To solve the problem of poor accuracy of robot localization and mapping caused by sample degradation and sample depletion caused by the traditional particle filter algorithm while re-sampling, a particle filter method is proposed based on an improved firefly algorithm. The algorithm introduces the observation and state information of the latest moment of the particle into the brightness formula, and the firefly's survival of the fittest and position update mechanism are integrated into the particle filter algorithm, which improves the filtering ability of the particle. To ensure the convergence speed and prediction accuracy of the algorithm, the adaptive adjustment step is introduced in the firefly position update process for immediate correction. Based on the defect of standard particle filter resampling, step-by-step resampling strategy is adopted, an efficient scheme for abandoning small and retain large values is established by the deviation correction exponentially weighted average algorithm. The remaining large weight particles are used to complete the copying and adding of the particle. The simulation results show that the proposed algorithm improves the prediction accuracy of the traditional particle filter, and it can maintain high positioning accuracy and stability when applied to the localization and mapping based on mobile robot motion models.