Ultra Compact powerful designed for space
PULSAR Edge computing module
→ Configurable EDAC scheme up to rad-hard level protection to SEEs→ 16 performance levels for most energy-efficient utilisation→ Reliable and consistent with unlimited on-orbit reconfiguration→ Radiation tolerant suitable for LEO and upgradable for MEO, GEO and deep space missions.
Download
Pulsar is a world-class edge computing module offering power efficient ultra-high-speed data processing of both conventional algorithms and artificial intelligence (AI) models.
With 90 TOPS peak AI performance, this device delivers powerful and timely processing in a small form factor and power consumption.
A unique range of 16 computing levels means power is never overspent and computation time is always optimal.
Another key benefit is a configurable Error Detection and Correction (EDAC) scheme providing lower power, rad-hard level (default) integrity checks of critical data such as ML weights, bootup parameters and more. The scheme trades off reliability, power consumption and computational latency withsimplicity through selectable settings.
Custom and third-party libraries can be installed through a software interface and via Docker containers to run a variety of models with leading AI frameworks such as Keras, TensorFlow, Pytorch, Caffe2, Darknet (YOLO) and more.
platforms
product applications
PULSAR performance
showcase
Real-time ship localisation
.jpg)
Processed 1500 MB of raw image data in 46 sec. Result: 800 times data reduction to 1.9 MBShip segmentation near docks and at sea
.jpg)
Accurate segmentation of ships and other vesselsActivity-based recognition
Recognition of vehicle direction and velocity; colour-coded traffic density; approaching vehicle alert and vehicle in zoneCoastal detection and obscurants
Detecting six classes: seaside, land, cloud, haze, dust, and smoke
What comes in the kit?
- Pulsar Engineering Model or Flight Model
- Pulsar SDK and AI/ML compiler
- Application-specific software libraries (optional)
- The AI/ML compiler supports over 20 of the leading-edge frameworks such as TensorFlow, Apache MXNet, Keras, Pytorch, Caffe2, Darknet, scikit-learn, Microsoft Cognitive Toolkit.
.png)
space product updates
Register to receive news about our products for space in your email