Surrogate-assisted evolutionary algorithms have been widely employed for solving expensive optimization problems. To address high-dimensional expensive optimization problems, we propose an evolutionary sampling-assisted particle swarm optimization method, termed ESPSO. ESPSO consists of some evolutionary sampling-assisted strategies. It first improves the initialized population with some elite samples by evolutionary sampling. Secondly, during the optimization process, the method builds a local radial basis function model using the personal historical optimal data of the population to approximate the objective function landscape. Finally, surrogate-assisted local search and surrogate-assisted trust region search are designed to find promising candidate solutions for replacing individuals in the population to accelerate the search process. Behavioral research experiments of ESPSO verified these strategies have led to improvements in the search efficiency of the algorithm in various aspects, such as initialization, population update, and optimal solution promotion. We compared ESPSO with five state-of-the-art SAEAs using 18 benchmark functions, which show that ESPSO outperforms the other compared SAEAs and get the best average ranking of 2.194.