Vicarious: AI for the robot age
Improving Baxter robot to work on the latest Gazebo simulator and ROS release.
ABOUT
Vicarious is developing artificial general intelligence for robots. By combining insights from generative probabilistic models and systems neuroscience, their architecture trains faster, adapts more readily, and generalizes more broadly than AI approaches commonly used today. Vicarious hopes to bring human-like intelligence to the world of robots by organizing around a set of questions and themes, while using datasets that are appropriately designed to probe those questions.
CHALLENGE
Vicarious needed their research to be enabled in embodied AI. They chose to use an industry standard by selecting Rethink Robotics’ Baxter robot as one of their platforms. However, they needed additional expertise to migrate Baxter to the latest version of the Robotic Operating System and the corresponding latest Gazebo Simulator.
SOLUTION
PickNik Consulting was engaged to migrate Baxter to ROS Kinetic and Gazebo 7 on Ubuntu 16.04. PickNik provided implementation advice and fixed code including upgrading various Gazebo plugins, creating custom Baxter plugins, and finalizing the integration through extensive testing.
HIGHLIGHTS
- Vicarious agreed to open source the effort and the results were announced on the Baxter mailing list
- 23 pull requests were created across 5 Github repositories
- Fixes to the upstream Gazebo project were addressed, cleaning up warnings, excess console output, and build errors
- Load time of the Baxter simulator was improved by 35 seconds!
“It was the best experience I’ve had working with an external consultant!”
Dr. Bhaskara Marthi
Head of Robotics, Vicarious