Rat cunning will improve GPS for in-car navigation and robots

HE navigational patterns of small mammals such as rats have inspired the development of technology to boost the reliability of global positioning systems.
Queensland University of Technology science and engineering faculty researcher Michael Milford is using camera technology and algorithms to improve the performance of GPS for in-car navigation and robots.
"For car GPS to work, you need three or four satellites to be able to send a signal to your smartphone or your navigation device," he said. "But if you are in a built-up city with tall skyscrapers and tunnels, the signal can disappear or get badly interfered with."
Dr Milford's approach to visual navigation algorithms, dubbed SeqSLAM (Sequence Simultaneous Localisation and Mapping), uses local best-match and sequence recognition components to lock in locations.
Top 50 Tech Rec Coverage
"For the first time, we have image data from most streets in Western countries because companies like Google have driven around and captured all of this data."
The "novelty" of the approach was that it used cheap, low-quality cameras and simple algorithms, he said.
"One of the inspirations is a lot of animals have quite low-resolution eyes, so they have managed to navigate quite effectively without having to resort to ever increasing megapixel counts," Dr Milford said.
"We use low-resolution, low-quality images so that the algorithms don't actually get distracted by little details, but just look at the overall image as a crude sort of whole."
The innovation would be targeted at in-car navigation systems and integrating navigation systems into smartphones, he said.
The navigation algorithms would also be usable in robots.
"So you could put a small camera in the front windshield of your car hooked up to a little computer and as you drive around, it compares what that camera sees to images it has accessed in a database," Dr Milford said.
"Over a sequence of images it gradually builds up an idea of where you are in the world by comparing what it is seeing now to what it has in that image database."
Dr Milford said the technology, which was at an early prototyping stage, was comparable with satellite-based GPS navigation and in certain conditions could be more accurate.
"It wouldn't replace GPS completely, but it would complement it," he said. "So you would have navigation everywhere instead of just in certain situations."
You May Also Like