The GPU-based parallel Ant Colony System

Abstract

The Ant Colony System (ACS) is, next to Ant Colony Optimization (ACO) and the MAX-MIN Ant System (MMAS), one of the most efficient metaheuristic algorithms inspired by the behavior of ants. In this article we present three novel parallel versions of the ACS for the graphics processing units (GPUs). To the best of our knowledge, this is the first such work on the ACS which shares many key elements of the ACO and the MMAS, but differences in the process of building solutions and updating the pheromone trails make obtaining an efficient parallel version for the GPUs a difficult task. The proposed parallel versions of the ACS differ mainly in their implementations of the pheromone memory. The first two use the standard pheromone matrix, and the third uses a novel selective pheromone memory. Computational experiments conducted on several Travelling Salesman Problem (TSP) instances of sizes ranging from 198 to 2392 cities showed that the parallel ACS on Nvidia Kepler GK104 GPU (1536 CUDA cores) is able to obtain a speedup up to 24.29x vs the sequential ACS running on a single core of Intel Xeon E5-2670 CPU. The parallel ACS with the selective pheromone memory achieved speedups up to 16.85x, but in most cases the obtained solutions were of significantly better quality than for the sequential ACS.

Implementing a GPU-based parallel MAX–MIN Ant System

“…In an ACO, a population of agents (ants) construct, in parallel, a set of solutions to the optimization problem begin tackled. Unfortunately, the inherent parallel nature of the ACO does not translate easily into an efficient GPU-based parallel implementation [7,12,39]. The difficulties arise partly from the fact that not all of the ACO computations are independent, e.g., pheromone trail updates; as well as from the computing restrictions inflicted by GPU architectures.…”

“…It is possible to use the current global best value instead of the iteration best [43]. It is worth noting, that in contrast to the Ant Colony System (ACS), parallelization is made simpler because the MMAS lacks a local pheromone update [39]. In fact, the pheromone trail values remain constant during the solution construction phase, allowing a beforehand computation of the product of the pheromone trails and the heuristic values required by Eq.…”

An improved Ant Colony System for the Sequential Ordering Problem

“…Because the proposed fusion between the ACS and SA is problem-agnostic one could try to apply it to solve other difficult combinatorial optimization problems. The performance of the proposed algorithms in terms of computation time could also be improved with the help of parallel computations, as the ACS is susceptible to parallelization even with modern GPUs [41]. Acknowledgments: This research was supported in part by PL-Grid Infrastructure.…”

An Approximate Algorithm for the Steiner Tree Problem based on Ant Colony Algorithm

“…If > 0 , we use the formula (2) to select the next node, this method is also called roulette select method. Skinderowicz R [9] based on the characteristics of the thread bundle proposed a reduction algorithm, the maximum size of the calculation is limited to 32. This is because, in the CUDA architecture, the number of threads in a thread bundle is 32.…”