As continuation of a series of one-screen perspective tracking experiments (Portal 2.0, Contra Base 1, Doge Chorus, etc.) I did in late 2014, Shawn Lawson (shawnlawson.com/) and I changed the lab layout into a three-screens box CAVE environment late last month. This video documentation covers multiple new demos I've developed for the new environment in the past couple of weeks.
Each screen is attached perpendicularly to its neighbor(s), aligned with its corresponding rear projection. Four Vicon Bonita cameras on top corners cover the central area surrounded by the screens. The origin point locates at center of the central area.
Data collected by Vicon is sent through VRPN (Virtual-Reality Peripheral Network), then OSC (Open Sound Control) protocol to Processing, and is mapped as frustum parameters to compute three projection matrices, which coincide with the screens' position in reality.