Robots are leaving their rigid roots behind, learning to roll with the punches, and adapt on the fly, thank to MIT engineers. In what could be a game-changer for domestic automation, researchers have created a system where robots can gain a pinch of common sense to tackle household chores without constant human oversight, according to a report by MIT News.
Previously, household robots had to mimic human actions to the letter, if they hit a snag mid-task, they'd have to start from square one. Now, utilizing large language models, or LLMs, these metal maids are taught to pick up where they left off, adjusting their actions after a blunder. This means, they won't throw in the towel if a nudge sends them off course or if they drop a marble while scooping. Instead, they'll make the necessary adjustments and proceed.
"Imitation learning is a mainstream approach enabling household robots. But if a robot is blindly mimicking a human’s motion trajectories, tiny errors can accumulate and eventually derail the rest of the execution," Yanwei Wang, a graduate student in MIT’s Department of Electrical Engineering and Computer Science (EECS), explained to MIT News. Wang and his team's study is slated for presentation at the International Conference on Learning Representations (ICLR) come May.
The approach they've developed sees robots capable of breaking down tasks into subtasks using LLMs suggestions. The robots are then trained to switch to different subtasks on their own, if necessary,. As they handle household tasks, like scooping marbles from one bowl to another, they're subtly nudged off track, only to self-correct and continue with the task at hand, showing a significant leap from their previously single-minded ways.
"With our method, when the robot is making mistakes, we don’t need to ask humans to program or give extra demonstrations of how to recover from failures," Wang told MIT News. "That’s super exciting because there’s a huge effort now toward training household robots with data collected on teleoperation systems. Our algorithm can now convert that training data into robust robot behavior that can do complex tasks, despite external perturbations." With such advancements, the dream of having a robot butler adept at dealing with the unexpected is inching closer to reality.
