Project information

  • Project date: Spring 2019
  • Techniques Used: Markov Decision Processes, Reinforcement Learning
  • Write Up:

Compiled MultiTaxi Project

Final Project for CMSC 478 (Introduction to Machine Learning) at UMBC. My goal was to explore the multi-agent version of the classic taxi domain, in which a robotic taxi must perform the functions of picking up passengers and dropping them off at the correct locations. Now, there are two taxis, compounding control difficulty and adding the issue of collisions. As this was an introductory course, I selected a centralized MDP and implemented basic Q-Learning, to decent results.