Project releases
ruv-drone gains a predictive advisory for fleet coordination
ruv-drone coordinates cooperative UAV fleets above the per vehicle autopilot: formation keeping, Raft consensus and a hard flight safety boundary. A September 2 commit wires RuForecast in as a predictive advisory.
GitHub activity: · Published:
What ruv-drone is
ruv-drone sits above each vehicle's autopilot and handles the part the autopilot does not: keeping a group of aircraft in formation, agreeing on shared state through Raft consensus, and enforcing a flight safety boundary that coordination logic cannot argue its way past.
It is a Rust workspace, feature gated so you can build only the coordination core or pull in MAVLink, ONNX, the demo, LatentMesh and RuForecast.
What changed
Commit eafd7e5 added a RuForecast predictive advisory behind the ruforecast feature, giving the coordination layer a forecast to reason with rather than only present state. The preceding commit added a LatentMesh explainer and operator guide, and a security commit removed an unused MAVLink dependency in PR #3.
An advisory is advice. The flight safety boundary remains the authority; a forecast does not get to override it.
Get started
A Rust toolchain and a checkout. Start with the coordination core only:
cargo build
cargo test
Expected result: the core coordination layer builds and its tests pass with no radio hardware present.
To include the advisory and mesh features:
cargo build --features full
cargo test --locked --features latentmesh,ruforecast --all-targets
No documented MCP server. The supported surface is the Rust workspace and its feature flags.
Use it today
Practical case: an inspection flight with several aircraft that must hold relative position. Input is vehicle telemetry. Workflow is run the coordination layer in simulation against the loopback example, then stage onto hardware. Output is formation commands plus a consensus view of fleet state.
Acceptance test: run cargo run --locked --features latentmesh --example latentmesh_loopback and confirm the signed message path completes end to end.
Push it further
Experimental commentary. The ambitious direction is the jellyfish work: a companion crate added on August 21 that borrows swarm behaviours from biology instead of prescribing formations.
Limitation: the interactive LatentMesh performance figures come from staged conditions, not a certified flight programme. Falsifiable test: run cargo test -p ruv-jellyfish standalone and confirm it builds and passes outside the main workspace.
Read the original on GitHub commit