It’s a math problem familiar to any maitre d’: You have a roomful of thirsty guests, one bartender, and a handful of servers. How soon before each guest is holding a drink?
A team from MIT claims that new algorithms and a crew of delivery robots can outdo humans at this challenge, serving up cold ones to partygoers more quickly than human servers.
In a demonstration this summer, the researchers showed how a double-armed robot could be instructed to play bartender and collaborate with a crew of two delivery bots — little wheeled platforms scooting around with party coolers — to ferry drinks to waiting people.
These are hardly the first robots to play bartender. Last year, a pair of robots made by Italian firm Makr Shakr mixed drinks and placed them on a conveyer belt leading to waiting passengers at the Sky Bar on the Quantum of the Seas cruise ship.
But the MIT demo is the first time booze bots have worked as a team, said University of New Hampshire computer science professor Christopher Amato, who began the work as a research scientist at MIT’s Computer Science and Artificial Intelligence Laboratory. The team presented its work last month at the Robotics Systems and Science Conference in Rome.
The robots also take a smarter approach to collaboration, preparing for possible missteps like dropping a bottle, or picking up the wrong one.
When programming multiple robots to collaborate, lots of other researchers take the “swarm” approach, where an army of some 1,000 robots executes a very simple task, despite the fact that each individual robot is oblivious to the movements of its teammates.
MIT’s bartender bots, on the other hand, are aware of their neighbors and adjust their behavior based on what they are doing. “Our robots are more like people, the swarm robots are more like ants,” Amato said.
The booze delivery task is just a dry run en route to harder real-world problems. Amato said that the new algorithms being used by the bartending robots could be applied to any situation that would benefit from a team of robots collaborating with each other.
For example, the researchers behind this project also are simulating search-and-rescue operations. In that scenario, flying drones could scour a disaster zone for survivors and then send a message to ground robots, which could be dispatched to pick them up. Another simulation is exploring how Amazon-style delivery drones could communicate with a ground crew to ferry a package to the correct doorstep.
“The beer problem is really emblematic of all these real-world problems,” Amato said.