Enhancing hybrid visual servo control by probabilistic techniques
Başlık çevirisi mevcut değil.
- Tez No: 611517
- Danışmanlar: PROF. DR. DANIŞMAN YOK
- Tez Türü: Doktora
- Konular: Bilgisayar Mühendisliği Bilimleri-Bilgisayar ve Kontrol, Computer Engineering and Computer Science and Control
- Anahtar Kelimeler: Belirtilmemiş.
- Yıl: 2007
- Dil: İngilizce
- Üniversite: Osmania University
- Enstitü: Yurtdışı Enstitü
- Ana Bilim Dalı: Belirtilmemiş.
- Bilim Dalı: Belirtilmemiş.
- Sayfa Sayısı: 236
Özet
Özet yok.
Özet (Çeviri)
Robotic research aims at performing tasks autonomously in complex and dynamic environments. Visual input has been demonstrated to be very useful in large number of these tasks. In robot vision systems, it has been argued that closing the control loop using visual feedback, has more effect than increasing the accuracy of the electro-mechanical components. Closing the control loop of a robot arm using visual information, is now referred to as visual servoing. Based on the nature of the visual information used, visual servoing algorithms are currently classied as 2D, 3D or hybrid. In this thesis, issues related to enhancing the visual servoing algorithms using probabilistic techniques are investigated. Most of the hybrid (2D/3D) visual servoing schemes are designed to obtain some desired characteristics like keeping features visible in the eld of view during the servoing process. In general, visual servoing control law is expected to produce a control signal with satisfactory performance on multiple fronts. Performance criteria could also related to the visibility of the features, shortest path in the 3D space, avoiding the local minima, faster convergence and continuity of the control signal. In addition to these issues, visual servoing systems have to perform well in presence of uncertainty that exist in the real world. We analyze the proposed visual servoing schemes in these dimensions. The uncertainty of robot vision system results from different sources including unpredictable environments, unreliable input of sensors and actuators, apart from the computational and modeling errors. Modeling uncertainty for performance enhancement can be of immense help in both indoor and outdoor robotic applications. Probabilistic techniques are shown to be useful for obtaining desirable performance in many critical tasks. Existing hybrid visual servoing schemes integrate partial information from 2D (image space) and 3D (Cartesian space). This thesis addresses two subproblems. First, design of a hybrid control law that uses the complete information in 2D and 3D to control the full degrees of freedom of the robot motion. Here, we consider two cases, (i) integrating features from different domains with an optimization framework and (ii) integrating actions produced by different algorithms using a boosting algorithm. Second, we propose algorithms for estimating the 3D parameters (pose and depth), which are involved in the control law using probabilistic technique (Bayesian ltering). We use particle lters due to its superiority over the common extended Kalman lter methods in taking care of non-Gaussian noise. In summary, a hybrid visual servoing method that addresses a set of challenges in the visual servoing process is presented in this thesis. Probabilistic methods are shown to be useful in enhancing the visual servo control algorithms
Benzer Tezler
- Güvenlik kameralarındaki yüz görüntülerinin süper çözünürlüklenetleştirilmesi
Face enhancement in surveillance systems using super-resolutiontechniques
ALİ HÜSAMEDDİN ATEŞ
Yüksek Lisans
Türkçe
2024
Bilgisayar Mühendisliği Bilimleri-Bilgisayar ve KontrolSakarya ÜniversitesiBilgisayar Mühendisliği Ana Bilim Dalı
DR. ÖĞR. ÜYESİ HÜSEYİN ESKİ
- Bayesian optimization strategies for human-in-the-loop systems: Theory and applications in physical human-robot-interaction
İnsanın optimizasyon döngüsüde olduğu sistemler için bayes optimizasyon stratejileri: Teori ve fiziksel insan-robot etkileşimi uygulamaları
HARUN TOLASA
Yüksek Lisans
İngilizce
2024
Mekatronik MühendisliğiSabancı ÜniversitesiMekatronik Mühendisliği Ana Bilim Dalı
PROF. DR. VOLKAN PATOĞLU
- Remote sensing, tracking and imaging inside MRI systems
MRI sistemlerinde uzaktan algılama, izleme ve görüntüleme
MEHMET BERK BİLGİN
Yüksek Lisans
İngilizce
2021
BiyomühendislikKoç ÜniversitesiElektrik ve Elektronik Mühendisliği Ana Bilim Dalı
PROF. DR. METİN SİTTİ
PROF. DR. HAVVA YAĞCI ACAR
- Deep learning approaches for autism spectrum dis-order diagnosis: Ensemble archtectures and multi-modal analysis
Otizm spektrum bozukluğu teşhisi için derin öğrenme yaklaşımları: Topluluk mimarileri ve multimodal analiz
ASSIL JABY
Yüksek Lisans
İngilizce
2023
Bilgisayar Mühendisliği Bilimleri-Bilgisayar ve KontrolBahçeşehir ÜniversitesiYapay Zeka Ana Bilim Dalı
PROF. DR. MD BAHARUL ISLAM