Model-Predictive Cyber-Physical Networking

MPCPN项目第一阶段已经确立了模型预测型网络-物理网络的基本设计原则。主要结果是在受控物理子系统和通信网络之间建立了一个新的接口，网络通过该接口将其预测的决策和延迟轨迹传达给子系统对象控制器。控制器可以使用这些信息来极大地提高其性能，并预见即将到来的瓶颈。第二阶段的目标是进一步开发框架，并考虑具有更大规模的系统，从而解决扩展、集群、学习等方面，以及更一般的系统互联拓扑。此外，第二阶段的研究将加强对预测网络控制和预测子系统控制器中的优化问题如何进一步交织在一起，以相互响应和适应其他部件的预测状态的问题的关注。预计增强的交织可以进一步提高系统性能，并可以避免遇到容量限制。第二阶段的研究重点受到两个用例的激励，也将得到验证和演示，这两个用例分别是“相互作用的自动赛车”和“成群的移动机器人”。

Phase 1 of MPCPN project has established the underlying design principles of model-predictive cyber-physical networking. Main result is a novel interface between the controlled physical subsystems and the communication network by which the network communicates its predicted decisions and delay trajectories to the subsystem plant controllers. This information can be used by controllers to greatly enhance their performance, and to foresee upcoming bottlenecks. The ambition of the Phase 2 is to further develop the framework and to consider systems with larger size, thus addressing the aspects of scaling, clustering, learning, etc., as well as more general topologies of system interconnection. In addition, the research in Phase 2 will strengthen the focus on the questions of how the optimization problems within the predictive network control and in the predictive subsystem controllers can be further intertwined to mutually respond and adapt to the predicted status of the other components. It is anticipated that an enhanced interleaving can further improve the system performance and can avoid to run into capacity limits. Phase 2 research focus is motivated by and will also be validated as well as demonstrated by two use cases, namely "interacting autonomous racing cars" and "swarms of moving robots".