İç ortamlarda 2 boyutlu mobil robot lokalizasyonu

Localization of an object is an important issue in order to achive navigation successfully. In this study, localization is used for a mobile robot navigation for indoor environment. A mobile robot is used in a wide range and also it has lots of missions. The mobile robot has to know its current poisition in order to perform its assignments such as path following, reaching target positon, etc. So, localization is a paramount case for a mobile robot in order to carry out its assignments. Localization with odometry technique is the most known in this area. Odometry is based on integration in order to find position parameters. So, this integration basis causes error parameter for each calculation. For each scan, the error paramater is added the previous error value and error is getting bigger with each environment. When the error increases, the difference between current position parameters and calculated position parameters also increase. Because of this, localization does not give reliable solution. In order to overcome this disadvantageous method, the necessity of new techniques occurs. One of the alternatives is scanning environment with a laser. To accomplish localization for mobile robots with laser scanning, position coordinates and rotation angle parameters are correlated during the process without any error parameters. In this study, localization technique for a mobile robot is performed for 2D indoor environment. In order to carry out this study, laser sensor is used to scan environment. When laser sensor scans environment and collect data, these data have to be transferred to the computer to process and apply localization procedure. The communation and transformation between laser sensor and computer is provided by serial interface RS422. In the computer, MATLAB programme is used to process data. The laser sensor scans the environment with 180̊ scanning angle and 1̊ resoluion. Thus, at the end of one scan 181 data point is transferred to MATLAB. First of all, these 181 points' are extracted to features with split and merge algorithm. So, less number of features are obtained instead of points. This split and merge algorithm is applied for each scan data. At the result of split and merge algorithm, there are feature maps for each scan. Features have to be associated between each consecutive two feature maps in order to obtain common fetaures. These common features provide calculating changes on the position in the next step, data relation. Data association is carried out by distance and angle algorithm. In this method, not for all features, only two chosen features for each maps are associated and the result is obtained via these features. At the end of this step, there is one couple random feature for the first map and another couple feature in the second map which correspond to the features in the first map. In this way, associated features are obtained. In the final step, it is known that which feature in the first map match which feature in the second map. Therefore, the changes between these associated features gives the changes for the position in the same time. These changes are deal with translational on x coordinate, translational on y coordinate and rotation. In this step first of all, it is assumed that the start point cartesian coordinates is known for the first map as a reference point (0,0) and also the rotation angle is 0. Thus, for the first consecutive scans, the difference is obtained according to this reference point assumption and calculated new position values. Then, for the second consecutive scan, the new position parameters is calculated with the previous position values. Finally, in order to obtain rotation and translational paramaters between two featıure sets 'MATLAB Estimate Geometric Transform' function is used. This function matches point pairs. Thus, current position can be calculated with these parameters with konwledge of translation from the reference point and also its rotation according to the reference point.