Today we learned how to create functions which greatly simplifies coding and speeds up the process.

1. Describe at least one “physical” iteration to your robot that you have made, answering these prompts: 

   1. How did you know that you needed to make a change to your robot (describe the testing process and what you learned from testing). Include a picture if you are able. 

I noticed that my robot’s wheels were slipping and did not have a good grip on the floor.

What specifically did you change?  

I added rubber bands around the wheels which added much needed traction.

2. How did you know/verify that your change helped your robot’s function? 

After this My robot went much faster and turned much faster as the wheels no longer slipped.

2. Describe at least three “iterations” to your code that you have made, in order (three successive changes): 

   1. How did you know that you needed to make a change to your code (describe the testing process and what you learned from testing). Include a line of code to aid in your description if you are able. Originally, when attempting to get my robot to go forward, I typed: 

for (pos = 0; pos <= 3600; pos += 1) { 

    myservo.write(pos);                delay(5);     

2. What specifically did you change?  Include a line of code in your description.

I changed this to a much more simplistic servo.write which helped me edit my code faster

3. How did you know/verify that your change helped your robot’s function? 

I was able to alter my robots code much easier, usually I only had to type two or three numbers.

Lock Research

Idea Development

I would like to create a lock using electromagnetism, I could affix the electromagnet on the lid of the box and have a switch turning it on and off for when you want it to be locked as well as a hidden override button.

MACHINE LEARNING PROGRAM in association with Georgia Tech

Brief

Machine learning is when a program uses mathematical models to learn things and improve itself. For this assignment, we were tasked with creating a machine learning model that would identify varying objects using prior knowledge and segments of code from previous assignments. For my model, I created a program that would identify four shapes, a cylinder, cube, star, and pyramid. I produced these models using tinkerCAD and the school's 3d printers.

Final Model

This is my model which identifies different shapes, it uses the above greyscale images and a machine learning model to improve it's detection of different shapes.

Used Dataset

This is the dataset used in the program, each folder holds 200 of the used pictures. The name of the folder is the true label in this instance which is called upon later.

Building the Model

The first layer is the data layer containing the training tensors named data. The second layer of this model is a dense input layer with a sigmoid activation code. The third layer is a dense layer with a softmax activation code which is the output. The first code retrieves the program's data and then the sigmoid function produces a value that is 1 or 0, essentially yes or no, telling the program whether its labeling was correct. The sigmoid function allows for generalization and for the program to make accurate guesses in the future as to what the object's label is.

Correct Labeling

These predictions are correct as the prediction matches up with the true label. The true label is the assigned label in the data set which can be pulled for reference against the prediction in order to calculate its accuracy.

Machine Learning Inconsistencies 

These pictures are predicted incorrectly as the prediction does not match up with the label. It seems the program has a problem distinguishing pyramids, most likely from the angles the pictures were taken from. This would easily be fixed by providing more or better pictures of the pyramid