I pledge my honor that I have abided by the Stevens Graduate Student Code of Academic Integrity.
The Machine Learning Agents for Self-driving Car is a project that uses machine learning technique to train an agent for a self-driving vehicle in a simulated environment build by Unity3D. The environment includes several pedestrians, traffic lights and other vehicles. All the pedestrians will be set an individual agent to decide if they can cross the road safely. The agent of walker will detect the distance between the certain pedestrian and car, then calculate his own speed and make decision. The other vehicles except the self-driving car will follow their own fixed routes.
The final goal of this project is to train a qualified self-driving car which can predict pedestrians' movement and avoid collision with other vehicles.
Key Words:
Machine Learning, Self-driving Car, Unity3D, TensorFlow, Python API, ML-agent.
Self-driving Car is a vehicle that is capable of sensing its environment and navigating without much human input. Self-driving cars combine a variety of techniques to perceive their surroundings, including radar, laser light, GPS, odometry, and computer vision. Advanced control systems interpret sensory information to identify appropriate navigation paths, as well as obstacles and relevant signage.
The Machine Learning Agents is a toolkit for Unity3D Game-engine that using Python API to enable games and simulations to serve as environments for training intelligent agents. Agents can be trained using reinforcement learning, imitation learning or other machine learning methods. They would be used for 3D games made of Unity3D. A simple 3D game will be set as training environment (the certain game is not determined yet). One single machine learning agent will be created and trained to be used for multiple purposes, including controlling NPC (Non-Player Character) behaviors, learning and simulating human operations, automated testing of game builds and evaluating the difficulty of games.