Backpropagation Algorithm Applied to a Line-Following Robot

Since the early 21st century, artificial intelligence has advanced at a dizzying pace thanks to the development of various learning algorithms, which have turned machines into extraordinary systems capable to interact with and help human beings in everyday and hazardous tasks. This evolution has sparked interest in building a low-cost plant known as a line-following robot (LFR), with the ability to autonomously navigating a platform marked with a colored line. This can be applied to the development of precision agriculture, material transport in industrial environments, among other human activities. This is achieved through the creation of an artificial neural network (ANN), which uses the backpropagation algorithm to adjust the weights of each neuron relative to the error of each output. This adjustment is possible by the mathematical method of gradient descent, aiming to minimize the error in each of the three layers of the neural network: Input and output layers that use the SIGMOID activation function and hidden layer with RELU activation function, which seeks to obtain information from the environment through its infrared (IR) sensors, process it in the neural network and activate the motors according to previous learning.