Stay on Top of Enterprise Technology Trends
Get updates impacting your industry from our GigaOm Research Community
Researchers at MIT have figured out how to make multiple robots talk to each other and collaborate in real time — a task equivalent to telling a bunch of toddlers to bake a cake and actually expecting them to do it. The MIT article on the research describes the problem well:
Writing a program to control a single autonomous robot navigating an uncertain environment with an erratic communication link is hard enough; write one for multiple robots that may or may not have to work in tandem, depending on the task, is even harder.
The MIT release highlighted how solving this problem will be advantageous in automated environments for controlling drones or operating shipping warehouses, but it could have a big impact in the smart home. As I’ve argued before, most connected devices — especially those that learn — are essentially robots. And getting those robots, whether they control your temperature or access to your home, to work together to make your home secure and comfortable today is a mess.
And because right now, the home has a variety of networks from Wi-Fi to Zigbee that information has to traverse, the communications can be unreliable or just add more latency. So if a light on one type of network hopes to communicate with a motion sensor on a different network, the added latency means that triggering the motion detector turns on a light a few seconds after the person has passed it, at which point they no longer need the light.
So MIT’s system, which stitches together existing control programs, could find a home outside of the warehousing example the researchers are showing off. The key to the system is that it automatically plans to route around dropped connections or algorithms that make a robot behave in a way that doesn’t achieve the end goal. From the release:
“In [multiagent] systems, in general, in the real world, it’s very hard for them to communicate effectively,” says Christopher Amato, a postdoc in CSAIL and first author on the new paper. “If you have a camera, it’s impossible for the camera to be constantly streaming all of its information to all the other cameras. Similarly, robots are on networks that are imperfect, so it takes some amount of time to get messages to other robots, and maybe they can’t communicate in certain situations around obstacles.”
Thus the system handles uncertainty in the real world by triage, quite literally. It takes three inputs and uses them to determine how the system as a whole should behave. The systems are low-level control algorithms that govern agents’ behaviors collectively or individually, stats about those programs’ execution in a particular environment and a scheme for valuing different outcomes that takes into account success and penalizes using energy without accomplishing the task.
Within these parameters the robots can “make decisions” based on the data gathered, and then take actions to achieve the pre-determined goal. If flexible enough, such a methodology could be used in a variety of systems ranging from the smart home, controlling automated manufacturing lines but even for deploying information technology in a virtualized environment. This provides an opportunity for companies like Splunk, which is collecting machine-level data and for the myriad companies that are researching AI to really shine.