IBM?s new purchase, SoftLayer, powered the DARPA Virtual Robotics Competition, and could do the same for the future of cloud robotics.
The DARPA Robotics Challenge (DRC) is currently a third of the way to completion, having selected finalists in both Track A and Tracks B/C to move on to phase two of the competition. All thirteen teams have out-programmed over 100 competitors and still have a year of unforeseeable challenges ahead, but it seems that in the DRC, small victories are just as valuable as the grand prize.
Just ask SoftLayer?big data mogul IBM?s newest acquisition?that won the competition?s bid to provide the cloud infrastructure services for the virtual robotics challenge (VRC).
When the original provider, Amazon Web Services (AWS), IBM?s top rival at the moment, was unable to support the rapid loop speeds required for data communication, the Open Source Robotics Foundation which organized the contest tapped SoftLayer. Due to the sheer bulk of programming needs?100 teams, spread across the globe, controlling their own virtual robots processing physical and sensory data in a simulated, 3D environment, in real time?required SoftLayer?s unique customization of bare-metal and cloud servers.
?The first VRC is an exclamation point in the evolution of the cloud, testing its performance limits and highlighting the need for bare metal servers and virtual environments to work in tandem,? said Nathan Day, SoftLayer?s chief scientist.
With communication loops between machines operating smoothly at 1kilobyte per second, it appears SoftLayer has put out one of the fires facing cloud robotics?and interestingly enough, in support of a contest for virtual robotic emergency responders.
For all of the industry excitement and potential surrounding cloud robotics, the very real concern of data relay speeds persists in raining on the otherwise enthusiastic parade. But this could very well be the platform that changes everything.
That is, if SoftLayer and IBM want to. They?re currently embroiled in a back-and-forth legal battle with AWS over a ?secret? CIA cloud bid, plus a race to win over as many government departments as possible. Refocusing their gaze, though, could make them instrumental in the future of cloud robotics.
Beyond business models and stiff competition in big data, SoftLayer has the most powerful cloud system currently available. The VRC proved that. And moreover, the VRC is the perfect example of where the cloud has a viable market in robotics?research and development.
Testing prototypes can be one of the most expensive stages of production, when unforeseen problems arise and when design flaws make themselves known in the most unfortunate of ways. Any time a physical prototype is made, money is on the line. But what if there?s a way to experiment and find the pitfalls before you throw money into them?
A highly controllable virtual reality, which is essentially what the VRC required, could provide this opportunity to iron out the kinks in programming and test multiple environmental factors at once, without risk to the prototype or the project budget. Granted, this is essentially a super-powered simulation, but there?s an added element that the cloud brings to it?a community.
Think of it as peer reviewing in real time. But the hypothetical beauty of it is the range; instead of a single group of developers or roboticists in a lab in California, for example, the cloud can extend that range globally, inviting multiple brains with different perspectives to contribute. And with SoftLayer?s servers, it could all happen in real time.
This could mean big things, especially for the future of autonomous robotics. Right now, the difference between programming-dependent and learning-independent robots is a matter of building a high-functioning cloud to act as a program library. Learning from demonstration (LfD) is certainly improving the viability of widespread autonomy, but very slowly.
The fix? There?s no singular option. But a cloud with unprecedented data relay capabilities would certainly be ideal. Instead of relying on its creators in-lab, a cloud-connected robot could get a thousand references from the global community, providing demonstrations in real time, and creating a catalog of possible action paths.
Robots, by nature of their programming, are good at precision and following directions. Humans, by nature, are bad at exactitude (especially repetition) and often give complicated directions. The way to close this gap, and make human-robot interaction more effective, is to provide robots with the resources to understand their world, including us.
SoftLayer, pretty uniquely, supplies the raw computing power to make this sort of cloud community possible. The question remains, though, if this is what IBM is looking to cultivate as a market. Right now, the VRC success seems more a portfolio padder than IBM?s break into robotics.
Currently, its cloud availability is rentable on a monthly basis, and is fully customizable, so researchers and roboticists can pursue the tech on an individual basis. But it remains unclear if SoftLayer and IBM will offer to bring the robotics industry under their umbrella.