Eş zamanlı konum belirleme ve harita oluşturma probleminin çözümünde kullanılan yöntemlerin iyileştirilmesi
Özet
Simultaneous Localization and Mapping (SLAM), was introduced in the early 90s, is known as a method for the building the map of the environment and simultaneously estimating the robot/ autonomous vehicle path in the unknown position information and unkown environment conditions. SLAM applications have some basis problems such as; measurement noise minimization, complexity of the data association in case of building map of the environment, building the large environment map and method complexity depending on the real time processing. In this thesis, to solve these basic problems encounterd in general practice SLAM, an improved FastSLAM based method has been proposed. Unlike the approaches that exist in the literatüre, adaptive central difference Kalman filter A-(CDKF) and joint probabilistic data association (JPDA) algorithms have been implemented in conventional FastSLAM structure. So, thanks to new structure of the FastSLAM, both presented as a solution for the problem of SLAM data association and obtained high accuracy results with less processing time. As well as the obtained results have been compared with the partical swarm optimization based and differential evolution based fastSLAM approaches which has been again improved in thesis studies. Experimental results have shown that the proposed method has property of the real time implementation, both has provided superior than optimize based FastSLAM approaches and although having less particle, minimizing the processing time and obtaining robust results than FastSLAM II and U-FastSLAM.