A robot has been taught to navigate using simulated neural cells.
A team at the Agency for Science, Technology and Research (ASTAR) in Singapore found that by simulating the behaviour of two types of neurons in the brain, a robot was able to navigate around a 35 square meter office.
The research, as reported by MIT Technology Review, replicated two vital elements of the animal navigation system -- place cells and grid cells. Place cells refer to neurons within the hippocampus that become active when an animal or human enters a particular place in its environment. Each place cell will have only a few fields within a small environment.
Grid cells, discovered in 2005, are similar, deriving their name from the triangular grid of points on which animals relate to their position in space.
Together, place and grid cells allow humans and animals to form a cognitive map of their environment, location and position, and are linked to regions of brain that perform spatial processing. They are also responsible for the innate sense of location that humans and animals experience.
The robot's cells were not simulated physically, however -- rather, researchers created a two-dimensional model of the cells in the robot's software. The robot -- a small machine with wheels -- was then set to roam around the research office space. To the researchers delight, the artificial place and grid cells functioned in a comparable way to those in humans and animals.
The robot's navigation system is still rudimentary, and does not yet fully represent the complexity of human or animal cognitive maps. It is also far more simple than conventional neural mapping techniques.
But it may offer advantages over these systems, which are often confused by changes to an environment. And researchers hope that the work will not only allow a more efficient way for robots to navigate but also provide neuroscientists a better understanding of place cells, grid cells and cognitive maps.
This article was originally published by WIRED UK
