Expected behaviour in Linkloss scenarios - UAVs

Question

Imagine a remoted piloted aircraft like MQ-9 being controlled by pilot on the ground. I was wondering what is the expected behaviour of the aircraft when the connection with the pilot is lost? Does the RPA keep flying with the old commands defined by the pilot before the connection is interrupted? Does it fly a defined trajectory while being tracked by the ATC?, or do they fly towards a preplanned safe route? (autoland?). If the second solution applies, the RPA would need a fast algorithm to calculate the path (while avoiding obstacles, restricted zones, ...). Any idea how this is done and which algorithm they might implement?

Answer 1

I would say - Any and all of the above, depending on circumstances.

Upon Loss of Signal:

• Hold in place (Circle for XX minutes, to see if signal returns)
• Return To Base (RTB) right now
• Go somewhere else and land
• Do some other mission, then RTB
• Do something else

And also, location dependent. What happens near home base in Nevada would be different than what happens over hostile territory.

There isn't a singular or "expected" default of "what happens". It all comes down to how it is programmed.

Even a little $300 consumer UAV is programmable as to a LOS event.

Answer 2

Every product/project will implement the approach that they think best, given that it is approved by the authorities.

For my current project, we still have to get the permit to fly by EASA, but we're going for an autoland.

The algorithm to compute the path has been developed by me, so it does not really have a name. Moreover, rather than being simply an algorithm, I would say it is a full fledged program, given the amount of possible input and scenarios.

Being a very early demonstrator/proof-of-concept kind of thing, we're not going to have obstacles in our testing area.

Answer 3

This differs how the drone is specifically programmed, but here are the modes I know have been used.

• Continue flight path: Often, the operator pre-specifies a mission with the help of a series of waypoints. If the telemetry-link to the drone is lost but it still has a valid navigation solution (e.g. GPS signal) the drone will simply continue this pre-planned mission. At the end of this flight-path, a holding pattern can often be introduced, on which the drone waits for a "go-ahead" signal (of course only when it has a valid radio signal). The reasoning is to not endanger your airbase or recovery site. If everything goes wrong the drone would crash at the location of that holding pattern.
• Hold pattern on location: In certain flight modes, or when no waypoints were planned, often the only course of action is to enter a holding pattern at the current location. This also is true, when the drone loses all its communication signals including GPS signal. Because in that case the drone doesn't know where it is, nor can it receive any information from the outside, this is the only option (besides crashing). The now infamous Beast of Kandahar was captured by the Iranias with GPS-spoofing, such that the RQ-170 entered a holding pattern until it ran out of fuel and glided to the ground (at least that is the speculation of what happend).
• Dynamic Return to Home: This is the case you mentioned about planning a way home. The drone would either re-track its flight path back to where it launched, or it would dynamically plan a path home. This case is often more trivial then you think, because on the one hand you have plenty of time to compute this flight path, and you can simply high enough up, such that the airspace can be considered obstacle free. In that case you only need some elevation map to ensure that you do not controlled fly into the terrain. The downside to this is that this drone who is trying to return home, does not react to air traffic, for this reason it is often deemed more safe to simply ditch the drone. A lot of research exists in which safe dynamic planning is attempted, but a big problem is always environment perception. A nice paper imo on this topic can be found here (no affiliation). At the heart of a lot of these dynamic planning algorithms is the well-known and relatively simple A* algorithm.