Indoor Navigation of a Service Robot Platform Using Multiple Localization Techniques Using Sensor Fusion

Abstract

The adversities of navigating in an unknown environment are challenging for autonomous robots as they not only rely on the quality of the sensor data but also on the type of environment. The errors can cumulatively result in a catastrophic failure of the purpose. This paper proposes a way for an autonomous agent to navigate in an indoor structured environment by minimizing the errors. A 2 DoF differential drive robot is developed and the map-building process is done wirelessly in the teleoperation mode using bluetooth. In the navigation mode, various approaches are tested for localizing such as the Extended Kalman Filter, Unscented Kalman Filter and a particle filter called the Adaptive Monte Carlo Localization which uses KLD (Kullback-Leibler Distance) as the sampling method. The designed navigation system is accurate and also time saving, thereby increasing the efficiency of the robotic system. All the implementation has been done using the Robotic Operating System because of the available packages that satisfy the technical aspects of the paper and further changes have been made in them to make them suitable for use in this scenario.