Distributed Time-Varying Optimization of Networked Heterogeneous Systems

This paper addresses the fully distributed time-varying optimization (DTVO) problem for networked heterogeneous systems, where each agent only has access to its own local time-varying cost function. Most existing distributed optimization studies focus on time-invariant objective functions and homogeneous agent dynamics, which cannot meet the requirements of dynamic optimization tasks in practical scenarios with diverse agent characteristics. To tackle this challenge, a fully DTVO algorithm is proposed by integrating output regulation techniques and adaptive parameter methods. This algorithm enables the outputs of networked heterogeneous systems to asymptotically converge to the optimal trajectory without relying on any global network information (such as topology structure or node quantity). Compared with existing DTVO studies, the proposed algorithm makes two key improvements: first, it expands the scope of agent dynamics from simple homogeneous models to more general heterogeneous dynamical equations, enhancing practical applicability; second, the introduction of adaptive dynamic gains eliminates the need for pre-known network global information, realizing a truly fully distributed control framework. Finally, simulation experiments on a multi-UAV system with heterogeneous dynamics verify that the algorithm can effectively solve time-varying optimization tasks, demonstrating its theoretical validity and engineering practicality.