GPU-based parallel multi-objective particle swarm optimization for large swarms and high dimensional problems

•The proposed parallel implementation of MOPSO using a master-slave model provides up to 157 times speedup compared to the corresponding CPU MOPSO.•This paper presents a new parallelized implementation of the multi-objective particle swarm optimization (GPU MOPSO) for large swarms and high dimensional optimization problems based on a master-slave model. This paper also presents a new serial implementation of MOPSO. The new serial implementation of MOPSO (CPU MOPSO) has faster performance for large swarms and high dimensional optimization problems.•Here, we investigate a large number of iterations to reach good nondominated solutions which achieve good Pareto fronts. Pareto fronts of both CPU MOPSO and GPU MOPSO implementations are very closed to the true Pareto fronts.•Performance of MOPSO is dependent upon an archiving technique. We have proposed a simple parallel archiving technique which significantly speeds up the process. Our serial archiving technique is the same as the parallel archiving. In our parallel implementation of MOPSO, PRNG and coalescing memory access have a positive impact which improves computational time. The single step of the combined taus worthe generator of PRNGs for GPU presents enough quality Pareto fronts.•In most of the serial and parallel implementations, they fixed dimensionality and a swarm at small size because the time complexity of archive handling is proportional to the sizes. If they consider a large number of iterations for best quality Pareto fronts, in the case of large dimensionality and a large swarm the time performance will decrease significantly. We demonstrated that our proposed implementations can handle both of cases with good Pareto fronts which are very closed to the true Pareto fronts. During the last couple of years, parallel MOPSO (Multi-objective Particle Swarm Optimization) with two or more objectives has gained a lot of attention in the literature on GPU computing. A number of implementations have been published for MOPSO on a GPU. However, none of them have been able to capture good enough Pareto fronts fast. In addition, the authors have pointed out their limitations in various aspects such as archive handling, picking up fewer nondominated solutions and so on. Previous literature also lacks evaluation of its MOPSO implementation with large swarms and high dimensional problems. This paper presents a faster implementation of parallel MOPSO on a GPU based on the CUDA architecture. We achieved our fast