MDS - Mission-Directed Swarm
Open-source MAVLink fleet operations for PX4 drones, SITL, drone shows, SAR, and cooperative autonomy.
MDS began as MAVSDK Drone Show and now covers a broader PX4-oriented fleet stack:
drone-side runtime, GCS/backend services, React operator dashboard, SITL, real
companion computers, sidecar connectivity, and mission execution workflows. It
is built to stay simple enough for a quick SITL demo while remaining structured
enough for serious multi-drone validation, operations, and custom deployments.
What MDS Covers
- Offline drone shows with SkyBrush-imported trajectories and synchronized execution
- Smart swarm missions with live leader-follower coordination and runtime control
- QuickScout SAR / recon with multi-drone coverage planning and PX4 Mission Mode execution
- Unified operations tooling with SITL, GCS services, Swarm Trajectory planning, and live/historical logs
MDS is a field-operations and research platform, not certified avionics. Flight
testing requires qualified operators, local aviation compliance, geofencing,
failsafe review, and independent safety validation.
Simurgh maturity boundary
Simurgh Operator is currently a demo / proof-of-feasibility beta, not a
production-ready or safety-certified autonomy product. The completed SITL
checkpoint demonstrates that typed intent, guarded confirmation, monitored
actions, telemetry evidence, and ULog review can work together in a repeatable
operator workflow. It does not establish production readiness, real-aircraft
approval, or commercial/field safety.
Use Simurgh only with the documented human-confirmation and circuit-breaker
controls in a trusted SITL or review environment. Treat telemetry gaps,
ambiguous state, provider behavior, and all remaining deferred items as
operator-visible limitations. Real-aircraft and commercial hardening are
explicitly deferred to a later phase; see the
Simurgh feasibility checkpoint.
Start Here
Quick Demo
Start with the SITL Comprehensive Guide.
It covers the official SITL archive, validation, bootstrap, dashboard launch,
and the current dashboard-first SITL Control workflow.
The stock lab/SITL profile is intentionally plug-and-play: dashboard and
machine API authentication are off on the trusted deployment network, and MDS
warns when a feature uses that open posture. Do not expose that profile to an
untrusted network. Before shared, field, customer, or commercial deployment,
follow the staged hardening path in the
GCS Auth Guide.
For normal demos, use the dashboard SITL Control page to reconcile/start
SITL drones instead of copying low-level shell commands from the README. Shell
and make wrappers still exist for automation and advanced operators; see the
Operator Makefile Guide when you need
scripted control.
The dashboard launcher keeps the FastAPI backend single-process by default, so
telemetry, heartbeats, command tracking, and other in-memory operational state
stay coherent during live SITL runs. Backend auto-reload is only for backend
code editing; the SITL guide covers that advanced mode.
For the stock official SITL package, Mission Config starts from the tracked Azadi Stadium demo origin in data/origin.sitl.default.json. If you later set a different origin in the dashboard or via the API, MDS writes a local runtime override to data/origin.json on that server. Remove that local file when you want the stock Azadi default to apply again.
For the default official HTTPS demo path, the GCS now treats git write-back as disabled by default (MDS_GIT_AUTO_PUSH=false). That keeps imports/config saves clean on read-only evaluation setups; move to the fork/SSH workflow when you want the GCS to commit and push changes.
If you skipped the bootstrap installer and do not have the repo yet, follow the
clone/bootstrap path in the SITL guide instead of copying ad-hoc commands from
the README.
For a first dashboard-driven Smart Swarm run after SITL launch:
- open
Overview and check aircraft identities, live telemetry and attention items
- open
Smart Swarm and review the saved formation in Design Swarm
- use
Operate to take off, or take off with QGC/RC and stabilize
- review
Start Swarm, then confirm the intended cluster
- monitor from
Operate; use Pause/Stop following, Land or Managed Return there when needed
Overview retains general Takeoff, Hold, Land and immediate PX4 Return controls.
Mission setup/start/monitoring belongs to its dedicated workspace. Overview
selection is explicit: first-use All or the remembered selection; filtering and
reconnecting never change its targets. See the
Overview operator guide.
The validated 5-drone Smart Swarm acceptance flow is documented in the Smart
Swarm and SITL guides, including the operator-safe dashboard path and the
automation validation helper for advanced users.
If you need your own fork, custom image, or a pinned redistribution workflow, do not improvise from the demo path. Use:
These advanced and real-hardware paths require stronger PX4/Linux/networking knowledge plus licensing, regulatory, and operational review. If you need private assistance or deployment consulting, contact Alireza on LinkedIn or [email protected].
More Guides
MDS can integrate optional companion-side tools when real hardware needs them:
- Smart Wi-Fi Manager for field Wi-Fi profile management on Linux companion computers
- MAVLink Anywhere for MAVLink routing over companion computers, LTE/Wi-Fi/VPN, UDP, and serial
They remain standalone projects and are not required for a first SITL demo.
Product Highlights
- Single operator surface: React dashboard for monitoring, control, QuickScout, Swarm Trajectory, and log review
- Governed AI operator demo: Simurgh combines live MDS evidence, multilingual semantic planning, human confirmation, durable progress, and circuit-breaker enforcement as an early proof-of-feasibility beta
- Modern SITL workflow: prebuilt PX4 Gazebo SITL image, fast container startup, and reproducible custom-image tooling
- Operational visibility: unified logging across GCS, drones, and frontend error reporting with exportable sessions
- Drone Show pipeline: staged SkyBrush ZIP import, processed trajectory plots, readiness gating, and synchronized launch control
- Scalable architecture: designed for anything from a small demo to large validated multi-container runs
Roadmap Direction
Active and planned work is tracked in ROADMAP.md. Current research
directions include:
- MCP-compatible AI-agent workflows for drone swarms and fleet operations
- automated SAR, surveillance, and cooperative mission workflows
- PixEagle integration for vision-assisted swarm control
- GPS-denied and indoor swarm feasibility using AI/image-processing pipelines such as monocular vision
- future ArduPilot support exploration
- optimized minimal SIH Docker simulation for lightweight validation
Dashboard Scope
The dashboard covers:
- live fleet monitoring and control
- mission upload and execution
- Swarm Trajectory planning and advanced route editing
- QuickScout mission planning and monitoring
- live and historical Log Viewer workflows
Map-heavy features such as Swarm Trajectory, the Advanced Route Editor, and QuickScout use Mapbox when available and Leaflet fallback otherwise. See the relevant guides in Documentation Index.
Videos
- Project history and walkthrough playlist: YouTube playlist
- MDS 5 walkthrough: coming soon, to be linked here when published
Documentation
Licensing
MDS uses dual licensing:
- PolyForm Noncommercial for education, research, and other non-commercial use
- PolyForm Small Business for qualifying small commercial operators
- Commercial license for larger commercial deployments
See LICENSE and LICENSE-COMMERCIAL.md.
Contributors should start feature branches from main. Use short-lived feature or client branches for work-in-progress, then merge back to main only after SITL/hardware validation.
Safety Note
Real-drone deployment requires appropriate flight-test discipline, regulatory compliance, and failsafe validation. Treat SITL success as a prerequisite, not as proof of real-world readiness.
评论