Vision-Based Macro-Micro Mobile Manipulation

Abstract

Increasing demand for robots with broader applications has created many challenges for the commercialization of recent achievements in the fields of robotics and perception. This research focuses on formulating a solution for these challenges and to integrate vision-based information in the real-time control loop of nonholonomic Mobile Manipulators (MMs) to fulfil requirements for modularity and reusability of control subsystems. The proposed solution answers two research questions: how can visual camera feedback be improved during motion and how can dissimilar subsystems of mobile platforms and manipulators be coordinated?<br/><br/>Seven publications have been developed to address these issues from various perspectives. It is necessary to carefully address changes in camera outputs, discontinuities, variable latencies, and quantization errors and uncertainties. In particular, there are challenging issues about output of numerical iterative methods used for classic controllers with assumptions of well-behaved and continuous setpoints and feedbacks. Due to the multidisciplinary nature of this problem, the primary objective is to ensure that benefits are realized by lessens learned from the everyday motions of humans as a smart agent. Thus, the camera should act as the eyes, the mobile platform (i.e., the macro robot) as the body trunk, and the manipulator (i.e., the micro robot) as the arm.<br/><br/>The thesis analyzes two main research questions considering practical issues caused by the position feedback of several object detection methods. It is shown that a complementary M-estimator and accurate synchronization of signals can significantly smoothen a visual feedback for the control purposes. Moreover, another focus of this research is to formulate coordination of macro robot and micro robot based on the performance of the visual feedback and vehicle capabilities.<br/><br/>The solutions addressed in this thesis suggest the use of online evaluations of sensor feedback to determine its quality. This evaluation plays a key role in both macro – micro coordination and integration of visual feedback into robots’ motion controllers.