Swarm

The open arena where AI learns to fly.
Train a drone pilot, prove it against the world, and watch it fly.

--- ## What Is Swarm Drones are moving into everyday life: delivering packages, inspecting bridges and power lines, searching for survivors after a disaster. The hardware is ready. The hard part is the **AI that flies them**, and today that AI is built behind closed doors, where no one can compare it or prove whose is best. **Swarm is the open arena that settles it.** Anyone can train a drone pilot, submit it, and watch it compete on a public [leaderboard](https://swarm124.com/benchmark) across thousands of fresh worlds it has never seen. No private test sets, no memorizing, no shortcuts. The best pilot wins, out in the open, and its rewards are paid automatically by the [Bittensor](https://bittensor.com) network (Subnet 124).

_{Search and Rescue. The mission Swarm is built around: teaching a drone to find people.}

_{5 missions · 1,100 fresh worlds every 14-day epoch · 60-second flights · one live leaderboard}

(back to top)

--- ## Five Missions Swarm is five separate competitions, each with its own champion and its own share of the rewards. Master one, or take on all five. | Mission | What your drone has to do | Reward share | |---------|---------------------------|:------------:| | **[Office Interceptor](docs/families/office_interceptor.md)** | Hunt a target drone inside a fixed office and catch it before time runs out |

| | **[Swarm Autopilot](docs/families/swarm_autopilot.md)** | Land a whole team of drones, fast and without collisions |

| | **[Swarm Search and Rescue](docs/families/swarm_sar.md)** | Send a team of drones to sweep the area and find the victim together |

| | **[Autopilot](docs/families/autopilot.md)** | Cross the world, find the landing pad, and touch down clean |

| | **[Search and Rescue](docs/families/search_and_rescue.md)** | Find a lost person and hold a steady hover above them |

_{Every mission has a full guide with everything a builder needs to start.}

_{Solved: Interceptor — open-terrain pursuit, cleared by its champion and closed. The crown is final and the winning solution is preserved as open source.}

(back to top)

--- ## Watch: Interceptor

_{Air-to-air pursuit, from a real benchmark run: close the gap and catch a fleeing drone before the clock runs out.}

(back to top)

--- ## How It Works Every mission runs the same simple loop:

Depth Camera _{what's in front of it}	→ Your Model →	Flight Commands _{where to fly next}	→ Drone
Flight State _{position and speed}

_{Third-person view}

_{What the drone sees}

The drone sees the world through a single depth camera and knows its own position and speed. Fifty times a second, your model looks at that and decides where to fly next: a direction, a speed, a turn. No map, no GPS, no list of obstacles. Just like a real pilot, it has to read what is in front of it and react. What changes between the five missions is the goal: land, rescue, coordinate a team, or give chase.

(back to top)

--- ## The Worlds None of these worlds exist until the benchmark builds them. Every 14-day epoch, each family receives **1,100 fresh seeds** across its supported map types, so nothing can be memorized.

City _{streets, buildings, intersections}	Ski Village _{snow-roofed streets, mountains}	Mountains _{peaks and valleys}
Warehouse _{indoor racks and cranes}	Forest _{dense trees, tight gaps}	Open terrain _{wide skies, no cover}

Forest, in four seasons

_Normal

_Autumn

_Snow

_Dead

(back to top)

--- ## Scoring & Rewards The standard score from 0 to 1 uses three things:

Success
_{did it finish the mission?}
45%

Speed
_{how fast, within the time limit?}
45%

Safety
_{how clear of obstacles?}
10%

A drone's rank is its **average across all 1,100 worlds**, so steady skill beats a few lucky runs. (Interceptor and Office Interceptor use pure pursuit: half success, half speed, with no safety term.) Rewards run on **King of the Hill**. Each mission pays its **last five champions**, not just the current one, and your share depends on how much you raised the bar when you won. Beat the record and you keep earning even after someone beats you, so a real breakthrough pays off for a long time. The full mechanics are in the [King of the Hill guide](docs/king_of_the_hill.md).

(back to top)

--- ## Built for Builders Everything happens from the terminal, with no dashboards to wrangle. Train, package, and score your model with a handful of commands: ```bash swarm doctor # check your setup swarm model package --source my_model/ # package your agent swarm benchmark --model Submission/submission.zip --workers 4 # score it locally swarm visualize --type 1 # fly a map yourself to see it swarm report # see how it did ``` The local benchmark uses the same simulator and scoring path as validators, so it is the right place to compare iterations. Leaderboard scores use independently generated epoch seeds, so a local score is not a guarantee of the final network score. Full reference in the [CLI docs](docs/CLI_readme.md).

(back to top)

--- ## Start Building > [!IMPORTANT] > **One shot per drone, per mission.** Once your model is scored, its slot is locked. Practice locally as much as you want; when you are ready, submit your best and it competes for good.

Train a Model

Build a drone pilot from scratch. The Miner Guide walks you through the whole path, from first install to a spot on the leaderboard.

Run a Validator

Help score the network on your own hardware. The Validator Guide covers setup, launch, and auto-updates.

(back to top)

--- ## From Simulation to Reality The pilots trained here don't stay in simulation.

**[Langostino](https://github.com/swarm-subnet/Langostino)** is the open-source drone we built to prove it flies for real: 3D-printed, off-the-shelf parts, full build guide included. Anyone can make one.

Train in simulation → Compete on the leaderboard → Fly on real hardware

(back to top)

--- ## Community

--- ## License Distributed under the MIT License. See [LICENSE](LICENSE) for details. ---

Swarm

Swarm: where AI learns to fly.
_{Subnet 124 on Bittensor}