Software · Vision System
Balor JUL 2026
Moving-object detection & tracking system for resource-constrained embedded targets.
| Function | Detect & track |
|---|---|
| RAM | 512 kB |
| Processor | ARM Cortex-M33 · dual-core |
| Frame rate | 60 FPS |
Release · Jul 2026
Aerial Robotics & Machine Vision
We develop ultra-performant state-of-the-art software and hardware for machine automation.
Our solutions work on mass-produced retail microcontrollers and do not require datasets and AI training, costly or limited-supply hardware (TPUs, NPUs, AI accelerators, GPS) or large RAM.
The ARX ARCENDA approach
Engineered for commodity hardware
Every product is engineered to extract maximum throughput from commodity hardware — eliminating the cost, supply risk and complexity of accelerator-bound platforms.
Runs on Mass-produced retail microcontrollers / ARM Cortex-M & Cortex-A / RISC-V / x86
01 / Machine Vision
Detection & tracking on the edge
Real-time moving-object detection and tracking that fits inside kilobytes of RAM — no neural accelerator, no training pipeline, no cloud.
Balor delivers a complete detect-and-track stack on a dual-core Cortex-M33 in just 512 kB of RAM. For higher-rate workloads, the AMV32H747V1 board pushes the same pipeline to 240 FPS on a Cortex-M7.
Hardware · Vision Board
AMV32H747V1 Q4 2026
High-performance board for moving-object detection & tracking at elevated frame rates.
| Function | Detect & track |
|---|---|
| RAM | 4 MB |
| Processor | Cortex-M7 · dual-core |
| Frame rate | up to 240 FPS |
Pre-order opens Jul 2026
$40Pre-order
02 / ImPro
Microcontroller-native computer vision
A computer-vision library built for CPUs — and a visual design tool to compose pipelines without writing accelerator code.
The ImPro library is optimised across Cortex-M, Cortex-A, RISC-V and x86, running morphology operations 4× faster on a Cortex-A CPU than Metal and OpenCV on GPU. Image Pro provides a CAD environment for designing and validating those pipelines on macOS and iPadOS.
Software · CV Library
ImPro Library
Ultra-fast CPU-based computer-vision library, architecture-portable by design.
| Compute | CPU-based |
|---|---|
| Optimised for | Cortex-M / -A · RISC-V · x86 |
| Morphology | 4× vs GPU (Cortex-A) |
| Baseline | Metal & OpenCV on GPU |
Core library
Software · Design Tool
Image Pro AUG 2026
CAD environment for computer-vision pipeline design and validation.
| Role | Pipeline CAD |
|---|---|
| Platforms | macOS · iPadOS |
| Output | ImPro pipelines |
| Planned release | Aug 2026 |
Planned · Aug 2026
03 / Avionics
Built for >400 km/h flight
A complete avionics stack — firmware, flight controller, speed controller and visual-inertial odometry — engineered from scratch for high-speed autonomous UAVs.
FlashFlight pairs the responsiveness of Betaflight with the feature depth of PX4 and ArduPilot. It drives the AFC32H747V1 controller, the AESC8AT328PV1 speed controller and the AVIO32H747V1 odometry unit — coordinated over MAVLink with inertial and visual navigation.
Software · Flight Firmware
FlashFlight
New-generation flight-controller firmware for autonomous UAVs, written from scratch for extreme speed.
| Target | > 400 km/h |
|---|---|
| Profile | Betaflight speed · PX4/Ardu depth |
| Protocol | MAVLink |
| Navigation | Inertial & visual odometric |
Flight firmware
Hardware · Flight Controller
AFC32H747V1 Q4 2026
Flight-controller board designed for FlashFlight and optimised for high-speed flight.
| Designed for | FlashFlight |
|---|---|
| Compatibility | Pixhawk 6X |
| Optimised for | > 400 km/h |
| Release | Q4 2026 |
Pre-order opens Sep 2026
$250Pre-order
Hardware · Speed Controller
AESC8AT328PV1
Electronic speed controller for high-performance propulsion.
| Motors | 4 × Brushless DC |
|---|---|
| Output | 480 A @ 24V · 12 kW |
| Protocols | DSHOT 600 / 300 / 150 |
| Feedback | Back-EMF |
Propulsion
Hardware · Odometry
AVIO32H747V1 Q4 2026
Visual-inertial odometry controller for GPS-independent navigation.
| Type | Visual-inertial odometry |
|---|---|
| Sensing | Binocular |
| Update rate | 1000 Hz |
| Compatibility | PX4 |
Release · Q4 2026
04 / Connectivity
Coordination at 250 km
A long-range mesh built to keep autonomous fleets connected across kilometres — and to hold a link when the spectrum is contested.
The Hydra protocol orchestrates autonomous robotic vehicles over LoRa-based mesh links in the 868 and 915 MHz bands, reaching up to 250 km. The AERIS32LR2022V1 transceiver carries those links reliably through noisy electronic-warfare environments at low bandwidth.
Software · Mesh Protocol
Hydra Q1 2027
Mesh network protocol for orchestrating autonomous robotic vehicles over kilometre ranges.
| Basis | LoRa |
|---|---|
| Bands | 868 MHz · 915 MHz |
| Range | up to 250 km |
| Release | Q1 2027 |
Release · Q1 2027
Hardware · Radio Transceiver
AERIS32LR2022V1 Q1 2027
Ultra-reliable long-range radio link transceiver for low-bandwidth comms in noisy EW environments.
| Protocol | LoRa |
|---|---|
| Bands | 868 MHz · 915 MHz |
| Bandwidth | 11 baud |
| Range | up to 250 km |
Release · Q1 2027
Contact
Contact sales
Discuss integration, evaluation units, or pre-order volumes for any product in the portfolio. Technical inquiries reach our engineering team directly.