Indoor navigation and mapping with a DaNI mobile robot

OData support
Supervisor:
Dr. Kiss Bálint
Department of Control Engineering and Information Technology

Abstract

Nowadays, mobile robots are quickly moving from the laboratories and factories to households, where they have to explore and navigate autonomously in different environments, knowing only their own properties and structure a priori.

This BSc thesis explores the feasibility of autonomous navigation with a low cost robot. Low cost means, that the robot uses only inexpensive sensors. Autonomous mobile robots need a map of the environment for navigation.

A good map requires precise computations on the robot position and orientation. In order to solve the task, I have chosen the methodology of the occupancy grid. The mapping and navigation approach of this method compensates the limited resolution and range of the ultrasonic distance sensors.

A simulation environment is used to implement and test the subprograms first, before using them in real environment. The real-time tests were executed once the algorithms where successfully implemented in simulation. In the simulator, the robot perfectly maps and navigates in the virtual area, but only after using simulated GPS data as reference. Without this information, the configuration data contains odometry errors, since the data is calculated only with internal sensors. After using the GPS data, all these problems were solved or reduced to an optimal level.

The real robot was tested in an indoor office environment. In the real environment, problems occurred with the primary internal sensors; the odometry error was even bigger and caused more problems than in the simulated environment. In the real robot, there was no way to use a GPS. However, the robot was equipped with a compass which could be used instead of the GPS as external reference. With the help of the compass, I created a Kalman filter to reduce the effects of the odometry errors. Therefore, despite the occurred problems, the robot was able to generate a map with acceptable precision.

Navigation, once the map is created, was working reliably, as long as the more accurate map is used, which was generated with the Kalman filter. Hence this BSc thesis also shows that despite the limited amount sensory capacity and precision, it is still possible to perform navigation and mapping in an adequate way.

Downloads

Please sign in to download the files of this thesis.