A team of researchers at Purdue University have devised a method to teach robots to navigate foreign environments by themselves. Using some very complex artificial intelligence, this robot can make educated “guesses” about what it might next encounter in its surroundings.
Rather than trying to just deal with differences in surfaces and obstacles as the robot approaches an area, the ‘bot can actually take data from its recent surroundings to predict what might be next and store that information in its internal map. By doing this, the robot can better prepare to navigate ahead of time, reducing the likelihood of getting stuck or slow progress through challenging obstacles.
Currently, the algorithm does add a bit of processing overhead to the robot’s software, meaning that it can’t cover as much ground as less intelligent robots. In future, the researchers plan to link up multiple robots to share their knowledge and to improve prediction capabilities.