo, OpenPI/pi0
Per-Arm Specs
- Payload Capacity
- 1.5 kg
- Arm Degrees of Freedom (per arm)
- 6.0
- Maximum Reach
- 700.0 mm
- Repeatability
- 1.0 mm
Perception
- RGB-D Camera Count
- 4.0