SEXPAND
One of the more serious limitations facing the robotics industry today is that each bot it produces is an island unto itself. Worse, robots' primitive AI doesn’t allow for intuitive thinking or problem solving — what’s known as artificial general intelligence. Looking to overcome this problem, researchers from several different European universities have developed a cloud-computing platform for robots that will allow them to collaborate — and make each other smarter — over the Internet.
Essentially, the new system, called Rapyuta: The RoboEarth Cloud Engine, is an open source repository of accumulated information for robots. Its name is taken from the movie Castle in the Sky by Hayao Miyazaki, in which Rapyuta is the castle inhabited by robots. The name is quite perfect, actually.
In terms of the technology required, the developers implemented a Platform-as-a-Service (PaaS) framework designed specifically for robotics applications. Each robot can kickstart its own computational environment or launch any node that has already been set up in advance (typically by another developer). It can also communicate with other nodes using the WebSockets protocol.
SEXPAND
Data stored in the cloud will include software components, maps for navigation (including the location of objects and world models), task knowledge (like action scripts and manipulation strategies), processing human voice commands, and object recognition models.
By allowing robots to collaborate and share information in this way, each bot will essentially offload its “brain” into the cloud. Moreover, each unit can be considerably “lighter” in terms of its processing and software requirements; when in doubt, it just needs to hit the cloud. Ultimately, this will make robots cheaper, more efficient — and more intelligent.
The platform will allow robots who are connected to the Internet to directly access powerful computational, storage, and communications technologies, including those of modern data centers.
"The RoboEarth Cloud Engine is particularly useful for mobile robots, such as drones or autonomous cars, which require lots of computation for navigation,” noted Mohanarajah Gajamohan through an official statement, and a researcher at the Swiss Federal Institute of Technology (ETH Zurich) and Technical Lead of the project. “It also offers significant benefits for robot co-workers, such as factory robots working alongside humans, which require large knowledge databases, and for the deployment of robot teams."
As exciting as this appears, the concept is not without its problems. Two things conern me in particular.
First, anything that’s connected to the Internet is inherently hackable. This system will need to be crazy secure, otherwise the robots could be controlled by a malicious source (either individually, or collectively).
And second, the query response-and-match algorithms will need to be very strict to prevent a robot from getting the wrong instructions. For example, a robot could ask the cloud for instructions on how to perform task x, but the cloud-engine could misunderstand and provide it with instructions for task y. The robot, because it’s stupid, will then execute task y. This could be dangerous, and even potentially catastrophic in some contexts.
To learn more about RoboEarth, check out their official website. Their study can be read here.
Image: agsandrew/Shutterstock, RoboEarth.