基于动态事件触发策略和异步周期检测的分布式凸优化

Distributed Convex Optimization Based on Dynamic Event Triggering Strategy and Asynchronous Periodic Detection

摘要: 本文研究了多智能体系统的优化问题，在局部成本函数可微并且强凸的条件下，提出了分布式凸优化算法，使得智能体的状态渐近收敛到全局最优点。为了节省系统的通信成本及降低系统的能量消耗，本文引入了积分型动态事件触发机制。不同于传统的周期检测，本文研究的事件检测具有异步性，即每个智能体基于自己的时钟周期地检查事件触发条件。当满足事件触发条件时，相应智能体及其邻居更新它们的控制输入。此外，由于事件触发时刻的最小时间间隔是检测周期h，Zeno行为可以很自然地被排除。最后，数值仿真验证了算法的有效性。

Abstract: In this paper, we investigate the convex optimization problem of multi-agent systems (MASs). Un-der the condition that the local cost function is differentiable and strongly convex, a distributed convex optimization algorithm is proposed, which makes the state of the agent asymptotically converge to the global optimal point. In order to save the communication cost and reduce the en-ergy consumption of the system, the integral dynamic event trigger mechanism is introduced in this paper. Different from the traditional periodic detection, the event detection studied in this paper is asynchronous, that is, each agent checks the event trigger conditions based on its own clock cycle. When the event trigger condition is satisfied, the corresponding agent and its neighbors update their control input. In addition, since the minimum time interval of the event trigger time is the detection period, the Zeno behavior can be naturally excluded. Finally, we illustrate the effectiveness of the proposed protocols by a numerical simulation.