Development of intelligent algorithms for simultaneously localization and mapping

Robots are autonomous mechanisms that can interpret the data they sensed from the environment and can decide how to react to the environment. The most common properties of an autonomous robot are: self sufficiency and sensing its surroundings. However, robots can have these abilities limited when compared to humans. It is unavoidable for the robots to interact with humans and the environment surrounding them. Commonly robots are used for risky processes in dangerous environment. Hence, they should be able to recognize the environment surrounding them and their own position within this environment. To achieve these abilities robots need to build a map of the environment. An autonomous robot, which is constructed in Yıldız Technical University, Computer Engineering Department, is used in this study. It begins its movement in an unknown position, builds a map of the environment, and at the same time estimates its new position. It is also able to recognize the starting point when it arrives that point again. In this study, a computer and the robot construct an autonomous system. This autonomous system successfully builds the map while calculating the new perceived robot location simultaneously. In literature the algorithms which provide these abilities to the robots, are called SLAM (Simultaneous Localization and Map Building) algorithms. The aim of this thesis is to develop a successful SLAM algorithm by using a robot equipped with only simple and cheap sensors. During map building and simultaneous localization, the robot can sense its environment by its infrared sensors and can decide the path to follow by using the developed SLAM algorithm. The most frequent problems in SLAM algorithms are sensors? noise and odometry errors. To solve this problem in the SLAM, statistical estimate methods are used very often. In the thesis Sequential Monte Carlo (SMC) algorithm which is a well known particle filter application is used and promising results were obtained for the SLAM problem. Keywords: Simultaneously localization and mapping algorithms, sequential Monte Carlo approach, particle filters, mobil robot kinematics.