Search results

This article is based on a lecture to the Fellowship of Engineering by the head of the Department of Engineering Science, Professor Mike Brady, and an interview with the department's specialist on non‐vision sensing and data fusion, Dr Hugh Durrant‐Whyte.

Examines one of the main problems of mobile robot navigation: determining exactly where the robot is at all times. Describes the most important algorithm in localization: the extended Kalman filter. Looks at the simplest type of navigation using a system of fixed beacons in conjunction with a special sensor located on the vehicle and also the use of “natural beacons”. Discuss the problems of building and maintaining a map for the vehicle. Concludes that a complete solution to mobile vehicle localization is a long way off.

A car capable of self‐diagnosis, a central heating system that actually keeps your house at a constant temperature, a pointer able to pick out individual atoms and a temperature sensor capable of detecting temperature changes that last only eight millionths of a second were just some of the revelations made at the Institute of Physics' Sensors Conference held in Edinburgh, 23–25 September 1991.

By tradition, sensors are used to measure one desired parameter; all other parameters influencing the sensor are considered as interfering inputs, to be eliminated if possible. Hence most of existing sensors are specifically intended for measuring one parameter, e.g. temperature, and the ideal temperature sensor should be as immune to all other parameters as possible. True, we sometimes use primitive sensor fusion, e.g. when calculating heat flow by combining separate measurements of temperature difference and of fluid flow.