AVEONI
Platform Vision Platform

The engineering environment we are designing for the whole drone system

Aveoni is designed as one environment for the flight controller, companion computer, AI models and video, so a team does not maintain a separate toolchain for each. This page shows that target architecture and how mature each part is, so the long-term vision and the verified scope stay separate.

Target architecture · vision

Six areas, one engineering layer

The intended shape of the platform. These are areas, not a status list. Which parts are delivered, and how far, is stated in the tables below.

Flight Systems

Flight controllers and their firmware, configured from one project.

Companion Computing

In Development Aveoni Companion is a first-party hardware companion computer. Raspberry Pi 5 and NVIDIA Jetson Orin Nano are separate reference platforms. Module runtime and CPU model deploy with read-back are recorded on a Raspberry Pi 5 reference companion. The Aveoni Companion target specification is in the companion computing table below.

AI & Autonomy

Model deployment, on-device inference, perception and autonomy.

Video & Connectivity

Cameras, live video and the links between aircraft and ground.

Cloud & Fleet

Operations across many aircraft, releases and device lifecycle.

Studio & SDKs

The engineering environment and the public interfaces for extension.

Studio & SDKs are designed as the engineering layer across all six areas: one project, one engineering workflow.

Five intended workflows

Architectural scenarios — illustrating the intended platform, not currently available end-to-end workflows.

  1. Connect & Configure IntendedConnect a flight controller and a companion, and configure them in one project.
  2. Build & Deploy IntendedBuild modules and models, and deploy them to a companion with read-back.
  3. Perceive & Decide IntendedCamera input feeding a model that informs decisions.
  4. Stream & Connect IntendedLive video and links to the ground..
  5. Monitor & Manage IntendedFleet status, releases and device lifecycle across aircraft..
Current Product: Aveoni Studio, by area

Flight hardware and flight software

CapabilityStatusScope and limits
Connect to Betaflight and read device identity over USBAvailableVerified on the Tykho Lit S4, Betaflight 4.5.1. Configuration writes are not hardware-verified. Other boards are not claimed.
ArduCopter 4.6.2 over USB: connect and read identityLimitedOne reference board (Tykho Lit S4). ArduCopter 4.6.2 is the recorded firmware. ArduPlane is not claimed as physically verified. Configuration writes are not hardware-verified. Other firmware versions are not claimed.
Read Betaflight identity through a companionLimitedRead-only: the board identity is read through the companion broker. Betaflight only, Tykho Lit S4 through a Raspberry Pi 5 or NVIDIA Jetson Orin Nano reference companion. Configuration through a companion is not verified. Studio offers the route; use through the production interface is not separately verified. ArduPilot through a companion is not supported. Studio refuses it by name.
Connect to iNAV and read device identity over USBPlannedNo iNAV support exists in Studio yet. Studio cannot connect to an iNAV flight controller today, and no board has been tested with it.

Safety, not a capability status: motor testing and firmware flashing are not listed in this table. A motor command can move a motor: remove propellers before any motor command. Firmware flashing is destructive: do not use it on a board you need.

Companion computing

CapabilityStatusScope and limits
Raspberry Pi 5 companionAvailableConnection and compute discovery. Reached over a direct Ethernet link in the 2026-09-16 session record. Local-network discovery is not shown in that record.
NVIDIA Jetson Orin Nano companionAvailableOn the NVIDIA Jetson Orin Nano. Connection and stack identification only. Not verified: model deploy, inference or TensorRT execution.
Aveoni Companion (first-party companion computer)In DevelopmentTarget specification, not a shipped product. Quad-core Cortex-A72 64-bit SoC at 1.5 GHz with 8 GB LPDDR4-3200 ECC memory and a Yocto-based Linux. VideoCore VI GPU with an integrated ISP and H.265/H.264 encode and decode up to 4Kp60. Hailo AI accelerator, 26 TOPS. Built-in IMU, magnetometer and barometer. One 4-lane MIPI CSI-2 camera interface and two analog video inputs, for digital, analog and thermal FPV cameras. Designed for Betaflight and iNAV flight controllers over CRSF, GHOST and MSP. Two UART/I²C/SPI/PWM ports and USB-C. 24 × 40 × 67 mm, 65 g (100 g with heatsink). Not yet tested with Studio.
Companion Module RuntimeAvailableInstall, start and stop modules on the reference companion. Starting a module is not a health check.
Full Drone RuntimePlannedThe deployment and execution foundation across the aircraft. Not the Companion Module Runtime above.

Models and AI

AI is where a drone earns its value in the field, and it is central to the platform. Studio already deploys a model to the companion and confirms the device holds exactly what was sent. The Aveoni Companion is being built with a 26 TOPS Hailo AI accelerator, so models can run on board, next to the camera, without a round trip to the cloud. Running a model on the device is not yet available.

CapabilityStatusScope and limits
CPU model deploy and read-backAvailableRaspberry Pi 5 reference companion only. Deployed means the intended file is installed and reads back as sent. It does not mean the model is running or the device is healthy.
Running inference on the devicePlannedNot reachable from Studio. Running inference on the device is not yet available.
Accelerator execution (NVIDIA, Hailo)PlannedAccelerator execution is not yet available. The Aveoni Companion is designed around a 26 TOPS Hailo accelerator; Hailo has not yet been tested with Studio on reference hardware. NVIDIA execution on the Jetson Orin Nano is not verified.

Cameras and video

CapabilityStatusScope and limits
UVC camera discovery through the companionAvailableVerified on a Raspberry Pi 5 bench: one UVC camera (046d:0821) reported by the companion, in a run artifact dated 2026-10-08. Discovery only. Video is not part of this record.
On-companion video runtimePlannedImplemented and tested in software. Camera capture on hardware is not verified.
Live camera video in the browserPlannedNot verified. Browser delivery of camera frames is not claimed.

Studio, projects and developer tools

CapabilityStatusScope and limits
Projects and component catalogueAvailableOne project per drone. The catalogue lists known hardware, not verified hardware.
Visual system editorAvailableDraw connections between components. Known linking gaps are recorded in the engineering record.
Public SDK and APIsPlannedNo public SDK exists yet.
Accounts and organisationsPlannedNot built.
Drone Kits (7″, 10″, Fixed-Wing)In DevelopmentScope and contents will be published when available.
Cloud and fleet toolsPlannedNot built. Multi-aircraft operations are not available.
Engineering Evidence

Proven on real hardware, run by run

Every route marked Available on this page comes from a dated run on reference hardware. Here are the flight-controller runs: what was connected, how, and what happened.

  • Pass

    Studio connected to a Betaflight flight controller over USB and read its identity.

    • Tykho Lit S4
    • Betaflight
    • Direct USB

    Run X3 · record c919132

  • Pass

    Studio reached a Betaflight flight controller through a Raspberry Pi 5 companion on the local network.

    • Tykho Lit S4
    • Raspberry Pi 5
    • Via companion

    Run X1 · record fccd105

  • Pass · 34/34 checks

    Studio reached a Betaflight flight controller through an NVIDIA Jetson Orin Nano companion.

    • Tykho Lit S4
    • Jetson Orin Nano
    • Via companion

    Run X7 · record b24f566

  • Pass

    Studio connected to an ArduCopter 4.6.2 flight controller over USB and read its identity.

    • Tykho Lit S4
    • ArduCopter 4.6.2
    • Direct USB

    Run X4 · record 88b10c7

  • Pass

    Studio worked with a Betaflight flight controller over USB while a Jetson Orin Nano companion was connected alongside.

    • Tykho Lit S4
    • Jetson Orin Nano
    • Direct USB

    Run X6 · record 5847357

  • Refused safely

    Asked to reach ArduPilot through a companion, a route it does not support, Studio refused it by name and never touched the board.

    • Tykho Lit S4
    • Raspberry Pi 5
    • Unsupported route

    Run X5 · record 88b10c7