The goal that this project hopes to accomplish is to create an autonomous RC car which makes driving decisions via a neural network. The car will be able to drive down a hallway while avoiding obstacles. Visual data collected by the car will be sent through a Convolution Neural Network (CNN) in order to determine where the car should go next.

 

There are many benefits to utilizing autonomous vehicles. Self-driving cars can automate tasks and increase safety, thus reducing human error and increasing efficiency. The implementation of neural networks within this project will allow our vehicle to push the boundaries of traditional automation.

 

In traditional programming, the system is given explicit commands based on a given input in order to output a desired result. In the case of automating the task of driving, this would include items such as object detection and decision making based on a large set of previous data. A traditionally programmed machine will do what it has been told to do and nothing more. Our neural network will be able to make decisions without having been previously told what the correct answer is. This means decisions can be made about a more real-world set of problems instead of just lab based “clean” environments. A machine learning self-driving car can take things it learns about driving in a certain location or certain conditions and apply that information to other situations it encounters. Additionally, as the car makes mistakes such as causing a collision or encounters new obstacles, the neural network will be able to take note of this and improve its decision making in future events. 

​

​