By Jeongyong Chris Yang, VI Form
Autonomous Navigation and Decision-Making Process Using Machine Learning and Deep Learning
Autonomous vehicles are self-driving cars that do not require human drivers. They use sensors that are attached to the vehicle as their vision to detect their environment. After the vehicle detects other objects or signals, computer programming (coding) allows them to react to the situations adaptively. Even though the sensors do not need to be improved, the millions of situations the cars can face on roads create difficulties for people to build a sophisticated computer program that makes the autonomous vehicles completely safe on roads.
First, I decided to build an algorithm pseudocode to help resolve this problem. During this process, I built mazes and followed the instructions based on the algorithm manually to check whether the algorithm is effective. I mainly used three different models for my mazes, each with different difficulty levels to ensure that the algorithm works every time. Then, I decided to record the information (velocity and displacement for both x and y directions) about the vehicle on the map so that the following vehicles can get a picture of the map automatically. However, if the subsequent vehicle detects a different or an altered map with its sensors, the new information will also be recorded on the map. Finally, the final vehicle will follow the path set by the first vehicle, but the map will guide the car with the most efficient path after completely learning and optimizing the possible paths.
To read the full project write-up, CLICK HERE.
Jeongyong(Chris) Yang is a VI Form boarding student from Seoul, Korea. He likes to study mathematics and computer science, plays video games, and enjoys spending time with his friends and family.