The 2014 General Video Game Playing Competition

Abstract

Abstract-This paper presents the framework, rules, games, controllers and results of the first General Video Game Playing Competition, held at the IEEE Conference on Computational Intelligence and Games in 2014. The competition proposes the challenge of creating controllers for general video game play, where a single agent must be able to play many different games, some of them unknown to the participants at the time of submitting their entries. This test can be seen as an approximation of General Artificial Intelligence, as the amount of game-dependent heuristics needs to be severely limited.The games employed are stochastic real-time scenarios (where the time budget to provide the next action is measured in milliseconds) with different winning conditions, scoring mechanisms, sprite types and available actions for the player. It is a responsibility of the agents to discover the mechanics of each game, the requirements to obtain a high score and the requisites to finally achieve victory. This paper describes all controllers submitted to the competition, with an in-depth description of four of them by their authors, including the winner and the runner-up entries of the contest. The paper also analyzes the performance of the different approaches submitted, and finally proposes future tracks for the competition.

Analysis of Vanilla Rolling Horizon Evolution Parameters in General Video Game Playing

“…The winner of the first edition of the competition in 2014, Adrien Couëtoux [23], employed an Open Loop technique quite similar to this algorithm.…”

“…The General Video Game AI Competition (GVGAI) [22,23] offers a large corpus of games described in a plain text language, making it easy to run general AI agents in several different environments and analyse their performance. The competition has already completed three editions of its single player track (starting in 2014), with two additional tracks running in 2016 for two player games [7] and level generation [11].…”

MCTS/EA hybrid GVGAI players and game difficulty estimation

“…The inherent nondeterminism discourages plain game-tree search methods and renders this environment suitable to non-deterministic learning algorithms such as Monte-Carlo Tree Search (MCTS) and Evolutionary Algorithms (EA). Both approaches have been shown to work well, but MCTS based controllers tend to exhibit the best overall performance [14]. Still, as of yet, no submitted controller has been able to consistently be successful on all games, showing that all controllers have their weaknesses and strengths.…”

“…GVGAI utilizes this implementation, providing a responsive forward model to simulate actions within the game and an interface for controllers to participate in an open competition. The results and rules of this contest, which was initiated in 2014, can be found in [14].…”

“…The single-player planning track was run as a competition in 2014 [14] and 2015, attracting more than 70 entries in total.…”