Research on SFSDP Optimization Based on Gradient Methods

Multi-agent systems can be applied in many fields and have good prospects for development. It is particularly necessary to obtain the location information of a single agent in the system. In this paper, a sparse version of full semidefinite programming(SFSDP) is used to solve the multi-agent system localization problem. The results can be applied in networks of both robotic vehicles and unmanned aerial vehicles with located sensors. In this paper, a model of multi-agent system localization problem was built and converted into a convex optimization problem by SFSDP relaxation. To address the poor performance of SFSDP optimization based on the gradient method in some situations, this paper creatively adopts Adagrad to optimize the SFSDP and verifies the superiority of Adagrad in optimizing the SFSDP through experiments.