This monograph introduces novel responses to the different problems that arise when multiple robots need to execute a task in cooperation, each robot in the team having a monocular camera as its primary input sensor. Its central proposition is that a consistent perception of the world is crucial for the good development of any multi-robot application. The text focuses on the high-level problem of cooperative perception by a multi-robot system: the idea that, depending on what each robot sees and its current situation, it will need to communicate these things to its fellows whenever possible to share what it has found and keep updated by them in its turn. However, in any realistic scenario, distributed solutions to this problem are not trivial and need to be addressed from as many angles as possible. Distributed Consensus with Visual Perception in Multi-Robot Systems covers a variety of related topics such as: * distributed consensus algorithms; * data association and robustness problems; * convergence speed; and * cooperative mapping. The book first puts forward algorithmic solutions to these problems and then supports them with empirical validations working with real images. It provides the reader with a deeper understanding of the problems associated to the perception of the world by a team of cooperating robots with onboard cameras. Academic researchers and graduate students working with multi-robot systems, or investigating problems of distributed control or computer vision and cooperative perception will find this book of material assistance with their studies.

Vision-based control of wheeled mobile robots is an interesting field of research from a scientific and even social point of view due to its potential applicability. This book presents a formal treatment of some aspects of control theory applied to the problem of vision-based pose regulation of wheeled mobile robots. In this problem, the robot has to reach a desired position and orientation, which are specified by a target image. It is faced in such a way that vision and control are unified to achieve stability of the closed loop, a large region of convergence, without local minima and good robustness against parametric uncertainty. Three different control schemes that rely on monocular vision as unique sensor are presented and evaluated experimentally. A common benefit of these approaches is that they are valid for imaging systems obeying approximately a central projection model, e.g., conventional cameras, catadioptric systems and some fisheye cameras. Thus, the presented control schemes are generic approaches. A minimum set of visual measurements, integrated in adequate task functions, are taken from a geometric constraint imposed between corresponding image features. Particularly, the epipolar geometry and the trifocal tensor are exploited since they can be used for generic scenes. A detailed experimental evaluation is presented for each control scheme.

This work examines the challenges of distributed map merging and localization in multi-robot systems, which enables robots to acquire the knowledge of their surroundings needed to carry out coordinated tasks. After identifying the main issues associated with this problem, each chapter introduces a different distributed strategy for solving them. In addition to presenting a review of distributed algorithms for perception in localization and map merging, the text also provides the reader with the necessary tools for proposing new solutions to problems of multi-robot perception, as well as other interesting topics related to multi-robot scenarios. The coverage is largely self-contained, supported by numerous explanations and demonstrations, although references for further study are also supplied. The reader will not require any prior background knowledge, other than a basic understanding of mathematics at a graduate-student level.