Numerical simulation of airflow in an indoor environment has been carried out for forced, natural, and mixed convection modes, respectively, by using the computational fluid dynamics (CFD) approach of solving the Reynolds-averaged Navier−Stokes equations. Three empty model rooms in two-dimensional configuration were studied first; focusing on the effects of grid refinement, mesh topology, and turbulence model. It was found that structured mesh results were in better agreement with available experimental measurements for all three convection scenarios, while the renormalized group (RNG) к − e turbulence model produced better results for both forced and mixed convections and the shear stress transport (SST) turbulence model for the natural convection prediction. Further studies of air velocity and temperature distributions in a three-dimensional cubic model room with and without an obstacle have shown reasonably good agreement with available test data at the measuring points. CFD results exhibited some unsteady flow phenomena that have not yet been observed or reported in previous experimental studies for the same problem. After analyzing the time history of velocity and temperature data using fast Fourier transformation (FFT), it was found that both air velocity and temperature field oscillated at low frequencies up to 0.4 Hz and the most significant velocity oscillations occurred at a vertical height of an ankle level (0.1 m) from the floor, where temperature oscillation was insignificant. The reasons for this flow unsteadiness were possibly a higher Grashof number, estimated at 0.5 × 106 based inflow conditions, and thus strong buoyancy driven effects caused the oscillations in the flow field. The appearance of an obstacle in the room induced flow separation at its sharp edges and this would further enhance the oscillations due to the unsteady nature of detached shear-layer flow.