Graph attention network-enhanced multi-agent reinforcement learning for dynamic interception task allocation in counter-drone defense
Dianbo Jia, Ganliang Wang, Hongfei Bu, Junwei Wang, Zhengkai Wang · 2026
Read Graph attention network-enhanced multi-agent reinforcement learning for dynamic interception task allocation in counter-drone defense by Dianbo Jia, Ganliang Wang, Hongfei Bu, Junwei Wang, Zhengkai Wang (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/PMC13547370.txt.
Description
Dianbo Jia and co-authors designed in 2026 a two-level learning system for assigning interceptor drones to incoming hostile drones when targets keep appearing, vanishing, and changing course. At the top level, a graph attention network treats interceptors and targets as a shifting web. It recomputes who should chase whom, using masks that absorb arrivals and departures without rebuilding the web. Edge features carry reachability and urgency straight into the decision.
At the lower level, a multi-agent policy steers the actual flying. Tested from four-against-four up to twelve-against-twelve engagements, the framework beat the strongest rival by about ten percentage points in the surprise-intrusion scenario. It reallocated targets successfully about 87 percent of the time and rarely flip-flopped assignments. Removing the edge features hurt most, marking them as the key ingredient.
Catalog record
| CREATOR | Dianbo Jia, Ganliang Wang, Hongfei Bu, Junwei Wang, Zhengkai Wang |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 117362 |
| LICENSE | https://creativecommons.org/licenses/by/4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |