Blog
Stay informed with the latest news, updates, and industry insights from Astral.
30 Minutes, 4.6 Kilobytes, Zero Collisions
We trained a ground rover to navigate cluttered environments, tight gaps, and multi-room mazes using only a 360° lidar and a learned GRU reflex — no map, no planner, no demonstrations. 120 trials, zero collisions. A VFH analytic baseline fails completely on dense fields.
Four Models, One Stack: Training the Full Perception–Reasoning–Action Pipeline for Autonomous Drones
After the domain detector, we trained three more models in a single session: a VLM action-LoRA that cuts malformed commands, a 121 KB reactive policy MLP that runs at 200 Hz, and a monocular depth fine-tune for rangefinding beyond stereo baseline. All four are now running on Jetson Orin Nano hardware.
We Trained a Domain Detector for Drones. One Class Collapsed to Zero.
COCO-80 has no drone class, no person_aerial, no landing pad. We trained a 9-class domain detector on 48,000 images of sim and real aerial footage — and learned why class imbalance is the dominant failure mode in aerial perception.
Domain-Specific Object Detection for Aerial Autonomy: Sim Data, VisDrone, and the Class Imbalance Problem
Technical report. YOLOv8n fine-tuned on a 9-class aerial schema across three training rounds: sim-only (v1), merged with VisDrone (v2), and 4× drone oversampling (v3). mAP50 0.471 → 0.376 → 0.384. Drone AP50 0.047 → 0.010 → 0.087.
Tower vs. Self-Organized Droneport ATC: What a 9-Cell Factorial Study Found
We ran 405 simulations across nine coordination architectures. Self-org with ADS-B matches tower throughput below 20 ops/hour — then falls apart. Silent-cruise drones exceed safe separation thresholds at just 12 ops/hour.
Why Every Drone AI We Tested Lost to Doing Nothing (And What Fixed It)
We ran 10,200 closed-loop flight trials across 25 vision-language models. Every single one lost to a drone that just hovered. Here's the metric gap — and the architecture that finally closed it.
18 Iterations to Beat Hover: What We Learned Engineering Drone Autonomy
We trained on 6.7 million frames. Detection improved 9.7×. Closed-loop navigation didn't budge. The story of the domain gap trap — and what actually limits autonomous drones today.
What 1,000 Drones Need to Coordinate: UWB Is Non-Negotiable Above 100
Camera-only swarms drop 15.8 percentage points in coverage and have 8× the collision rate at 1,000 drones. Here's what the sensing infrastructure for large-scale drone fleets actually requires.
We Attacked Our Own Drones 11,340 Times. Here's What We Learned.
GNSS spoofing. RF jamming. PN interceptors. Control takeover. We ran every major counter-UAS attack against our autonomous drone swarm. None of them degraded mission success — and that's the problem.
94% Success Rate: What Happens When You Add a Human to the Loop
Full autonomy gets 57.6% on hard warehouse tasks. Add a human for novel situations, and it jumps to 94.4%. The right architecture isn't fully autonomous — it's autonomy-aware.
How to Make Autonomous Drones Smarter (Without Wishful Thinking)
A practical stack for AI drone autonomy: simulation-first iteration, metric grounding, modular perception and planning, and closed-loop evaluation.
The Metric Gap in Vision-Language Drone Navigation: What Actually Breaks
Why general-purpose VLMs struggle as end-to-end drone controllers, what fails in closed loop, and why separating semantics from geometry is the pragmatic path.
Yonder: A Large-Scale Drone Navigation Dataset and Why Offline mAP Lies to You
What Yonder contains, who it is for, and how cross-simulator evaluation prevents false confidence when training perception for embodied flight.