Before you even have an algorithm you need to know what the flight environment and positive separation requirements are. A spec is needed. Is this going to be a

Before you even have an algorithm you need to know what the flight environment and positive separation requirements are. A spec is needed. Is this going to be a centralized system (like air traffic control) or a decentralized one (like TCAS) or some hybrid approach? A whole bunch of rules need to be sorted out about obstacle clearance, speeds, overflight of people's homes or roadways, and emergency procedures. What do emergency procedures even look like for an autonomous aircraft? A bunch of failure vectors and likelihoods need to be defined, e.g. since a partial power failure can occur anytime. Where should it set down if it has that much control? If it lacks sufficient control to get to a safe setdown zone, how does it crash land? i.e. better to crash in the middle of the road than on top of a car or person. Progressive failure. The certification and currency process for your algorithm should be more like a pilot because the autonomous UAS software is in effect a substitute pilot. Pilots are certified for life in the U.S. but they have 6, 12, and 24 month (depending on privileges) medical evaluations, and 30 day, 90 day, and 2 year currency requirements for various things. It's not legal for a pilot to fly (at certain times, environments, or at all) if they lack currency requirements, even though they're certified. Similar for UAS. You'd have to have a certain software update that ensures conformance with current version of the public use UAS operating environment behaviors. That way everyone is always on the same page. Autonomous flight in a shared airspace is non-trivial.