Akademik Veri Yönetim Sistemi
Tezin Türü: Doktora
Tezin Yürütüldüğü Kurum: Orta Doğu Teknik Üniversitesi, Mühendislik Fakültesi, Elektrik ve Elektronik Mühendisliği Bölümü, Türkiye
Tezin Onay Tarihi: 2014
Öğrenci: MUHAMMED ORKUN ÖĞÜCÜ
Danışman: AFŞAR SARANLI
Özet:Legged robot platforms have distinct advantages over wheels in rough terrain and provide better mobility. Also in many applications, including military reconnaissance, disaster relief, hazardous site inspection, search and rescue applications benefit from the legged robots which is capable of moving safely at high speeds through rough natural terrain. However, with the increasing speed, fundamental difficulties like dynamic and mechanical limitations as well as control and computational limitations arise. Also sensor uncertainty and limited terrain data can result in unexpected and dangerous situations. At such situations, the data obtained by a single sensor may not be reliable. To provide certain and robust information, it is conventional that to integrate multiple sensor data which is known as sensor fusion. Gyroscopes and accelerometers are frequently used motion sensors at robotic applications. However, these sensors can only provide an accurate solution for a short period of time because of integration drift. To correct these bias errors, cameras are frequently used in robotics applications with integration of accelerometer and gyro data. However, vision sensors are considerably sensitive to shocks and instantaneous movements especially at high speed and accelerations resulted with motion blur effect. Due to the limited performance of deblurring methods, it is crucial to avoid the captured images affecting from visual disturbances. In this thesis, head stabilization problem of SensoRHex mobile robot platform which is dynamically stable autonomous hexapod robot with six compliant legs will be examined. Thesis works can be categorized into two groups; Sensing and Control. Accelerometer and gyroscope data will be integrated with camera data according to the sensor fusion algorithms. These sensory data will be used to estimate state space vector partially. At the final stage of the sensing works is disturbance and signal characterization. Control strategy will be determined after the examination of disturbance characteristics. But there are intuitive ideas about the disturbances which presence of predictable cyclic characteristic due to the compliant legs and passive filters which the camera will be mounted on. In that case, model predictive control can be used as a controller structure. Anyhow, it would be a good challenge to compare the different control strategies according to the performance evaluation by considering the effects of passive filter parameters. After selecting the control strategy, it can be attended to solve the main problem of thesis works; head stabilization. For this purpose, the primary objective is to maintain the state vector of the camera which attached to the head, as close as possible to the origin.