Edge AI on the orbit - Bruhnbruhn case study
High‑Performance Stereo Matching for Autonomous Navigation
SoftServe has conducted rigorous benchmarking of stereo‑matching algorithms—both classical and AI‑based—on radiation‑hardened edge hardware, demonstrating that modern space-grade processors like the Blue Marble Space Edge Processor (SEP) can handle advanced perception tasks without needing specialized FPGA implementations. Classical algorithms (e.g., block‑matching, SGBM) show real‑time throughput, while deep learning approaches (like Fast‑ACVNet+) deliver denser, higher‑quality depth maps suitable for detailed navigation, especially in lunar-like terrains where lighting and texture can be challenging.
Streamlined Development with COTS Hardware and Simulation Tools
SoftServe highlights the transformational potential of modern rad‑hard platforms such as the SEP, which integrate x86 CPU, GPU, and FPGA—all supported by Linux and mainstream AI/ML ecosystems (e.g., ROS, TensorFlow, PyTorch). This compatibility allows developers to build, test, and iterate using the same codebase across simulators and target hardware, enabling “test‑as‑you‑fly” practices. By bridging simulation environments (e.g., Gazebo, Isaac Sim) and real deployment quickly, teams can reduce development time, lower risk, and reuse prototypes seamlessly in flight-grade systems.
Edge‑Ready Intelligence for Autonomous Exploration
Combining efficient stereo‑matching performance with a flexible, simulation‑friendly compute architecture means SoftServe equips teams to develop rich, autonomous capabilities directly onboard their space systems. This includes real-time perception, hardware‑in‑the‑loop testing, and the ability to deploy advanced vision algorithms at the edge. It paves the way for surface exploration that’s robust, adaptive, and less reliant on ground intervention—crucial for lunar or planetary missions.
🏆 How it helps you WIN?
With SoftServe, your team masters perception and autonomy in simulation—then confidently deploys the same high‑quality stereo vision and AI‑powered navigation onto your hardware. That means fewer surprises during the competition, smoother HIL integration, and a mobile system that navigates rugged terrain like a pro.
Want to win? Contact us and get ahead of competition!🚀
Further reading
🛰️ Stereo matching algorithms for space use