Bracu DUBURI 3.0 Software Architecture

Duburi's Computer Vision subteam is in charge. Its job is to choose, train, and test Duburi's Machine Learning Algorithms for identifying objects underwater, which is an important aspect of Duburi's autonomous operation.

Simulation

The environment was built using Unity. The model of Duburi was placed in this simulation and based on the model’s location, different stimulus data was fed into Duburi’s microcontroller code and machine-vision algorithms to simulate the input from the sensors. The simulation was run on a sufficiently capable laptop and the microcontroller code and machine vision algorithms were run on the Duburi’s Arduino Mega and Jetson Nano respectively.

Interfacing

Duburi's microcontroller board, the Arduino Mega, strikes at the core of its operation. It functions as a control surface between the sensor payload's inputs and the actuators and end effectors on the Duburi, it manipulates the Duburi's thrusters in line with the data signals it receives to maintain steady locomotion.

OBJECT DETECTION