To overcome the problem that the algorithm is not stability because the observation dimensions increases in gravity gradient aided positioning, the system posterior probability density narrows and the particle degradation of the conventional particle filter intensifies, an improved particle filter based on artificial physics is introduced to optimize the particle distribution. The improved method incorporates Artificial Physics Optimization into resampling process of the generic particle filter to overcome the problem of particle degradation and sample impoverishment. The improved particle filter is applied into gravity gradient aided positioning by combining the sea experiment data of an inertial navigation system. The artificial physical optimization enables to improve particle degradation and optimize algorithm stability. After optimizing, the algorithm is adopted in gravity aided positioning, which has better estimation precision.