Researchers give robot simulated neurons for travel

Humans and animals alike are able to travel around in familiar places without using a map or getting lost; put the GPS away and pay attention to your surroundings, and you'll quickly form a mental map of whatever unfamiliar place you're visiting. This ability is due to two neuron types: grid cells and place cells, as they're called. Scientists have recently used simulated versions of these cells to help a robot navigate.

A robot is unique from humans in many ways (obviously), not the least of which is its ability to have pre-defined maps and routes uploaded to its "brain". That's great if the robot is tasked with traveling the same limited, already-mapped route, but ineffective for instances where a robot must learns its own way around in unfamiliar territory.

In that case, robots could benefit from the same type of neurons humans and animals have — something a team of scientists have given them via simulations. The work was done by a team of researchers from "A*STAR", Singapore's Agency for Science, Technology and Research.

Grid cells and place cells are similar, but have their own specific purposes. Of them, grid cells were the first to be identified, with a scientist discovering their purpose back in the 70s; these cells fire when a human or animal is at a place, such as when they come across a road sign or tree they've previously passed. Place cells were only identified back in 2005, and are concerned with providing a human or animal with knowledge of their approximate location within a particular place (where your room is within a house, for example).

After being given a simulation of these neurons, a robot was placed in a 35 square meter office to roam around, and the researchers found the simulated cells were firing in a way similar to those in animals and humans. Improvements still need to be made, and that will, in part, depend on a better understanding of how the cells work in living creatures, however.

SOURCE: MIT Technology Review