Assistance Needed for Autonomous Underwater Vehicle (AUV) Development

General Discussion
1

Hi Blue Robotics Community,

I’m currently working on a university project to build an Autonomous Underwater Vehicle (AUV) designed for navigation in a glass tank measuring approximately 1.5m x 4m x 1m. As I’m relatively new to this area, I would appreciate your insights on a couple of key points.

Background:
We have two ROVs to work with: a BlueROV 1 and a smaller custom-built ROV equipped with four T100 thrusters. We are using a Raspberry Pi 4 for motor control and plan to implement object detection using OpenCV, along with an e-con Raspberry Pi Camera.

Questions:
Flight Controller Recommendations: The Raspberry Pi struggles with PWM control (only allowing 5% increments). We’re considering using a flight controller to improve control and integrate features like gyro and compass functionality. Which flight controllers would you recommend for seamless integration with the Raspberry Pi? I noticed the Pixhawk is mentioned often, but I’d like to know if it’s user-friendly for object detection tasks. Other options like Naivo 2 and PXF Mini have also caught my attention—what are your thoughts?
Close-Range Detection Solutions: We are looking for effective methods to detect the glass walls of the tank. Ultrasonic sensors have proven ineffective due to their minimum range limitations. We are considering laser detection but would appreciate any suggestions on what types or models to use for close-range detection.
Any guidance or recommendations would be greatly appreciated! Thank you for your help.

2

Hi @johnz -
Welcome to the forums!
It sounds like quite a cool project. Blue Robotics would of course recommend our Navigator flight controller, and BlueOS on your Pi4!
It’s worth noting that flight controllers are not typically where object-detection tasks would be executed - microcontrollers don’t typically have enough computational power for this!
Making an extension for BlueOS that handles the openCV object recognition and generates control messages for the system should be possible!
Detecting the walls of the tank will likely be quite tricky - is placing locator QR tags (like april tags or other similar projects) possible? The vehicle could know its relative position to these tags and avoid moving too far in the wrong direction from them!

This github repository may be useful as well!