<p>In this paper, aiming at further strengthening the capability of particle swarm optimization (PSO) algorithm to deal with the search dilemmas, that is, premature convergence and stagnation, a novel variant of PSO is proposed by devising an individual-based adaptive learning strategy. In this new strategy, to ensure the search efficiency of particle and ameliorate the robustness and adaptivity of algorithm for various search environments and stages well, the search status of every particle is first measured by its update situation, and two special pools of velocity updating formulas are accordingly designed for different search requirements characterized by fitness and/or diversity. Also, Q-learning technique is then employed to dynamically choose the suitable formula among the corresponding pool for each particle to find its new position. Therefore, this strategy can adaptively assign the proper search formula for different particles. Moreover, in order to maintain the vitality of population during the whole search process, a population restart mechanism based on random opposition learning is further introduced, where the stagnant particles are greedily replaced with the randomly ones generated by the personal historical best positions. Resultantly, the proposed algorithm can effectively improve the performance of the algorithm. Finally, a series of ablation and comparison experiments are conducted on the IEEE CEC2017 benchmark functions, and the numerical results demonstrate the effectiveness and superiority of the proposed strategy and algorithm.</p>