RIVER SAVER

River Saver is a robot that born with a mission to assist people collecting floating debris. Approaching the situation, I've combined both Artificial Intelligence and Automation to form a robot. I've worked through this project with my co-worker: Hoang Nguyen and my mentor: Triet Do

Proposed Designs

Throughout the duration, we've tested out 3 designs in total, mostly using Fusion 360 and TinkerCAD to draw and visualize model

Model 1

At first, we intend to use pump motors placed under the bucket, but it's not power efficiency

Eco-friendly feature

Overall, the way that our project stands out is because of its unique method of collecting the debris, without using any energy at all!

In this approaches, we've used a passive door that could rotate and collecting the debris while it's moving.

Range table control method

Left/Right Motor

PID Controll, source: PID controller - Wikipedia

Speed

(rad/min)

Controlling System

To make the robot rotates in the desired angle, we've tested between PID Controlling and manual range control with the magnetic sensor as input

In the end, we chose range control for its simplicity and easier to manipulate the robot.

Object detection

Model optimization

In order to detect debris, we used object detection model Yolov4 - tiny version so that it can process in real-time on mobile platform.

In order to optimize the model faster and more light weight, I've optimized the weights from FP32 (Float 32 bits) to INT8 (Integer 8 bits), which only creates a small fluctuation in prediction, but boosts the time taken to process per image.

Object detection - Optimization, tracking, and navigation

05 Results

Back