Skip to Main Content
As a successful metaheuristic, ant colony optimization (ACO) performs excellently in solving most combinatorial optimization problems. However, the ACO algorithm needs considerable computational time and resources when the complexity of the problem increases. Parallel implementing is a good ideal to speedup it. A new parallel ant colony optimization (PACO) algorithm is presented, which has the characteristics of coarse-granularity interacting multiant colonies, partially asynchronous parallel implementation and a new information exchange strategy. The code is written in C and MPI and the main application has been executed on the dawn 4000 L parallel computer. We evaluate the PACO algorithm proposed in this paper by study the convergence speed, parallel size scalability and problem size scalability of it. The numerical results indicate that: (1) the PACO algorithm can construct solution better than the sequential ACO (SACO) algorithm and converge faster then SACO; (2) more computational nodes can reduce the computational time obviously and obtain significant speedup; (3) the PACO algorithm is more efficient for the large scale traveling salesman problem with fine quality of solution.