Enhanced Gold Mining Optimization with Dual-Layer Information Sharing and Dynamic Control for Multi-Objective Scheduling of Heterogeneous UAV

Imagine a world where a fleet of drones doesn't just fly around aimlessly, but works together like a perfectly choreographed dance troupe. This is the realm of swarm intelligence, a branch of science that studies how groups of simple agents (like bees, ants, or in this case, drones) can solve complex problems by following simple rules and sharing information. Think of it like a school of fish turning in unison without a single leader shouting orders; they just react to their neighbors. In the real world, we want these drone swarms to handle tricky jobs like delivering packages, searching for survivors after a disaster, or monitoring forests. But here's the catch: not all drones are built the same. Some are fast but short-ranged (like a sprinter), while others are slow but can carry heavy loads (like a marathon runner). When you mix these different types of drones together, figuring out who does what, and in what order, becomes a massive mathematical puzzle. You have to balance three things at once: finishing the job as fast as possible, using as little battery power as you can, and making sure no single drone is working itself to death while others sit idle.

Enter the Gold Mining Optimization Algorithm (GMO). You can think of this algorithm as a digital simulation of a group of gold miners. In the original version, these miners wander around a digital landscape looking for the "gold" (the best solution). They share what they find, move toward the richest spots, and dig together. It's a clever idea, but when researchers tried to use it to schedule these mixed-up drone teams, it stumbled. The miners were too scattered, they got stuck in local "pockets" of gold, and they didn't know how to handle the fact that there wasn't just one "best" answer, but a whole range of good trade-offs.

This paper introduces a new, upgraded version called IGMO-DP (Improved Gold Mining Optimization with Dual-Layer Information Sharing and Dynamic Control). The authors didn't just tweak the miners; they gave them a complete makeover. First, they added a two-layer communication system. Imagine the miners now have two ways to talk: a "local gossip" network where they chat with the people standing right next to them to refine their digging spots, and a "global megaphone" that broadcasts the location of the best gold found so far to the whole group. This keeps the team diverse (so they don't all dig the same hole) while still moving quickly toward the best solutions. Second, they made the miners' steps dynamic. Instead of taking the same-sized step every time, the miners now take giant leaps when they are far from the gold to explore new areas, and tiny, careful steps when they are close to the prize to polish the result. Finally, they introduced a smart filter that changes over time. At the start, the filter is loose, allowing many different ideas to survive to keep the group creative. As time goes on, the filter gets stricter, weeding out the weak ideas and focusing only on the absolute best ones.

The researchers tested this new system in three different "battlefields" of increasing difficulty: a small group of 5 drones doing 10 tasks, a medium group of 10 drones doing 20 tasks with tricky no-fly zones, and a large, chaotic group of 20 drones of four different types handling 40 tasks. They compared their new algorithm against the original Gold Mining method and other famous competitors like NSGA-II and MOPSO. The results were clear: IGMO-DP won. In the most complex scenario, it found solutions that were 14.2% to 17.6% closer to the perfect balance than the next best algorithm, and it covered a wider area of good solutions by about 5%. It also managed to keep the drones' workloads much more even, preventing any single drone from burning out.

The paper explicitly rules out the idea that the original Gold Mining algorithm is ready for this kind of complex, multi-drone job on its own. It argues that without these specific upgrades—especially the two-way communication and the changing step sizes—the algorithm gets stuck and fails to find the best trade-offs. The authors are very confident in their findings based on these computer simulations, showing that their new method is more stable and consistent than the others. They suggest that while this is a huge step forward, the real test will be putting these algorithms on actual physical drones in the real world, where wind and broken sensors might throw a wrench in the works. For now, though, this new "super-miner" algorithm proves that with the right mix of local chat and global vision, a swarm of different drones can work together better than ever before.