To date, teaching a robot to perform a task has usually involved either direct coding, trial-and-error tests or handholding the machine. Soon, though, you might just have to perform that task like you would any other day. MIT scientists have developed a system, Planning with Uncertain Specifications (PUnS), that helps bots learn complicated tasks when they’d otherwise stumble, such as setting the dinner table. Instead of the usual method where the robot receives rewards for performing the right actions, PUnS has the bot hold “beliefs” over a variety of specifications and use a language (linear temporal logic) that lets it reason about what it has to do right now and in the future.
Comments are closed.