Distributed Feedback-Feedforward Algorithms for Time-Varying Resource Allocation

Abstract: In this paper, we address distributed Time-Varying Resource Allocation (TVRA) problem, where the local cost functions, global equality constraint, and Local Feasibility Constraints (LFCs) vary with time. To track the optimal trajectories, algorithms that mimic the structure of feedback-feedforward control systems are proposed. We begin with their conceptual design in the absence of LFCs, developing a feedback-feedforward algorithm that is fixed-time convergent. For cases with LFCs, existing approaches predominantly rely on constructing a time-dependent barrier function, which may impede the design of fixed-time convergent algorithms. Therefore, by exploring the connection between projection and penalty functions, switched feedforward laws are tailored to handle LFCs, with projection used in conjunction. Based on this, we develop a projection-based feedback-feedforward algorithm, which converges to the exact optimal trajectories, possibly along with a number of switching instants, while exhibiting fixed-time convergence between consecutive switching instants. Numerical experiments verify the effectiveness of the proposed algorithms.