Li S., Zhang Y. Neural Networks for Cooperative Control of Multiple Robot Arms

Springer, 2018. — 86 p.In the past decades, recurrent neural networks have been widely investigated by many scientific and engineering communities. In particular, Hopfield neural network, originally designed for real-time optimization, triggers the studies on recurrent neural networks as a powerful online optimization tool. Recurrent neural network-based optimization methods have become independent research direction in online optimization field. It was later found that the kinematic control of robot arms can be formulated as a constrained convex quadratic programming (QP) problem, which can be efficiently solved via neural network approaches. Although many results have been reported on the kinematic control of a single robot arm, there are much less existing results on the cooperative control (especially fully distributed cooperative control) of multiple robot arms. It should be noted that the cooperation of multiple robot arms becomes demanding when tasks in industry to be completed by robot arms become more complex. In view of the successful applications of neural networks on the kinematic control of a single robot arm, it is interesting to extend existing results to the case with multiple robot arms.
In this book, mainly focusing on the cooperative control of multiple robot arms, we design, propose, develop, analyze, model, and simulate various decentralized or distributed neural network models and algorithms. We first present a neural network model for the visual servoing of a single robot arm so as to provide some basic ideas about the kinematic control of robot arms for the potential readers. Then, we move on to the cooperative control of multiple robot arms. Specifically, we investigated neural network models and algorithms for decentralized control of a group of multiple robot arms with a star control topology, hierarchical control topology, and fully distributed control topology, respectively. For all the models and algorithms, the corresponding theoretical analyses are presented, and the corresponding modeling is illustrated. Besides, the related computer simulations with various illustrative examples (most of which are related to the PUMA 560 industrial robot arm) are performed to show the effectiveness of the recurrent neural network models in achieving the cooperative control of multiple robot arms.
The idea for this book on solving cooperative control of multiple robot arms was conceived during the classroom teaching as well as the research discussion in the laboratory and at international scientific meetings. Most of the materials in this book are derived from the authors’ papers published in journals, such as IEEE Transactions on Neural Networks and Learning Systems. It is worth pointing out that since the early 1980s, research on recurrent neural networks has undergone exponential growth, and many new theoretical concepts and tools (including the authors’ ones) have been obtained. Meanwhile, the new results have already been successfully applied to solving practical problems. To make the contents clear and easy to follow, in this book, each part (and even each chapter) is written in a relatively self-contained manner.Neural Networks Based Single Robot Arm Control for Visual Servoing
Neural Networks for Robot Arm Cooperation with a Star Control Topology
Neural Networks for Robot Arm Cooperation with a Hierarchical Control Topology
Neural Networks for Robot Arm Cooperation with a Full Distributed Control Topology