Skip to content

Repository files navigation

minirov

A tethered underwater ROV with LLM-assisted supervised autonomy, built for UK coastal and lake shore deployments.

A locally-hosted LLM (Ollama / qwen2.5:14b) running on an M4 Pro Mac interprets natural language commands and executes them via ROS 2 middleware. The operator retains full manual override at all times.


System Architecture


M4 Pro — Ollama + minirov-brain (operator interface) │ │ ROS 2 over network │ Lenovo Ubuntu — ROS 2 Jazzy (serial bridge) │ │ JSON over USB serial │ Pico 2W — custom flight controller (IMU + attitude + PWM) │ 4× T200 thrusters \


Status

Phase Description Status
0 Software foundations and SITL ✅ Complete
1 Bench build and electronics 🔄 In progress
2 Enclosure and frame assembly 🔲 Pending
3 Pool and tank testing 🔲 Pending
4 Field deployment 🔲 Pending

Phase 0: Full LLM → MAVLink pipeline verified against ArduSub SITL. Natural language → Ollama (qwen2.5:14b) → JSON → pymavlink → ArduSub SITL → thruster RC override confirmed.

Phase 1 (current): Pico 2W flight controller in development.


Hardware

Component Part
Topside compute Apple M4 Pro MacBook
Middleware Lenovo laptop — Ubuntu 24.04, ROS 2 Jazzy
Flight controller Raspberry Pi Pico 2W (custom firmware)
IMU GY-521 MPU-6050
Thrusters 4× Blue Robotics T200
ESCs Blue Robotics Basic ESC
Tether Blue Robotics Fathom 50m
Tether interface Blue Robotics Fathom-X ×2
Enclosure Blue Robotics 4" Watertight (acrylic)

Full BOM: docs/hardware/bom.md


Software Stack

Layer Technology
LLM inference Ollama / qwen2.5:14b
Autonomy brain Python — minirov-brain
Middleware ROS 2 Jazzy
Serial bridge ROS 2 node — JSON over USB
Flight controller MicroPython — pico-fc
SITL comms pymavlink (simulation only)

Author

Hari Aditya github.com/hariaditya97

About

Tethered ROV with LLM-assisted supervised autonomy for UK coastal and lake deployment.

Topics

Resources

Stars

0 stars

Watchers

2 watching

Forks

Releases

Packages

Contributors

Languages