Multi-resource constrained flexible job shop scheduling problem (MRC-FJSP) is a kind of complex combinatorial optimization problem. A multi-population particle swarm optimization algorithm with random network (MPSO-RDnet) was proposed for solving MRC-FJSP with the objective to minimize makespan. First, a new decoding method which combines semi-active decoding and heuristic rule decoding was designed. The original solution space was cut out effectively. Second, two neighborhood structures based on the critical path were designed to improve the local search ability of the population, and a multi-population strategy based on the random network structure graph was added to improve the global search ability of the algorithm. A reinitialization strategy for the algorithm search stagnation was proposed to enhance the robustness of the algorithm. Numerical experiments verified the effectiveness and efficiency of proposed algorithm.