A Platform for Indoor Localisation, Mapping, and Data Collection using an Autonomous Vehicle

Abstract: Everyone who has worked with research knows how rewarding experimenting and developing new algorithms can be. However in some cases, the hard part is not the invention of these algorithms, but their evaluation. To try and make that evaluation easier, this thesis focuses on the collection of data that can be used as positional ground truths using an autonomous measurement platform. This should assist Combain Mobile AB in the evaluation and improvement of their Wi-Fi based indoor positioning service. How and which parts of the open-source community’s work in the Robot Operating System (ROS) project to utilise is not obvious. This thesis therefore sets out to build a Minimum Viable Product (MVP) which is capable of supporting two different use cases: measure and explore inside an unknown environment, and measure inside a known environment given a map. This effectively leaves Combain with a viable product, and indirectly helps the community by aiding it in comparing and recommending the best tools and software libraries for the task. The result of this thesis ends up recommending the following for measuring inside an unknown environment: the Simultaneous Localisation And Mapping (SLAM) algorithm Google Cartographer for navigation, and the exploration algorithm Hector Exploration for planning the exploration. To measure inside a known environment the following is recommended: the Adaptive Monte Carlo Localisation (AMCL) positioning algorithm and the Spanning Tree Covering algorithm.