A sensor-based personal navigation system and its application for incorporating humans into a human-robot team

Abstract

In this thesis methods for the sensor-based localisation of human beings are studied. The thesis presents the theory, test results and a realisation of the methods, which is called PeNa. PeNa is further applied to incorporate a human into a human-robot team that performs a simulated search and rescue task. 
\n Human-robot teamwork provides the vision for this thesis. Furthermore, the PeLoTe project and its search and rescue task provided the primary motivation for the research. However, the major part of this work and contribution is on sensor-based personal navigation. 
\n The approaches studied for personal navigation systems are based on sensor-based dead reckoning, laser-based dead reckoning, and map-based localisation. Sensor-based dead reckoning is based on heading estimation using a compass and gyro and step length estimation. Two alternative step length estimation methods are presented, ultrasound-based and accelerometer-based. Two laser dead reckoning methods are presented; a pose correlation method and a combined angle histogram matcher with position correlation. Furthermore, there are three variations for map-based localisation based on the well-known Monte Carlo Localisation (MCL): topological MCL, scan-based MCL, and a combined MCL method. 
\n As a result of the research it can be stated that it is possible to build a personal navigation system that can localise a human being indoors using only self-contained sensors. The results also show that this can be achieved using various combinations of sensors and methods. 
\n Furthermore, the personal navigation system that was developed is used to incorporate a human being into a human-robot team performing a search and rescue task. The initial results show that the location information provides a basis for creating situational awareness for a spatially distributed team.