Directional transmission and reception using multiple-input multiple-output (MIMO) technologies allows spatial multiplexing in a wireless network. However, multipath propagations in radio channels usually result in the angular spread of signal power, which causes interference among the sectors. Moreover, the random movement of the nodes leads to the variation of the interference. An accurate and mathematically tractable packet-level channel model is a key issue to evaluate the capacity and performance of the spatial multiplexing networks. In this paper, we have considered an indoor mobile ad hoc network where the sink node is equipped with an antenna array for beamforming. First, based on the angular power spectral density (APSD) of the indoor channels obtained by our field measurements, the interference among two nodes is evaluated. Second, by assuming two-dimensional random walk of the network nodes, a finite state Markov chain (FSMC) model is proposed to present the random variation of the signal-to-interference ratio (SIR) of the directional links from the nodes to the sink. The closed-form expressions of the state transition probabilities and steady state distributions are derived. Simulation results are also presented to demonstrate the channel evolution process. The proposed packet-level channel model can be conveniently incorporated into analytical frameworks and fast simulations of upper-layer network protocols.