Reinforcement Learning Project - Mountain Car

This repository contains the code and report for the Reinforcement Learning project about Mountain Car.

Mountain Car

The goal of this problem is to get the under-power car to the top of the mountain. The states consisted of 2D continuous values of position (-1.2 to 0.6) and velocity (-0.07 to 0.07), and the available actions are move left (0), stay (1) and move right (2). For each time step, there is a reward (penalty) of -1 until the goal is reached.

Tools and Libraries

• OpenAI Gym
• NumPy
• Scikit-learn

Algorithms

• Q-Learning
• SARSA
• Radial Basis Functions

Learning Curve