This study presents an acoustic virtual reality technique based on geometrical acoustics simulation and high-order Ambisonics to improve the performance of sound source localization and the realism in enclosed 3D virtual spaces. This tool is based on pre-calculated B-Format impulse responses at a grid of receiver positions that is decoded into a virtual array of loudspeakers fixed at the listener’s head. The virtual loudspeaker signals are synthesized for a binaural representation via head-related transfer functions. This tool is assessed in terms of sound localization. The 1st and 2nd order Ambisonics are compared by varying different visual and head movement conditions to see if they affect the localization performance. Moreover, this study tests the localization in full 3D challenging reverberation time conditions (0,6s up to 2s). The perceived realism, sound source width, and sound continuity are tested under different grid resolutions. Results show that significant outperformance of the 2nd order technique over the 1st order is found in all head movement and visual conditions. It has also been discovered that when using a coarse grid, the perceptual sound source size is increased and spread out. Moreover, realism and sound continuity are not affected significantly by the resolution of the grid.