دانلود کتاب ادراک و یادگیری ربات – دیدگاه ناوبری آگاه از انسان و استقلال بلندمدت

In 2015, I received a task from Dr. Nicola Bellotto to let a professional cleaning robot identify the points representing humans in the point cloud generated by 3D lidar. Then, based on the research progress at that time, I developed a point cloud annotation tool, collected point cloud data, manually annotated thousands of point cloud frames, and then trained an SVM model, tested with the robot, tuned the training parameters, and trained the model again. Then suddenly the robot changed its deployment environment, and I needed to collect data again and annotate them again, over and over again. Later, in Cyprus, I complained to Nicola, why can’t the robot learn these boring point clouds by itself? After returning to Lincoln, Dr. Tom Duckett joined the discussion, and finally in July 2016, we completed the draft of our first “robot online learning” paper together.
Since 2015, robot perception and learning have been the main line of my research. Ten years later, the results of my own research and those I supervised were recog-nized at the Institut Polytechnique de Paris, and I was finally enabled to obtain the Habilitation à Diriger des Recherches in France. To this end, I needed to prepare a thesis to systematically summarize my research since my doctoral defense. This led to the writing of this book.
The pursuit of truly intelligent robots capable of operating effectively in complex and unstructured environments has driven significant advancements in both percep-tion and learning. While traditional robotics often relied on pre-programmed instruc-tions and carefully engineered environments, the increasing demand for mobile robots that can adapt, interact, and learn from their environment and experience has propelled the field towards embodied intelligence. This paradigm emphasizes the crucial role of physical embodiment in shaping perception, action, and cognition, leading to more robust and adaptable robotic systems.
This book aims to provide an overview of the latest research at the intersection of these crucial areas, with a particular focus on 3D lidar-based perception and robot online learning methods, with downstream tasks including human-aware robot navi-gation and long-term robot autonomy. It is structured as follows: Chap. 1 introduces the background of the research, the content positioning, and some related open-source resources. Chapter 2 discusses some benchmarking issues related to the field of embodied intelligence and mobile robotics. Chapter 3 introduces robot perception, especially the object detection and tracking based on 3D lidar with contemporary characteristics. Chapter 4 introduces robot learning, especially robot online learning methods with strong embodied intelligence features. Chapter 5 summarizes the book and provides prospects for future research and application directions.
This book is intended for researchers and practitioners already familiar with the fundamentals of mobile robotics, including graduate students, Ph.D. candidates, post-doctoral researchers, and industry professionals working in related fields. This book is not a textbook in the traditional sense; it does not delve into basic introductory material. Instead, it focuses on presenting the core methodologies and principles that drive cutting-edge research, providing a structured overview of key concepts and recent developments. Detailed experimental results and specific implementation details are intentionally omitted to maintain a focus on the broader theoretical and methodological underpinnings, encouraging readers to consult the original research papers for in-depth analysis and practical implementation strategies.
This book would not have been possible without the contributions of numerous researchers whose work forms the foundation of this field. We gratefully acknowledge their invaluable contributions and hope that this book serves as a useful resource for those seeking to understand and contribute to the future of robot perception and learning. We also extend our sincere thanks to any reviewers who provided feedback on drafts of this material. We sincerely hope that this book inspires new research directions, facilitates collaboration within the community, and ultimately contributes to the advancement of embodied intelligence and mobile robotics.