Adaptive Positioning System Design Using AR Markers and Machine Learning for Mobile Robot

Positioning is an important issue in areas such as navigation, robotics, automation, aerospace and aviation. Indoor positioning systems are becoming increasingly important, especially in areas such as industry, resource management and crowd sourcing. the study a system for mobile robot positioning is designed, the working area is fixed, but the visual positioning system can be moved. The site estimate of the designed system was carried out by Machine Learning. Successful results were achieved with the MLP and SVM methods used. It has been observed that MLP type networks provide lower RMSE values due to the algorithms used in position estimation. Although this value is taken from very different positions and from different angles, the achievement of low results proves the adaptability of the system. Long-term operation of the designed system was also tested and it was observed that the system was able to estimate the position of the robot that passed the route points with an average error of 0.895 cm.