Arming rotors00
Palm‑sized drones. Planet‑scale intelligence.
AERKI sends swarms of small, smart drones over forests, coasts and unstable ground. They see under canopy, sample the air and water, and flag wildfire, pollution and landslide risk early, without waiting on a person or the cloud.
Field crews can't walk every hectare of forest, see beneath a closed canopy or stay on a reef around the clock. So warning signs are caught late. Fuel load and moisture loss build up under the canopy until smoke shows. Ground shifts and coastal pollution go unseen until they spread. The drones meant to extend human reach fly uniform fleets on basic grid searches, stream raw data to a distant cloud, and still need people to run every step.
Hidden hazards under canopy, in rough terrain and along coasts sit beyond human eyes and reach, so they are found only after they become visible.
One swarm mixes radar, spectral and physical-sampling drones. Each runs its own models on board, and they hand tasks to each other over the mesh with no operator in the loop.
Most AI lives in data centres and reads the world through someone else's photos. AERKI's AI flies. It senses the forest and the sea first-hand, decides on board, and acts on what it finds, all in the same few seconds.
Radar, thermal, spectral and chemical sensors give the swarm direct readings of canopy, ground, water and air, not second-hand images.
Models run on each drone and on the Jetson-equipped compute drones, so the swarm can judge what matters without waiting on the cloud.
It changes course, drops low, lowers a sampler, plants a geophone or calls in a teammate, turning a signal into evidence on the spot.
AERKI fields two swarm types built from the same drone family. Each mixes the sensors its environment needs, so one deployment covers mapping, prediction and on-the-spot threat analysis.
Radar and spectral drones map beneath thick canopy to find dry fuel, moisture loss and smouldering roots before smoke appears. Gas samplers then confirm where a fire would run.
Radar drones keep flying through fog and storms while spectral and sampling drones track water quality, shoreline change and shifts in the coastal shelf.
Most fleets fly identical drones. An AERKI swarm combines specialised variants that split sensing, sampling and tracking between them in real time, and overlap their data into one diagnostic picture.
See through obstructions with no moving parts, and build fast structural maps.
Watch thermal signatures and surface chemistry across wide areas.
Touch down or fly low to test soil, water, ground and air directly.
The swarm's on-board brain: fuses every drone's data and plans the next move.
Tap the map to place an anomaly. The right drone detects it, then cues the specialists it needs over the mesh. Remove a drone to watch the rest re-route.
The value is in the hand-offs. When one drone spots an anomaly, the swarm re-routes the others to collect exactly the data needed to confirm it.
| Payload drone | Primary data output | Core environmental metric | Role in the swarm |
|---|---|---|---|
| PAPhased array | 3D structural point clouds | Undergrowth biomass density | Flags structural anomalies and cues spectral scans. |
| CPCompute (Jetson Orin Nano Super) | Fused detections and task plans | Anomaly confidence, swarm-wide | Fuses every payload's data on board and assigns the next task in real time. |
| SESeismic sampler | Acoustic waveforms (Hz) | Micro-seismic shift velocity | Maps high-risk zones for landslides. |
| AQWind / AQI gas | Gas profiles (ppm / ppb) | PM2.5 / PM10 particulates | Works out fire spread direction from wind vectors. |
| XRX-band radar | High-resolution topographic maps | Ground surface displacement | Keeps navigation and tracking going in zero-visibility storms. |
| EOEO/IR multispectral | Radiant heat maps | Relative thermal signatures | Triggers immediate low-altitude gas and physical sampling. |
| HSHyperspectral | Chemical spectral signatures | Foliage stress and moisture | Predicts wide disaster zones by finding dry fuel grids. |
| SWSoil / water sampler | Chemical composition | pH, heavy metals, fuel leakage | Confirms suspected ground-level pollution clusters. |
AERKI avoids the delay of remote cloud processing. The swarm thinks for itself, and keeps working when drones are lost.
Each drone runs light machine-learning models on its own edge hardware, instead of streaming gigabytes of raw sensor data back to a base station. Dedicated compute drones carry an NVIDIA Jetson Orin Nano Super for the heavier fusion and planning work.
The swarm trades compressed metadata peer to peer. If an EO/IR drone sees a sudden thermal spike, it asks a nearby Wind/AQI drone to fly over and take a gas profile, with no operator involved.
Processing and routing are fully decentralized. If drones are lost to dense canopy or harsh coastal weather, the rest close the gaps and keep full coverage of the area.
Each drone carries three vertical 8×8 patch arrays set 120° apart, so together they see all the way round. A four-lens EO stack sits flush in the nose: a thermal microbolometer, an AR-coated zoom lens, and two multispectral bands (red-edge and near-infrared). The whole drone, props included, fits inside 100 × 100 mm (96 × 91 mm) and runs on a 2S 18650 pack mounted under the frame.
Drag the slider. Each antenna element fires with a small timing offset, and the combined wavefront tilts toward the target. Retargeting happens in microseconds, entirely in electronics.
Simulation for illustration. Element spacing and count are chosen to show the principle, not AERKI hardware values.
Four capabilities that come from putting phased array sensors on autonomous swarms.
The arrays see through thick forest canopy electronically, mapping dense undergrowth to find hidden fire hazards, fuel load build-up and moisture deficits before visible smoke appears.
Phase shifts across the antenna elements point the sensor's focus in microseconds. No rotating parts means better power efficiency and a tougher drone.
Inside the swarm, sensors cross-reference data in real time. One drone can sweep wide terrain angles while tracking several geological or coastal anomalies, with no tracking blind spots.
Spatial filtering in the array isolates weak target returns from heavy background noise such as moving leaves, heavy rain or crashing waves, for high-accuracy disaster forecasting.
The same drone family, re-tasked for each terrain. Pick a sector to see what AERKI watches for there.
01 / 07
Wildfire starts under the canopy.
02 / 07
Catch crop stress before it shows.
03 / 07
Watch every asset, onshore and off.
04 / 07
Know what moved before it fails.
05 / 07
Coasts, reefs and water, mapped live.
06 / 07
See the ground shift before it slides.
07 / 07 · New
The hyperspectral advantage for defense and intelligence
Seeing the unseen through material-level intelligence.
The pieces needed to put radar on a micro drone have only recently come together.
Phased array front ends now fit on small, flat boards instead of large rotating assemblies.
Small, low-power modules such as NVIDIA's Jetson Orin Nano Super can now run useful machine-learning models on board, so each drone makes decisions without a link to the cloud.
Mesh RF and EO links let a swarm share data in real time without a central node.
Hyperspectral and multispectral cameras now fit on a micro drone, so a swarm can read plant stress, water quality and soil from ultraviolet to infrared, far beyond what the eye can see.
Tell us the ground you need to watch and what you are trying to catch early. We will walk you through how AERKI's sensing stack would cover it.
