Skip to Content

Making Robots Talk to Each Other

Robots that can work together and communicate are not only more efficient, they’re cheaper—since not every robot has to excel at every task.
August 4, 2015

When you give two robots the ability to communicate in real-time, the possibilities for their teamwork are vast. Researchers at Carnegie Mellon University have done just that: enabled two types of robots with very different capabilities to collaborate in order to fulfill people’s requests.

These two robots can communicate in real time as they work together. Baxter (right) hands over objects for CoBot to deliver to humans.

Baxter is a stationary robot, equipped with two arms that can delicately manipulate objects, while CoBot has no arms but is adept at navigating indoor spaces and can reliably deliver objects using its front-end basket. The researchers wanted each robot’s strengths to make up for the other’s shortcomings, so they could work together to relieve humans of menial duties such as fetching and delivering objects throughout a building.

The robots talk wirelessly, using a common domain language to convey events as they occur, and, crucially, they also provide feedback to each other, which allows them to work together even when things don’t go exactly as planned. When the two robots coӧrdinate, they have three options for deciding what to do next: one robot can tell the other robot to wait until a certain moment to act; one can instruct the other to repetitively carry out the same activity until a particular moment; or one robot can simply ask the other what to do.

“There’s actually a sort of conversation going on here,” says Manuela Veloso, a professor of computer science who is involved in the research. “The robots can adjust to each other and optimize their work.”

Previously successful robot teams have either involved robots of the same type, or have involved heterogeneous robots acting in fixed scenarios. The trick to adaptive, diverse robot teams is to have the bots interact sparingly, says Veloso. The individual robots work independently until they absolutely must interact to complete a task, leaving fewer opportunities for mistakes and providing more flexibility.

These teams work best when one specialized robot like Baxter is the nexus of a group of simpler robots – Baxter carries out the fine-tuned parts of a task, and then enlists the help of any available nearby CoBot to take on the grunt work of delivery.

“Imagine Baxter is cooking breakfast for someone,” says Steven Klee, a graduate student working on the project. “The process of cracking an egg and cooking it shouldn’t require [Baxter] to interact with CoBot. However, the robots need to coӧrdinate when CoBot delivers the egg.”

Reducing the amount of interaction in the system also means that if one team member malfunctions or breaks, a ripple of problems won’t ensue. And the broken bot can easily be replaced without its collaborators noticing. Although CMU’s project only focuses on work done between Baxter and CoBots, Klee says the same team structure could work with any number of robot combinations.

This kind of collaboration might not work for tasks that aren’t easily divvied up or that require more constant communication, but teamwork between robots like Baxter and CoBot is overall very efficient, says Julie Shah, who heads MIT’s Interactive Robotics Group. Shah was not involved in the CMU project, but says, “Designing every robot in a team to be good at everything is like the saying, ‘Jack of all trades, master of none.’ It ends up being very expensive, and redundant.”

Keep Reading

Most Popular

Large language models can do jaw-dropping things. But nobody knows exactly why.

And that's a problem. Figuring it out is one of the biggest scientific puzzles of our time and a crucial step towards controlling more powerful future models.

The problem with plug-in hybrids? Their drivers.

Plug-in hybrids are often sold as a transition to EVs, but new data from Europe shows we’re still underestimating the emissions they produce.

Google DeepMind’s new generative model makes Super Mario–like games from scratch

Genie learns how to control games by watching hours and hours of video. It could help train next-gen robots too.

How scientists traced a mysterious covid case back to six toilets

When wastewater surveillance turns into a hunt for a single infected individual, the ethics get tricky.

Stay connected

Illustration by Rose Wong

Get the latest updates from
MIT Technology Review

Discover special offers, top stories, upcoming events, and more.

Thank you for submitting your email!

Explore more newsletters

It looks like something went wrong.

We’re having trouble saving your preferences. Try refreshing this page and updating them one more time. If you continue to get this message, reach out to us at with a list of newsletters you’d like to receive.