AUTONOMOUS ROBOT NAVIGATION USING THE UTILITY FUNCTION METHOD AND MICROSOFT KINECT

Abstract

In this thesis, a system for autonomous navigation using the Microsoft Kinect sensor and the utility function (UF) method for decision-making, has been developed. In the UF method an artificial brain decides what control system procedures to activate or deactivate, based on their utility. The system uses the Kinect sensor for obstacle avoidance and localization. The Kinect sensor readings are matched to a predefined map, using a scan matching algorithm based on the Hough transform, in order to correct pose estimates acquired through odometry. The A-algorithm is used for path planning and the algorithm is applied on a grid representing the operating area accessible to the robot.The results show that the UF method can be applied in systems for autonomous navigation, using the Kinect sensor for localization and obstacle avoidance. The Kinect sensor has proven to be a useful alternative to more expensive range sensors, despite its limitations in terms of field of view, range, and accuracy. On the other hand, the results indicate that improvements are needed, in order to improve the robustness of the system. During testing, the robot completed on average 87.6 % of a 45.1 m long, predefined course and on average 121 m when choosing random targets. Tuning the parameters of the decision-making system and improving the maneuverability at low speeds are examples of future work that could increase the performance of the navigation system.