Moving Robotic Toilet

Description

To refine my skills and test myself, I set on the ambitious journey of building an automatically aiming toilet. Using a "controller" I created using a repurposed siracha bottle, the user would need to fill the bottle with water, stand infront of the machine, hold the bottle near waist height, aim at any position within a 1x1 meter square, and the machine will never let the user spill water on the floor. Using a variety of sensors, the machine will rapidly move the bowl underneath the stream of water.

Computer Vision

Using a GoPro, the Intel RealSense Depth Camera, retroreflective tape, and linear algebra matrix math, I used the Python OpenCV library and depth camera information to detect the X, Y, and height position and angle of the bottle, and toilet bowl. Based on all this informaiton, as well as a flowsensor to detect initial velocity, I was able to use kinematic physics equations to predict the landing position of the liquid before landing, as well as what direction and how far to move the bowl to catch it.

Arduino/Electronics

To better organize the project, I isolated all movement programming to the Arduino, and computer vision and physics calculations to a separate, more powerful PC. The Arduino controlled 2 DC motors, reading rotary encoder information for each, as well as flowsensor data from the bottle to calculate initial velocity using Bernoulli's principle and the diameter of the bottle opening. The arduino was responsible for controlling the bowl using a system of belts and pulleys, which a similar layout commonly used by 3D printers (CoreXY). The arduino has a function that receives movement coordinates and moves the bowl accordingly, and the movement coordinates are received via USB from the computer vision software, which runs physics calculations.