SkyLattice

Inspiration

SkyLattice started with the Canadian Army's MINERVA UAS Deconfliction Challenge.

The problem is simple: as more drones enter the same airspace, it becomes harder for one operator to make sure every flight can safely happen together.

A drone may have a perfectly safe mission on its own, but once multiple teams are operating at the same time, conflicts can appear in location, altitude, time, restricted airspace, or radio usage.

And even if every mission is safely planned before takeoff, the real world can change after the operation begins.

We wanted to explore one question:

Can we connect what was planned with what is actually happening in one shared operating picture?

That became SkyLattice.

What it does

SkyLattice is a UAS coordination and deconfliction prototype with two complementary demonstrations: a 3D planning and flight simulation environment, and a live tabletop tracking environment.

Before takeoff, SkyLattice brings multiple drone missions into one shared 3D operating picture. Flights are represented as corridors through both space and time, allowing the system to identify conflicts between planned routes, restricted areas, active time-based constraints, and radio requirements. Mission priorities help determine which flights should yield and how alternatives are ranked.

For example, a route may be safe at one time but conflict with a fictional exercise artillery zone when that zone becomes active. Two drones may also be physically separated while still interfering with each other's radio links.

When a planning conflict is identified, SkyLattice can propose changes to departure time, altitude, radio channel, or route for human approval. It also checks terrain clearance, wind conditions, and aircraft constraints.

But planning is only half of the problem.

Once a simulated mission begins, SkyLattice monitors whether aircraft remain inside their approved corridors and predicts potential encounters. The simulator can automatically apply tactical responses, with the decisions and simulated actions shown to the operator.

To demonstrate live observation physically, we built a separate tabletop tactical MVP using camera-tracked vehicles as stand-ins for UAS. OpenCV detects ArUco markers, estimates vehicle motion, and feeds collision predictions and guidance such as SLOW, STOP, and HOLD into the tabletop interface.

The vehicles themselves are not the product — they allow us to reproduce the physical-to-digital coordination problem without flying multiple drones inside a hackathon venue. The tabletop demonstration operates in two dimensions and shares collision-prediction logic with the flight simulator.

Together, the two demonstrations explore two simple questions:

Before takeoff: Can all of these drones safely operate together?

After takeoff: Is that still true right now?

Why it is dual-use

Although SkyLattice began with a defence challenge, the coordination problem is not unique to defence.

During a wildfire, mapping drones, emergency-response drones, helicopters, and water bombers may all need to operate in or around the same airspace.

The same underlying problem appears in search and rescue, public safety, infrastructure inspection, and increasingly complex commercial drone operations.

The rules and missions may change, but the core need remains the same:

many aircraft, one shared environment, and conditions that can change in seconds.

SkyLattice is designed around that common coordination layer rather than around a single type of mission.

How we built it

We built SkyLattice as an offline-capable command-centre prototype.

The backend uses Python, FastAPI, and Uvicorn, with SQLite storing flight-plan revisions and audit records. A JavaScript, HTML, and CSS frontend uses CesiumJS for the 3D map, vis-timeline for scheduling, and milsymbol for military symbols. WebSockets deliver live state updates, and pywebview provides the desktop window.

The strategic demonstration uses a 3D map of CFB Gagetown with locally stored terrain and satellite imagery, published airspace data, OpenStreetMap-derived features, and fictional exercise constraints. It visualizes planned UAS corridors, time-based restrictions, airspace conflicts, radio conflicts, and human-approved planning resolutions.

NumPy, Shapely, and pyproj support numerical and geospatial calculations. We implemented time-dependent A* and safe-interval path planning, and used Google OR-Tools for joint fleet scheduling, altitude, and radio-channel choices. Radio checks use a simplified interference model, with itmlogic providing a Longley-Rice terrain-based propagation option.

For the tactical MVP, we created a physical tracking environment where small vehicles represent moving UAS and a camera acts as the live sensing layer. OpenCV ArUco detection and Kalman filtering turn marker observations into motion estimates, predicted encounters, and driver guidance.

The repository also includes a pymavlink adapter for mission upload and telemetry ingestion, with testing documented against the bundled MAVLink emulator.

The goal was not to simulate every capability of a production aviation system in 24 hours. Instead, we focused on proving the core workflow:

Plan → Deconflict → Operate → Observe → Respond

Challenges we ran into

One of our biggest challenges was realizing that a technically powerful interface is not automatically an understandable interface.

Our early demonstration contained too much information at once. A judge could be looking at a 3D map, reading a flight form, listening to an explanation, and trying to understand a conflict simultaneously.

We redesigned the demonstration around a much simpler principle:

the screen should visually prove the sentence being spoken at that moment.

Another challenge was demonstrating live UAS coordination without access to a fleet of aircraft that we could safely operate inside the venue.

That led to the physical tabletop MVP. Instead of trying to reproduce aircraft dynamics, we reproduced the information problem: multiple moving physical assets being observed and represented digitally.

Accomplishments that we're proud of

We are proud that SkyLattice became more than a static dashboard.

Within the hackathon, we were able to demonstrate complementary workflows covering:

a 3D operating environment built from real geographic data
space-and-time based flight corridors
restricted and time-sensitive airspace
planned flight conflict detection
modelled radio-resource conflicts
mission priority and human approval of planning changes
live camera-based physical asset tracking
monitoring and responses to deviations and predicted encounters

Most importantly, we were able to turn a very complex defence and aviation problem into something that can be explained simply:

help lots of drones share the same sky safely.

What we learned

We learned that UAS coordination is much broader than collision avoidance.

A safe operation depends on understanding not only where an aircraft is, but where it intends to go, when it will be there, what restrictions apply, what communications resources it needs, and what happens when reality no longer matches that plan.

We also learned that the human operator should remain central to consequential decisions.

SkyLattice requires human approval for strategic planning changes, while its simulator demonstrates automatic tactical responses and its tabletop interface provides guidance to drivers. Making those decisions and actions visible is central to the experience.

What's next for SkyLattice

The hackathon version is a proof of concept. Its approvals and separation parameters serve the demonstration and do not constitute operational flight clearances.

Next, we would like to develop SkyLattice into a modular coordination layer with validated real-world telemetry and aviation-data integrations. An initial MAVLink adapter already exists; the next step is to test and extend that integration beyond the bundled emulator.

Future work includes stronger live-data integrations, scaling the existing multi-aircraft deconfliction methods, authenticated asset identities, more robust handling of missing or stale data, richer radio-spectrum coordination, and integration with existing command-and-control or drone traffic management platforms.

Our longer-term goal is simple:

connect planned airspace with live reality so that operators can coordinate more aircraft without losing the shared picture.