This paper introduces a Multi-objective PSO algorithm (P-AMOPSO) based on Two Stages-guided and cross-mutation. There are four improved strategies in this algorithm. First one is to construct external data set based on the strategy of combining strong predominance ranking and crowding distance ranking, which can control congestion of particles effectively. Another one is two Stages-guided strategy, which accelerates the convergence and guarantees the diversity of Pareto optimal set as well. The other one is a cross-mutation operator combining Gaussian distribution mutation and uniform distribution mutation, which overcomes prematurity and reduces side effect of mutation operators. The rest one is to update personal best particle based on the strategy of neighborhood consciousness, which widens self-consciousness range of particles so as to improve its abilities to escape from local optima and conduct local search. Some benchmark functions are tested for comparing the performance of P-AMOPSO with DCMOPSO and MM-MOPSO. The results show the feasibility of P-AMOPSO.