There is no incremental path from UAV control to certification for personal transport (DAL-A), nor one from current DAL-A level avionics to full autonomy. We bridge that chasm, relying on solid execution with qualified people in a lean and agile organisation.
We need to build a system that is sufficiently deterministic to pass certification, yet can deal with unexpected situations to justify the absence of a human pilot. Ultimately this is decided by regulatory bodies, not by science or engineering, so the main risk is the legal landscape. It is clear though, that some things will move, one way or another.
We have experts from robotics, computer vision, machine learning and avionics, and we have a pilot whose job it is to go through the requirements for human pilots and turn them into system requirements to which we can engineer. A foundation of thorough classical systems engineering complements modern insights of machine learning and computer vision.