Autonomous systemsDarkStar and EdgeRunner by domain
Built for air, land, sea, and the field.
Every domain fights a different constraint. In the air, it’s weight. On the ground, it’s the link. At sea, it’s silence. In the field, it’s the months between visits. DarkStar and EdgeRunner are built for those environments: compact edge intelligence designed to keep critical capability on the machine when power, space, bandwidth, and connectivity are limited.
01 / Air
In the air, every gram costs flight time.
A drone pays for everything it carries in minutes of flight. Yet the farther it gets from infrastructure, the more intelligence it needs onboard.
- Size, weight, and power. Every gram is flight time, range, or payload.
- Cooling. Limited space and limited thermal headroom.
- The link. Jammed, contested, intermittent, or simply out of range.
- Cost. The airframe has to stay affordable enough to deploy at scale.
DarkStar is designed to keep more critical AI capability on the aircraft, within the constraints of edge hardware. The system can continue operating locally when the network is unavailable or intentionally silent.
A compact edge-compute unit designed to connect to the aircraft’s existing sensors and systems. Configure it for the mission, integrate it with the platform, and take the intelligence with you.
- A sortie that keeps working when the spectrum goes quiet.Critical AI capability stays onboard.
- More intelligence without turning the aircraft into a flying data center.Designed around the power, weight, and thermal limits of the platform.
- Operations beyond the reliable link.The network becomes an asset rather than a requirement.
- A smaller communications burden.Process locally and transmit only what the mission requires.
TARGET APPLICATION. IN DEVELOPMENT AND EVALUATION. INTEGRATION DEPENDS ON THE AIRFRAME, SENSOR, WORKLOAD, AND TASK.
02 / Land
On the ground, the link comes and goes.
A ground robot lives where signals don’t: inside buildings, under canopy, in valleys, underground, and across complex industrial environments. It may have more room than a drone, but it also has more to do — and it cannot stop thinking every time the connection disappears.
- Connectivity. Indoors, underground, under canopy, or simply unreliable.
- Complexity. A platform may operate across many environments and tasks.
- Long shifts. Hours of autonomous operation without constant supervision.
- Dust, heat, and vibration. Hardware has to survive the work.
DarkStar is designed to keep critical AI capability local to the platform, allowing the system to continue operating through degraded or lost connectivity.
A ruggedized edge-compute unit designed to integrate with the robot’s existing cameras, sensors, and autonomy systems. One platform. Different operational roles. Intelligence that stays with the machine.
- More capable autonomous platforms without data-center infrastructure following them around.Keep more of the intelligence onboard, closer to the sensors and the mission.
- Deployment in places conventional cloud-dependent AI struggles to reach.Indoors, underground, under canopy, or anywhere connectivity cannot be guaranteed.
- Operations that continue through a lost link instead of stopping for it.The link is a convenience, not a dependency.
Target application. In development and evaluation. Integration depends on the platform, the sensor, and the task.
03 / Sea
Under water, there is no signal at all.
Radio doesn’t travel through water. Below the surface, a vehicle is alone with whatever intelligence it brought with it. On the surface, connectivity thins as the horizon gets farther away. Anything that depends on a data center eventually hits a wall.
- No radio underwater. Acoustic links are constrained and expensive.
- Past the horizon. Surface connectivity fades with distance, weather, and operating conditions.
- Endurance. Missions can run for days on a fixed energy budget.
- Pressure and salt. Everything has to survive inside the platform.
DarkStar is designed for local operation when remote compute simply isn’t available. Critical AI processing remains with the vehicle so it can continue operating beyond the reach of persistent communications.
A compact edge-compute platform designed for integration into surface and subsea systems. The vehicle carries the intelligence it needs for the mission instead of depending on a continuous connection to shore.
- A vessel that keeps working past the horizon.Critical AI capability stays onboard when shore-side connectivity disappears.
- A subsea platform that can operate without waiting for a data center to answer.Local processing keeps the vehicle useful even where communications are slow, limited, or impossible.
- More useful intelligence within the vehicle’s existing power and compute envelope.Designed to do more without demanding data-center-class hardware at sea.
- Silent operation when the mission requires it.Local processing reduces dependence on continuous transmission.
TARGET APPLICATION. IN DEVELOPMENT AND EVALUATION. INTEGRATION DEPENDS ON THE VESSEL, SENSOR, WORKLOAD, AND TASK.
04 / Field and IoT
In the field, nobody is coming to reboot it.
A sensor on a hillside might sit for months. It may run on a battery or solar panel, communicate over an expensive or intermittent link, and see a technician only a few times a year. The intelligence has to live within those limits.
- Power. A battery, a panel, and whatever the weather allows.
- Bandwidth. Expensive, intermittent, or nonexistent.
- Months unattended. No operator standing nearby.
- The site visit. Every truck roll costs time and money.
DarkStar brings more AI processing onto the device itself, reducing dependence on continuous cloud connectivity and data transmission.
A compact edge-compute unit designed to connect directly to the system’s existing sensors. Local processing lets remote systems do more where the data is created.
- More intelligence at the sensor.It ignores the ordinary and reports the unusual.
- Less dependence on continuous streaming.Send what matters instead of relying on a constant connection to the cloud.
- Long-duration deployments in bandwidth- and power-constrained environments.Designed for systems that have to do more with limited energy, compute, and connectivity. Remote systems that keep operating when the network does not. Critical AI capability remains available through intermittent or lost connections.
Target application. In development and evaluation. Integration depends on the enclosure, the sensor, and the task.
Four walls, one answer
The constraint changes. The system doesn't.
| Domain | The wall | What changes | EdgeRunner form |
|---|---|---|---|
| Air | Weight, power, cooling, contested connectivity | More critical AI capability stays onboard | Compact unit integrated with the airframe and its sensors |
| Land | Intermittent connectivity, complex environments | Local operation continues when the link does not | Ruggedized unit integrated with the platform |
| Sea | No underwater radio, limited surface connectivity | Intelligence travels with the vessel | Integrated compute designed for the operating environment |
| Field | Battery power, constrained bandwidth, months unattended | More processing happens where the data is created | Protected edge-compute unit connected to the sensor |
How an evaluation works
It starts with your hardware, not ours.
An evaluation runs on the platform you actually fly, drive, sail, or install, against its real workload, inside its real limits. Here's what each side brings.
You bring
- The target hardware, and access to it
- A representative workload
- The known limits: memory, power, cooling, link
- Someone who owns integration
We bring
- A build appropriate to the evaluation
- Integration support
- A repeatable test setup
- A technical analysis of the result
We define together
- The current baseline
- What success looks like
- The test conditions
- The next decision
How much memory and storage the system requires while operating.
How quickly it performs the task within the platform’s real power and thermal limits.
How well it performs the target task compared with the current baseline.
How the system performs as connectivity, compute, power, and other deployment constraints change.
The compute platform with DarkStar running onboard. The fastest path to evaluating Morphos technology on an autonomous system.
For teams building their own hardware, DarkStar can be integrated into the target platform through Morphos software and engineering support.
Morphos has not published performance results outside language workloads. Results for perception, sensor, and autonomous-system tasks are established on the actual platform, data, and task. Why we expect it to travel: Green Vectors operates on vector representations, not on one particular kind of source data. Text, imagery, sensor data, and other inputs can all become vectors. That makes the architecture portable across workloads; performance on each platform still has to be measured and validated.
Measured on public and enterprise retrieval workloads and available to evaluation partners through Morphos software.
Embedded-hardware benchmarking is underway, with initial results targeted for publication following validation.
In prototype, with evaluation programs being developed for defense, autonomy, and other edge-compute applications.
Start an evaluation
Tell us where your machines need to think.
Air, ground, surface, subsea, or a sensor on a hillside. Bring the platform, the task, and the limits it has to live within. We'll bring the intelligence.