Quadratic programs arise in robotics, communications, smart grids, and many other applications. As these problems grow in size, finding solutions becomes much more computationally demanding, and new algorithms are needed to efficiently solve them. Targeting large-scale problems, we develop a multi-agent quadratic programming framework in which each agent updates only a small number of the total decision variables in a problem. Agents communicate their updated values to each other, though we do not impose any restrictions on the timing with which they do so, nor on the delays in these transmissions. Furthermore, we allow weak parametric coupling among agents, in the sense that they are free to independently choose their stepsizes, subject to mild restrictions. We show that these stepsize restrictions depend upon a problem's condition number. We further provide the means for agents to independently regularize the problem they solve, thereby improving condition numbers and, as we will show, convergence properties, while preserving agents' independence in selecting parameters. Simulation results are provided to demonstrate the success of this framework on a practical quadratic program.
I. INTRODUCTIONConvex optimization problems arise in a diverse array of engineering applications, including signal processing [1],
robotics [2], [3], communications [4], machine learning [5], and many others [6]. In all of these areas, problems can become very large as the number of network members (robots, processors, etc.) becomes large. Accordingly, there has arisen interest in solving large-scale optimization problems. A common feature of large-scale solvers is that they are parallelized or distributed among a collection of agents in some way. As the number of agents grows, it can be difficult or impossible to ensure synchrony among distributed computations and communications, and there has therefore arisen interest in distributed asynchronous optimization algorithms.One line of research considers asynchronous optimization algorithms in which agents' communication topologies vary in time. A representative sample of this work includes [7]- [12], and these algorithms all rely on an underlying averaging-based update law, i.e., different agents update the same decision variables and then repeatedly average their iterates to mitigate disagreements that stem from asynchrony. These approaches (and others in the literature) require some form of graph connectivity over intervals of a finite length. In this paper, we are interested in cases in which delay bounds are outside agents' control, e.g., due to environmental hazards and adversarial jamming