Projects

What we are building next: AI in orbit, lunar rover avionics and next-generation space GNSS.

Alongside our products, we run R&D programmes that push domestic space technology forward. Neuro-Orbit is supported by TÜBİTAK-TEYDEB.

AI in orbit · 2026–2028 · supported by TÜBİTAK-TEYDEB

Neuro-Orbit: AI-Enabled Satellite Flight Computer

Today’s satellites send almost all raw data to the ground and depend on operators to spot problems. Neuro-Orbit changes that. It is a flight computer that runs artificial-intelligence workloads in orbit, in real time, while keeping the determinism and reliability a spacecraft needs. The goal is to take the technology from TRL 3 to TRL 6.

Neuro-Orbit processing module 3D render

One board, many jobs

  • AI Inference Engine: common on-board inference service
  • Anomaly Detection: learns healthy telemetry and flags faults autonomously
  • AI Star Tracker: attitude at ≥ 20 Hz from star images
  • AI Navigation: GNSS orbit and time determination, ≥ 1 Hz PVT + PPS
  • Image Processing: classification, cloud screening, compression, encryption
  • Model Manager: uploads, validates and swaps AI models in orbit
  • Radiation-Aware Scheduler: adapts inference to radiation events (FDIR)
  • Federated Learning: shares model improvements across a constellation
Neuro-Orbit architecture HARDWARE Flight Controller Radiation-tolerant FPGA SoC RISC-V cores SpaceWire MIL-STD-1553B MRAM/DDR/EDAC CAN / CAN-FD Flight SW · FDIR AI Accelerator Adaptive SoC with AI engines AI engines ARM cores FPGA fabric ECC memory · redundant boot flash High-speedserial interconnect AI APPLICATIONS Inference Engine Anomaly Detection AI Star Tracker (≥ 20 Hz) AI Navigation (GNSS PVT) Image Processing Model Manager Radiation-Aware Scheduler Federated Learning FLIGHT SOFTWARE cFS core services · executive · events · software bus · tables · time · FDIR OS abstraction layer · platform support package RTEMS real-time operating system
Neuro-Orbit roadmap 2026-2028 DEVELOPMENT ROADMAP Q3 2026Kick-offTRL 3 Q4 2026Systemrequirements Q1 2027Hardware & softwarearchitecture Q2 2027Prototype boardinterface tests Q4 2027Engineering modelAI models on board Q1 2028HIL/SIL campaignECSS V&V Q3 2028TRL 6

Less downlink

Only useful results go to the ground, with a target of ≥ 95 % bandwidth saving on hyperspectral data.

Faster reaction

On-board fault detection in under 100 ms instead of hours.

Fewer boxes

One board can replace separate OBC, star tracker processor, GNSS receiver and payload computer, saving harness, mass and power.

Updatable intelligence

AI models are uploaded, validated and improved during the mission.

Neuro-Orbit is supported by the TÜBİTAK-TEYDEB 1501 Industrial R&D Projects Support Programme (Project No: 3261148).

Moon surface · engineering design complete

Lunar Rover Avionics: TTM-RVR-KIT

A rover on the Moon has no mechanic nearby. Every board has to survive radiation, lunar day-night temperature swings and long periods of unattended operation. Our lunar rover avionics kit is a set of SpaceVPX-compatible modules on a common backplane. You scale it by adding modules, not by redesigning.

TTM-PM-SVPX processing module

TTM-PM-SVPX · Processing

Quad-core COTS SoC, antifuse FPGA system controller, ECC-protected DDR and MRAM, SEL protection with LCL.

TTM-ADIM-SVPX analog-digital interface module

TTM-ADIM-SVPX · Analog-Digital I/O

12 × RS-422 RX, 12 × RS-422 TX, 2 × SpaceWire, 4 × CAN, 4 × RS-485, 16 opto-isolated GPIOs, 8 analog inputs.

TTM-MDM-SVPX motor drive module

TTM-MDM-SVPX · Motor Drive

8 × MOSFET motor-drive channels and 8 analog measurement channels.

TTM-PDM-SVPX power distribution module

TTM-PDM-SVPX · Power Distribution

Redundant 28 V inputs with EMI filter; monitored and protected 5/12 V rails; configurable 28/±15/±12/5 V outputs.

Each module: 160 × 100 × 20 mm. The same kit also fits landers, other surface missions and terrestrial robotic test beds.

Navigation · FPGA IP core in test

TetraGNSS: Space GNSS Receiver on FPGA and ASIC

GNSS receivers for space have to track satellites while moving at 7.5 km/s. They also have to survive radiation and fit a small power budget. Our team designed a flight-proven space GNSS receiver before, and TetraGNSS is the next generation. It is being developed in three steps:

1 · FPGA IP core

Multi-constellation (GPS / Galileo / GLONASS / BeiDou) baseband with high-dynamics tracking for LEO. It drops into our flight computers, so an OBC can also act as the GNSS receiver.

2 · GNSS receiver ASIC

The proven FPGA baseband moved to a space-compatible ASIC for lower power and mass.

3 · Receiver equipment

Complete unit with RF front-end, SoC/SoM processing and TetraGNSS software. It is verified on a GNSS simulator with static, automotive, aircraft and spacecraft scenarios.

IP cores · in verification

Space-Ready FPGA IP Cores

ECSS-compliant MIL-STD-1553B BC/RT and SpaceWire/RMAP IP cores. We will use them in our own flight computers and license them to other equipment developers.

Hardware platform

Modular Flight Boards

Every product is built from the same family of boards: processing, FPGA/SoC, SDR, GNSS ASIC, RF front-end, power, platform-interface, discrete-interface and ADCS-interface modules. See the modules →

Concept · AI payload

OrbitEye: AI-Based RF Spectrum Monitoring from Orbit

OrbitEye is an on-board AI payload concept for LEO satellites at around 600 km. During each pass it scans the RF spectrum up to 6 GHz, then detects, classifies and geolocates emitters and maps RF activity. Each pass is compared with earlier ones to reveal new signals, changes and anomalies, and only prioritised, high-value data is sent to the ground.

OrbitEye concept: a LEO satellite scanning the RF spectrum and mapping activity on the ground

Detect & classify

Wideband spectrum monitoring with AI-based signal detection and classification on board.

Geolocate & map

Emitter geolocation from orbit and georeferenced maps of RF activity.

Compare over time

Multi-pass comparison shows new or vanished signals, trends and anomalies.

Smart downlink

The AI selects which IQ data to record and which results to send, making the most of limited downlink.

OrbitEye concept pipeline: detect, classify, geolocate, map, compare, discover

Concept pipeline: from a raw RF spectrum to a georeferenced activity map.

OrbitEye concept: RF activity maps compared across three satellite passes

Concept renders: comparing RF activity across passes highlights what changed.

Interested in partnering on one of these projects?

We work with integrators, research institutes and new space companies. Let’s talk.