CoCPN: Cooperative Cyber Physical Networking

拟议项目CoCPN的目标是研究一种新的合作方法，用于控制系统和通信系统的综合操作。我们设想，控制系统和通信系统都可以对当前情况的变化做出自适应反应。所谓的CoCPN-translator将两个系统耦合起来，并将控制需求转换为通信属性，反之亦然。从控制系统方面，我们采用分散的方法局部近似地解决优化由通信系统连接的所有控制回路行为的全局问题。我们采用基于事件的方法来调整通信需求。采用基于序列的方法考虑了通信系统的不确定性。我们研究了假设时不变通信特性和时变通信特性。通过将系统建模为马尔可夫跳变线性系统，推导出最优线性控制器，解决了第一类问题。后者涉及一个非凸优化问题，被建模为一般随机混合系统（SHS）。优化将通过成本和策略近似策略以及采用基于场景的方法进行研究。对于网络，我们假设需要服务CPS和非CPS混合流量。因此，我们采用流量区分，为了使其有效，需要一个允许控制的支持。在交通管理方面，我们采用集中和分散的混合方法。所谓的观察者对无线域有一个大致的看法，因此，可以提供合适的路由，然后将通过源路由使用。为了正确处理网络中的竞争流量，我们将研究一种机制4拥塞控制，该机制利用网络中的知识，并将针对基于事件的流量而不是通常在互联网中看到的流量流进行定制。因此，CoCPN应该为网络物理网络提供一个灵活的框架，可以应对不同类型的流量，不需要网络过度供应，控制回路对通信系统的实时性要求更宽松。

The goal of the proposed project CoCPN is to investigate a novel cooperative approach for an integrated operation of control systems and communication system. We envision that both, the control system as well as the communication system can react adaptively to changes in the current situation. The so called CoCPN-translator couples both systems and translates control requirements into communication properties and vice versa. From the control system side, we employ a decentralized approach to locally approximately solve the global problem of optimizing the behavior of all control loops, connected by the communication system. We apply an event-based approach to adjust the communication demand. The non-deterministic properties of the communication system are taken into account by a sequence-based approach. We study assumed time-invariant communication properties, as well as time-varying communication properties. The problem induced by the first is solved by modelling the system as a Markov Jump Linear System and derive an optimal linear controller. The latter invokes a nonconvex optimization problem and is modelled as general Stochastic Hybrid System (SHS). The optimization will be studied by cost and policy approximation strategies and by employing a scenario-based approach.Regarding the network we assume that mixed CPS and non-CPS traffic needs to be served. Therefore, we apply traffic differentiation, which in order to be effective needs to be supported by an admission control. For traffic management, we follow a mixed centralized and de-centralized approach. A so-called observer has an approximate view on a wireless domain and, thus, can provide suited routes that then will be used via source routing. In order to properly deal with competing traffic in the network, we will investigate a mechanism four congestion control that utilizes knowledge from the network and will be tailored to event-based traffic instead of traffic streams as usually seen in the Internet. As a result CoCPN should provide a flexible framework for cyber physical networks that can cope with different types of traffic, does not need network overprovisioning, and the control loop has looser real-time requirements regarding the communication system.