Simulation is an indispensable tool in protocol design and evaluation. Typically, simulations of mobile wireless sensor networks rely upon mobility models. This paper presents a smooth Gauss-semi-Markov mobility model (SGM) that can reflect the realistic node movement and has more controllable parameters. Time stationary average speed and uniform spatial node distribution of SGM are proved by Markov process and renewal process theory. Thus, SGM model can be applied to simulation of mobile wireless sensor networks.