Toward a symbolic implementation of Active Inference for Lego robots

I program autonomous Lego robots to explore concepts of cognition and make these concepts tangible. My goal is for robots to conform to the principles of Active Inference and learn as much as possible on their own. As a software developer and not a neuroscientist, I take an unorthodox approach. Instead of encoding the mathematics of the Free Energy Principle, I implement models of cognition as collections of concurrent processes that carry out "symbolic" computations and communicate via events, something akin to Marvin Minsky’s "societies of mind".  This paper gives an account of my efforts to date.