Paper story
When the air traffic controller becomes a shepherd
Package-delivery drones are multiplying faster than the people who could ever direct them one by one. Our Springer chapter proposes a different control room: the controller moves a single logical shepherd, and the whole fleet responds. Here is the story, without the mathematics.
The problem
Too many aircraft, one pair of hands
Air traffic control was built around a simple loop: select an aircraft, issue an instruction, wait for confirmation, repeat. That loop works when the sky holds dozens of large, crewed aircraft. It cannot survive the sky that is coming, where delivery fleets, inspection drones, and autonomous platforms put hundreds of small aircraft into the same airspace at once. No human can run the select-and-command loop hundreds of times over, and hiring a controller per drone is not a plan.
Yet removing the human is not a plan either. When risk emerges in a monitored sector, someone accountable must be able to act on a large number of aircraft at once. The question is how one controller can meaningfully steer many.
The idea
One shepherd for the whole sky
Shepherding already solves a version of this problem: one sheepdog steers a hundred sheep by influence rather than instruction. So we gave the air traffic controller a sheepdog. The controller moves a logical shepherd, a purely digital agent in the airspace picture, and its influence rotates the intended flight plan of the nearest aircraft. That aircraft then passes the influence on, aircraft to aircraft, until the whole affected fleet has adjusted. One gesture from one human updates everyone, with no per-aircraft commands and almost no communication overhead.
The translation from paddock to airspace is not one-to-one, and that is the point. A sheep needs five instincts; an aircraft already owns most of them. Attraction to the flock becomes the flight plan. Heading is managed by the autopilot. Collision avoidance belongs to onboard detect-and-resolve systems. The random jitter that keeps sheep from deadlocking has no place around aircraft and is removed. What remains is the one force worth borrowing: repulsion from the shepherd.
Top: the five shepherding forces, translated. Only repulsion from the shepherd is kept; the aircraft’s own systems absorb the rest. Bottom: the controller moves the logical shepherd, and the influence propagates aircraft to aircraft, rotating each intended flight plan.
The design choices
Steer plans, not positions
Two decisions make this safe enough to contemplate. First, the shepherd never pushes an aircraft’s position around, the way a dog displaces sheep; it rotates the aircraft’s intended flight plan, once, and the autopilot flies the amended plan. Continuous force on positions would mean continuous, jittery corrections in a safety-critical domain. Second, the propagation is asynchronous and previewable: the controller can see the predicted effect of a shepherd move on every affected flight plan before committing it, so a gesture never launches a cascade the human did not intend.
What we demonstrated
A new control interface, in a real simulator
We built the approach into ATOMS, an air traffic operations simulator used in ATM research, alongside its original one-command-one-aircraft interface. The new interface lets the controller display the shepherd, drag it, preview the influence on the displayed flight plans, and accept the change, at which point the modified asynchronous shepherding algorithm updates the affected fleet. The contrast with the original interface is the argument in miniature: the same human, given a shepherd, controls a crowd instead of a queue.
Who should care
The people designing tomorrow's airspace
Uncrewed traffic management is being designed right now, and most proposals push the human toward strategic oversight only, because tactical control of hundreds of aircraft seems impossible. This chapter argues it is not impossible; it just needs the right lever. The same pattern extends beyond airspace to any setting where one accountable human must steer many autonomous agents. It also connects to the rest of my shepherding work: the limits study maps where reactive influence works, and the patrolling study turns the same lever to defence.
Cite & explore
The formal version
H. El-Fiqi, K. Kasmarik and H. A. Abbass, “Logical Shepherd Assisting Air Traffic Controllers for Swarm UAV Traffic Control Systems,” in Shepherding UxVs for Human-Swarm Teaming: An Artificial Intelligence Approach to Unmanned X Vehicles, H. A. Abbass and R. A. Hunjet, Eds. Springer, 2021, pp. 245–263. doi:10.1007/978-3-030-60898-9_11
Where reactive shepherding breaks: the companion story →
Back to the paper gallery →
How this page was written. The research, the results, and the ideas here are mine and my co-authors’. To retell them in plain language, I worked with an AI writing assistant that helped draft the text and render the diagrams in this site’s style. I reviewed and edited everything, and the technical responsibility rests with me. If the prose reads a little differently from my papers, that is why.