Multisensor-based Navigation and Map Building
Prof. Huosheng Hu
A crucial capability for an autonomous mobile robot is the estimation of its position in the real world. It is well known that a mobile robot cannot rely solely on the dead-reckoning method to determine its position because dead-reckoning errors are cumulative. Using external sensors to observe useful features from the environment becomes necessary. There are two kinds of the environment features that can be used in localisation: one is the artificial beacon populated at a known position in a structured environment, and another is the natural beacon that is abstracted from the real world. The use of artificial beacons is to simplify the localisation process by measuring the robot motion relative to pre-placed beacons.
The use of natural beacons enables mobile robots to operate in an unstructured environment, but pays a high cost to interpret sensory data and obtain useful features. This project is to investigate how to integrate multiple simple and inexpensive sensors in the localisation process for a mobile robot to observe both features in an environment. An extended Kalman filter will be used to fuse the data from multiple sensors. As a mobile robot traverses its environment, it should be able to observe both artificial and natural beacons to update its position and the environment map continuously.
This project is funded by Robotic Sciences Ltd. (£8,000) and collaborated with NavTech Electronics Ltd.
|Monday, 15 October 2007|