In millimeter wave communication systems, channel estimation consumes many resources, in terms of time, bandwidth, and hardware. However, the good news is that the millimeter wave channels are generally sparse in space and channel states at consecutive time slots are highly correlated in angle domain. Therefore, it is possible to track the channel state with less resources, instead of re-estimating it for every time slot. For millimeter wave communications, the antenna array form is important, and we employ a lens antenna array in this paper, because it has much lower complexity and it allows the base station to use only a few RF (Radio Frequency) chains for beamforming. Based on the lens antenna array and the spatial sparsity of the millimeter wave channels, we propose a data-aided channel tracking scheme for vehicle-to-infrastructure (V2I) communication systems, where high-mobility of mobile users is expected. The basic idea is to employ the channel estimate from the previous time slot for the data detection at the current one. After data detection, the detected data sequence can be used for channel update, and this process iterates. By doing this, the overhead of channel estimation can be reduced to great extent. We evaluate the performance of the proposed algorithm through theoretical analysis and simulations, and they both show promising results.