For the past few years, a consortium of six European research institutes has been collaborating on a project known as RoboEarth. Essentially a “worldwide web for robots,” the idea is that it will allow robots to access a shared online database of each others’ software, thus allowing them to learn how to perform new tasks from one another. The first phase of the project, Rapyuta: The RoboEarth Cloud Engine, is now up and running.
Named after a fictional castle in the sky inhabited by robots, Rapyuta’s main purpose is to allow robots’ data-processing functions to be performed in the cloud. This means that the robots themselves won’t require as much onboard computational hardware, and will thus be lighter, less expensive and more robust.
Each robot using the service has its own secure cloud-based computing environment. Using a wireless connection, the robot uploads data to that environment, where it’s processed at a rate that’s reportedly much faster than would be possible using robot-based hardware. The results are then downloaded back to the robot, which acts upon them.
Additionally, the environments of different robots can be linked together, allowing them to work as a team. Plans call for each environment to also be linked to the RoboEarth knowledge repository, allowing robots to draw upon the system’s “collected wisdom” when necessary.
Rapyuta would be most useful for particularly complex tasks, such as mapping, navigation, and human voice command recognition. It’s been suggested that it would be particularly useful for mobile robots, that require a lot of computational power for navigation.
A basic overview of the system is provided in the video below.
Sources: RoboEarth, ETH Zurich
