As a part of the multi-source cooperative navigation scheme, data fusion has significant impact on the quality of state estimation. Because of its unique theoretical advantages in the nonlinear non-Gaussian system, the particle filter has gradually become the focus of many fusion methods. However, particle degradation and sample depletion restrict the application of particle filter in complex engineering. In this article, a robust cubature fission particle filter is proposed to solve the above problems. Firstly, in the framework of cubature rule, the Huber function is used to combine L2 norm with L1 norm to improve the importance density function, suppress the observation noise, and further optimize the proposed distribution by integrating Gaussian distribution with Laplace distribution. In this way, the particle degradation is alleviated. The particle swarm is fission derived before resampling, and the sample depletion is suppressed by fission of high weight particles and covering low weight particles to reconstruct particle weights. The vehicle experiment of multi-source cooperative navigation shows that under the same conditions, compared with extended Kalman filter, cubature particle filter and strong tracking particle filter, the root mean squared of the proposed algorithm is improved by 23. 04% , 42. 62% and 37. 74% , respectively. It provides a new idea for alleviating particle degradation and multi-source cooperative localization.