Sensor fusion is the tricky task of combining different data streams to give an output that is better than the sum of the parts. Let me tell ya, it ain’t easy. Too often, too much data just confuses things.

So hats off to Cherif “Very” Smaili at The French Institute for Research in Computer Science and Control in Nancy and his mates who have squeezed, merged and melded the data from several sources to improve the way autonomous navigation systems determine where they are.

Anybody who’s ever used GPS knows that at best, it provides intermittent data on position and direction. So Very Smaili has combined GPS information with data from the car’s odometer to give continuous navigation infomation.

If the GPS signal goes down, the onboard computer works out where it is on a digital map using the distance travelled by the car.

Of course, that still leaves some uncertainty should the car come to a junction – there ain’t no way of telling from distance measurements which way the vehicle has gone. So the onboard keeps every possibility open until it can unambiguously work out where it is.

That looks simple and useful, which ain’t a bad combination.

Ref: arxiv.org/abs/0709.1099: Multi-Sensor Fusion Method using Dynamic Bayesian Network for Precise Vehicle Localization and Road Matching

This entry was posted on Tuesday, September 18th, 2007 at 12:25 am and is filed under Mean machines. You can follow any responses to this entry through the RSS 2.0 feed. Both comments and pings are currently closed.