This story is over 5 years old.

Before They Take Our Jobs, Robots Have To Become Better Coworkers

MIT is working on getting robots to talk to each other and notice things around them more smoothly.
Image: WowWee

At (smart) home and at work, one problem that holds our new robot overlords back from true world domination is their inability to contextualize space and their position within it with the same cognitive dexterity that comes so naturally to us humans. This creates many concerns about safety and efficiency when it comes to working—or living—alongside potentially clumsy or lethal machines, or even just leaving a room full of robots to their own devices.

The solution to this dilemma, until now, has been to robot-ify the humans in the equation by, say, attaching sensors to them so the robots can better interpret the location and movement of their meatbag coworkers. But new research from MIT approaches the problem from the opposite direction: developing new programs to help groups of robots communicate with one another more clearly.


"Writing a program to control a single autonomous robot navigating an uncertain environment with an erratic communication link is hard enough," the press release describing the research said. Writing one "for multiple robots that may or may not have to work in tandem, depending on the task, is even harder."

The problem, Christopher Amato, a postdoc student at MIT's Computer Science and Artificial Intelligence Laboratory (CSAIL) and lead author on the new paper, said in the release is that it's already "very hard" for robots to "communicate effectively" in multiagent systems.

 “If you have a camera, it’s impossible for the camera to be constantly streaming all of its information to all the other cameras," Amato explained. "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.”

What the new program does, therefore, is account for uncertainty—in the actions of individual robots, in the network of communications between multiple robots, and with environmental contingencies. To create the new system, the researchers took three inputs: the “macro-actions” of the robots themselves, statistical information about how these actions will be performed in the environment in question, and a valuation system to weigh outcomes—tasks accomplished versus energy expended, for instance. A video accompanying the research shows some roombas operating with the programming to collaborate on pushing various boxes around a simulation warehouse.


The idea, then, is a sort of big data approach: one could let the robots run free in a particular environment to gather information about potential snags in the system, then use that to optimize any future deployment.

As the warehouse simulation suggests, the MIT researchers are thinking about enterprise applications here for businesses like Amazon, which are constantly trying to refine the management of machines and humans to process inventory as fast as post-humanly possible. But there any number of applications for personal use as well. As Motherboard's own Brian Merchant noticed during his tour through a smart home last month at the Consumer Electronics Show (CES) in Las Vegas, most of the "smart" features that are now being promoted are actually just gussied up notification systems that can tell you that, say, you left the wine cooler open by delivering a message straight to the TV you're watching.

That's clever, sure. But not exactly revolutionary. A swarm of tiny little robot butlers that can communicate amongst themselves to run in the background of your house is—provided they don't accidentally set fire to the couch when racing to close the window or reset the router.

Interest in both kinds of applications is thriving. Google alone bought eight different robotics companies in 2013. Just this week, James Dyson revealed he was investing £5 million (around $8.3 million) to create a new robotics lab in partnership with the Imperial College London to investigate new "vision systems" for robots. And as Colin Lewis noted in a recent post on Robohub, stocks in robotics companies are beating current market averages, so Wall Street is getting bullish on the entire scene as well.

All of this raises the compelling possibility that smart, interconnected robots could alleviate the myriad pressures being placed on people like warehouse workers, who may now be facing increased risk of mental illness due to the harsh demands of working alongside automatons. But then again, that still leaves the looming question of where yet another crop of displaced workers will be able to find jobs.