Networked Collaborative Multi-Agent Cyber-Physical Systems

信息物理系统（CPS）涉及计算、通信、网络和控制等技术。尽管在开发CPS技术方面付出了巨大的努力，但在CPS技术的基础研究方面，跨学科的努力有限。在这个由德累斯顿工业大学、南佛罗里达大学和北卡罗莱纳州立大学共同参与的联合项目中，我们将在计算机、通信和控制这三个网络物理领域的融合领域进行创新，简称CoCoCo。在所有这些学科中，跨领域的相互关系将被考虑在内。计算问题将在工作包A （WP）中解决。接下来的方法将主要是人工智能，特别是强化学习。原因是强化学习已被证明是一种使代理能够与其环境进行交互的适当方法。通信将在WP中进行探讨[B]。网络虚拟化是我们选择的方法，因为它允许对控制代理和网络方面的复杂动态和相互依赖性进行适应性。WP将考虑控制[C]。研究无线网络控制的框架将是网络控制系统理论，因为它联合研究了环上网络控制系统设计中的通信和控制两个方面。最后，在WP[D]中，我们将把每个工作包的主要成果放在一起，并通过实验提出联合评估。我们期望通过对当前理论极限的实验评估来获得和提供见解，所有这些都汇聚在这个引人注目的网络化CPS领域。

A cyber-physical system (CPS) involves technologies of computing, communication, more precisely, networking, and control. In spite of great efforts in developing technology for CPS, only limited cross-disciplinary efforts toward the fundamental investigations on CPS technology. In this joint project, involving TU Dresden, University of South Florida and North Carolina State University, we will pursue innovation in the three disciplines converging in Cyber-Physical Networking, namely computing, communication and control or CoCoCo for short. In all of these disciplines, the cross-domain interrelationships will be taken into account. Computing will be tackled in Work Package A (WP). The methodology followed will be mainly artificial intelligence, specifically reinforcement learning. The reason is that RL has proved to be an appropriate methodology for enabling agents to interact with their environment. Communications will be approached in WP[B]. Network virtualization is our method of choice since it allows adaptability to the complex dynamics and interdependencies of control agents and networking aspects. Control will be considered in WP[C]. The framework used for studying control over wireless networks will be network control system theory, since it jointly studies the communication and control aspects in the design of control systems with networks on the loop. Finally, in WP[D] we shall bring the main outcomes of each of these work packages together and propose a joint evaluation through experimentation. We expect to obtain and provide insights through experimental evaluation of the limits of the current theories, all converging in this compelling field of networked CPS.