pedrodbs / frogger-study

Videos of RL agents playing the game of Frogger

Introduction

This repository contains several video clips depicting the game-play of different autonomous agents in the game of Frogger.

The agents' behavior was acquired through reinforcement learning (RL).

The video clips correspond to visual explanations produced by an eXplanation RL (XRL) framework. Each video captures a different aspect of an agent's interaction with the environment.

Structure

• 1-maxima: all observation maxima
• 2-minima: all observation minima
• 3-certain: all certain execution observations
• 4-uncertain: all uncertain execution observations
• 5-frequent: all frequent observations
• 6-infrequent: all infrequent observations
• 7-transition-value: half minima, half maxima
• 8-execution certainty: half certain-execution, half uncertain
• 9-frequency: half frequent, half infrequent
• 10-all: one of each type
• 11-sequence: sequence from minima to maxima