Cooperative Consensus-based Control of Multiagent Systems over Wireless Channels

在这个项目的第一阶段，我们研究了沟通和寻求共识之间的协同作用。在一个场景中，多个智能体的目标是在一个无线信道上达成共识，我们设计并评估了利用无线信道的上位特性进行函数计算的分布式算法。这些措施可以显著增加资源，特别是能源效率。这直接影响了我们计划在基于共识的队形控制背景下研究的两个主要权衡：(I)无线信道上信息交换的频率和准确性直接相关。在信息控制情景中，这可能意味着在某些情况下(例如，当代理人即将发生碰撞和需要快速反应时)，前者可能不得不以后者为代价增加前者。为了利用这种权衡，我们将研究基于事件触发的基于广播的通信体系结构。(2)增加发射功率总体上将改善网络的连通性。前者意味着单个广播的能量消耗增加，而后者可能导致更快的融合，因此所需的广播数量更少。通过调整发射功率来使用这种权衡的信息控制问题也将被正式调查。除了利用干扰，我们的目标是利用(可能是隐藏的)不同代理人传递的信号结构。例如，这种时空结构的知识可以用于在传输之前压缩稀疏信号，或者从噪声传输中恢复更准确的信息，从而减少在资源有限的无线信道上交换的信息量。最后，我们打算通过研究一个实际的队形控制问题来证明我们的概念的适用性。这些个体机器人是专门设计的轮式机器人，具有非完整动力学，可以配备无线电收发机。为了利用物理层上的信号干扰，我们将采用基于模拟的全双工系统。

In the first phase of this project, we have investigated synergiesbetween communication and consensus seeking. In a scenario, wheremultiple agents aim at achieving consensus over a wireless channel, wehave designed and evaluated distributed algorithms that exploit thesuperposition property of the wireless channel for functioncomputation. These allow for a significant increase in resource, andin particular energy efficiency. This directly affects two majortrade-offs that we plan to investigate in the context ofconsensus-based formation control: (i) frequency and accuracy ofinformation exchange over wireless channels are directly related. In aformation control scenario, this may imply that in certain situations(e.g., when agents are about to collide and rapid reactions arerequired), the former may have to be increased at the cost of thelatter. To exploit this trade-off, we will investigate event-triggeredbroadcast-based communication architectures. (ii) Increasing transmitpower will in general improve connectivity of the network. The formerimplies an increase in energy consumption for individual broadcasts,whereas the latter may lead to more rapid convergence and therefore asmaller number of required broadcasts. The use of this trade-off information control problems by adapting transmit power will also beformally investigated. In addition to harnessing interference, we aimat taking advantage of (possibly hidden) structures in signalstransmitted by different agents. The knowledge of such spatio-temporalstructures can, for example, be utilized for compressing sparsesignals before transmission, or recovering more accurate informationfrom noisy transmissions, hence reducing the amount of informationexchanged over the resource-limited wireless channels. Finally, weintend to demonstrate the applicability of our concepts byinvestigating a practical formation control problem. The individualagents are specifically designed wheeled robots with nonholonomicdynamics that can be equipped with radio transceivers. To exploitinterference of signals on the physical layer, we will employanalog-based Full-Duplex systems.