An introductary view into Simultaneous LocalizationAnd Mapping
Abstract: In recent years autonomous vehicles have gained attention as being the potential future for the automotive industry. A central part of an autonomous vehicle is the way that the vehicle locates itself relative to the surrounding environment. This localization can be efficiently performed simultaneously as building a map of the environment, a process called Simultaneous Localization And Mapping or SLAM. In this thesis, the basic algorithms that lay the foundation for the current state of the art of SLAM, and their refinements are studied with the goal of procuring a comprehensive list. The secondary goal is to create a simulation platform for assessing the performance of different algorithms. The thesis is commissioned by automotive consultancy AVL that will use the results to build their knowledge base within the autonomous field. Each algorithm that is mentioned has been looked at from a historical, mathematical and application perspective. This has resulted in a literature study containing thirteen algorithms and a table containing a categorization for easy and quick comparison. Categories of SLAM algorithms are identified to be map type, complexity, hypothesis and basic algorithm. Among the basic algorithms, the classic Kalman Filter and Particle filters have been found, along with modern Maximum a Posteriori algorithms. Furthermore, EKF and fastSLAM have been simulated in a Matlab environment for comparison. A development and simulation platform has also been developed in software frameworks Gazebo and ROS. This platform ensures modularity where algorithms can be exchanged and simulations can be run simultaneously as a real life implementation. The results from the simulations show that the simulation in Gazebo can produce similar results as the traditional simulations in Matlab. It is difficult to know exactly which algorithms that are used in commercial applications. Using the information from the literature review, a discussion is made regarding which algorithms may be suitable for real life usage. The conclusion is that most likely, combinations of several algorithms are used in order to leverage the benefits from each, as well as mitigate weaknesses.
AT THIS PAGE YOU CAN DOWNLOAD THE WHOLE ESSAY. (follow the link to the next page)