Speeding Up GDL-Based Message Passing Algorithms for Large-Scale DCOPs

Abstract

This paper develops a new approach to speed up Generalized Distributive Law (GDL) based message passing algorithms that are used to solve large-scale Distributed Constraint Optimization Problems (DCOPs) in multi-agent systems. In particular, we significantly reduce computation and communication costs in terms of convergence time for algorithms such as Max-Sum, Bounded Max-Sum, Fast Max-Sum, Bounded Fast Max-Sum, BnB Max-Sum, BnB Fast Max-Sum and Generalized Fast Belief Propagation. This is important since it is often observed that the outcome obtained from such algorithms becomes outdated or unusable if the optimization process takes too much time. Specifically, the issue of taking too long to complete the internal operation of a DCOP algorithm is even more severe and commonplace in a system where the algorithm has to deal with a large number of agents, tasks and resources. This, in turn, limits the practical scalability of such algorithms. In other words, an optimization algorithm can be used in larger systems if the completion time can be reduced. However, it is challenging to maintain the solution quality while minimizing the completion time. Considering this trade-off, we propose a generic message passing protocol for GDL-based algorithms that combines clustering with domain pruning, as well as the use of a regression method to determine the appropriate number of clusters for a given scenario. We empirically evaluate the performance of our method in a number of settings and find that it brings down the completion time by around 37-85% (1.6-6.5 times faster) for 100-900 nodes, and by around 47-91% (1.9-11 times faster) for 3000-10 000 nodes compared to the current state-of-the-art.

Collision Avoiding Max-Sum for Mobile Sensor Teams

“…Some of them, including the methods proposed by (Macarthur et al, 2011;Pujol-Gonzalez, Cerquides, Meseguer, Rodriguez-Aguilar, & Tambe, 2013), were implemented in the Max-sum version that was proposed for solving DCOP_MST in (Yedidsion, Zivan, & Farinelli, 2014). Others, which were proposed recently, make an immense reduction of the computational cost in such scenarios (Khan, Tran-Thanh, Ramchurn, & Jennings, 2018;Chen, Jiang, Deng, Chen, & He, 2019). In our work, we prove that such exponential computation is not required, and therefore these methods, which are most useful in standard scenarios, are less relevant.…”

On Population-Based Algorithms for Distributed Constraint Optimization Problems

“…T = [0. 1, 11.1, 22.2, 33.3, 44.4, 55.5, 66.6, 77.7, 88.8, 100] Let the feedback from each point be (using the modified ALS as described Algorithm 3: line 12): 40,30,25,32,42,57,70,95,130] The selected sample points (top G = 3 points) will be (Algorithm 3: lines 15-16):…”

“…As a consequence, diverse classes of incomplete algorithms have been developed to deal with large-scale DCOPs. Until recently, these incomplete algorithms could be classified into three classes: local search-based algorithms [20,21,22,23], inference-based algorithms [24,25,26,27,28] and sample-based algorithms [29,30].…”

GDPx: An Application Independent Pruning Technique to Reduce Computation Cost of Max-Product Belief Propagation Algorithm

“…The Max-Product algorithm has received particular attention amongst all of the existing message passing algorithms. Similar to other such algorithms, Max-Product performs inference on a graphical model by either following a synchronous or an asynchronous message update protocol [6,13]. The messages here are generated using the GDL framework that has an axiomatic tendency of computational savings [12].…”