REKF-Localization: Node localization using mobile robots in delay-tolerant sensor networks

Introduction

We are developing a novel localization system for sensor networks in which we can use a mobile robot to perform location estimation for sensor nodes it passes by, using the radio signal strength of the messages received from them. Thus, we eliminate the processing constraints of static sensor nodes; and also the need for static reference beacons. By using a mobile robot, we also eliminate many of the problems associated with using RSSI measurements, such as small-scale fading.

To solve the localization, we use a novel mathematical technique, the Robust Extended Kalman Filter (REKF). REKF is computationally efficient and more robust than the more commonly used traditional Kalman Filter. We have implemented the REKF localizer on a LegoRobot, and Crossbow's motes and the Stargate platform. Localization accuracies vary between 30 cm (small indoor areas) to 1m (large indoor areas).

People (in alphabetical order)

Nirupama Bulusu , Sanjay Jha, Budi Mulyawan, Pubudu Pathirana, Andrey Savkin, Dang Xuan Thanh

Papers

Node localization using mobile robots in delay-tolerant sensor networks (PDF)
IEEE Transactions on Mobile Computing (TMC), to appear, Fall 2004.

Demos

Demonstration of Sensor Localization Using Mobile Robots (Windows media player)

Software

Download localization software here.
Warning, this is a very early release! We are currently working on a more adaptive version of the localization software to be released later.

This research is supported by National ICT Australia Limited.