The following competitions are being held at CIG 2018:
- Short Video Competition
- Hearthstone AI
- The Ms. Pac-Man Vs Ghost Team Competition
- Fighting Game AI Competition
- microRTS Competition
- Hanabi Competition
- StarCraft AI Competition
- The General Video Game AI Competition – Learning Track
- 3rd Angry Birds Level Generation Competition
- The Text-Based Adventure AI Competition
- Visual Doom AI Competition 2018
Computational Intelligence in Games – Short Video Competition
- Simon Lucas (Queen Mary University of London, UK; firstname.lastname@example.org)
- Alexander Dockhorn (University of Magdeburg, Germany; (email@example.com)
- and Jialin Liu (Southern University of Science and Technology, China; firstname.lastname@example.org)
Summary: The first IEEE CIS Short Video competition took place as one of the competitions for IEEE CIG 2018 in Maastricht on August 15th. The aim is to run these short video competitions in conjunction with all IEEE CIS sponsored conferences, as a way to provide source of interesting videos showcasing CI.
The videos were presented in a plenary session, with the audience voting to deciding the ranking. Congratulations to the winner, runners up and indeed all the entrants for their excellent videos.
The results were:
- (1st place) Evolving Mario Levels in the Latent Space of a Deep Convolutional Generative Adversarial Network, Vanessa Volz, https://youtu.be/NObqDuPuk7Q
- (2nd place) Automated Curriculum Learning by Rewarding Temporally Rare Events, Niels Justesen and Sebastian Risi,https://youtu.be/H6K0bOCCnQ8
- (3rd place) General Win Prediction: CIG 2018 Short Video Competition, Raluca Gaina and Diego Perez-Liebana, https://youtu.be/zq9zaEjspUY
Other entries (no particular order)
- Coupled Empowerment Maximisation for Adaptive, Believable Game Characters, Christian Guckelsberger, https://www.youtube.com/watch?v=r0JYg3PRFV0&
- Overview on the shallow decision-making analysis in GVGAI, Ivan Bravi, https://youtu.be/8fUAkvJaxzI
- Tilt, Giovanni Rubino, Simon Colton and Joan Casas-Roma, https://vimeo.com/285266868
- Reusability of Evolved MCTS Tree Policies for General Video Game Playing, Ivan Bravi,https://youtu.be/26qKn_jG7VQ
- Showcase of AI techniques in StarCraft, Martin Rooijackers, https://youtu.be/5guDwZf_txA
- Alexander Dockhorn (University of Magdeburg, Germany; (email@example.com)
- and Sanaz Mostaghim (University of Magdeburg, Germany; firstname.lastname@example.org)
Description: The collectible online card game Hearthstone features a rich testbed and poses unique demands for generating artificial intelligence agents. The game is a turn-based card game between two opponents, using constructed decks of thirty cards along with a selected hero with a unique power. Players use their limited mana crystals to cast spells or summon minions to attack their opponent, with the goal to reduce the opponent’s health to zero. The competition aims to promote the stepwise development of fully autonomous AI agents in the context of Hearthstone. During the game, both players need to play the best combination of hand cards, while facing a large amount of uncertainty. The upcoming card draw, the opponent’s hand cards, as well as some hidden effects played by the opponent can influence the player’s next move and its succeeding rounds. Predicting the opponent’s deck from previously seen cards, and estimating the chances of getting cards of the own deck can help in finding the best cards to be played. Card playing order, their effects, as well as attack targets have a large influence on the player’s chances of winning the game. Despite using premade decks players face the opportunity of creating a deck of 30 cards from the over 1000 available in the current game. Most of them providing unique effects and card synergies that can help in developing combos. Generating a strong deck is a step in consistently winning against a diverse set of opponents. Tracks: The competition will encourage submissions to the following two separate tracks, which will be available in the first year of this competition: Premade Deck Playing”-track: the participants will receive a list of decks and play out all combinations against each other. Determining and using the characteristics of player’s and the opponent’s deck to the player’s advantage will help in winning the game. User Created Deck Playing-track: invites all participants in creating their own decks or choosing from the vast amount of decks available online. Finding a deck that can consistently beat a vast amount of other decks will play a key role in this competition track. Additionally, it gives the participants the chance in optimizing the agents’ strategy to the characteristics of their chosen deck.
Competition Webpage: http://www.is.ovgu.de/Research/HearthstoneAI.html
Competition Entry Deadline: July15th 2018 23:59 UTC-12
The Ms. Pac-Man Vs Ghost Team Competition
- Piers Williams (University of Essex, UK; email@example.com),
- Simon M Lucas (Queen Mary University of London, UK; firstname.lastname@example.org)
- and Diego Perez-Liebana (Queen Mary University of London, UK; email@example.com)
Description: The aim of this competition is to investigate co-operation in a fairly complex environment. This competition is a revival of the previous Ms Pac-Man versus Ghost Team competition that ran for many successful years. The previous two competition tracks are being altered into two different tracks. The first track in the new competition will ask competitors to submit controllers for Ms Pac-Man operating under a Partial Observability constraint. The second track will ask competitors to submit 4 controllers to control a ghost each under Partial Observability constraints. The game will enable controlled communication to allow co-operation without dictatorship control. Adding Partial Observability to the game forces the ghosts to co-operate and communicate in order to reach their full potential against Ms Pac-Man.
Tracks: Two tracks are available:
1-track: Ms. Pac-Man Track
2-track: Ghost Team Track
Competition Webpage: http://www.pacmanvghosts.co.uk
Fighting Game AI Competition
- Ruck Thawonmas (Ritsumeikan University, Japan; firstname.lastname@example.org)
Description: What are promising techniques to develop fighting-game AIs whose performances are robust against a variety of settings? As the platform, Java-based FightingICE is used that also supports Python programming and development of visual-based AIs. Two leagues (Standard and Speedrunning) are associated to each of the three character types: Zen, Garnet, and Lud (data unknown in advance). Standard League considers the winner of a round as the one with the HP above zero at the time its opponent (another entry AI)’s HP has reached zero. In Speedrunning League, the league winner of a given character type is the AI with the shortest average time to beat our sample MCTS AI. The competition winner is decided based on the 2015 Formula-1 scoring system.
Competition Webpage: http://www.ice.ci.ritsumei.ac.jp/~ftgaic/
- Santiago Ontañon (Drexel University, US; email@example.com)
Description: The microRTS competition has been created to motivate research in the basic research questions underlying the development of AI for RTS games, while minimizing the amount of engineering required to participate. Also, a key difference with respect to the StarCraft competition is that the AIs have access to a “forward model” (i.e., a simulator), with which they can simulate the effect of actions or plans, thus allowing for planning and game-tree search techniques to be developed easily.
Competition Webpage: https://sites.google.com/site/micrortsaicompetition/home
- microRTS setup:
- Creating a bot:
- Joseph Walton-Rivers (University of Essex, UK; firstname.lastname@example.org)
Description: Write an agent capable of playing the cooperative partially observable card game Hanabi. Agents are written in Java and submitted via our online submission system. In Hanabi, agents cannot see their own cards but can see the other agent’s cards. On their turn, agents can either choose to play a card from their hard, discard a card from their hand or spend an information token to tell another player about a feature (rank or suit) of the cards they have. The players must try to play cards for each suit in rank order. If the group makes 3 errors when executing play actions the game is over. Agents will be paired with either copies of their own agent or a set of unknown agents. The winner is the agent that achieves the highest score over a set of unknown deck orderings.
Tracks: Two tracks:
Mixed-track: Agents will play with a set of unknown agents.
Mirror-track: Agents will play with copies of the submitted agent.
Competition Webpage: hanabi.aiclash.com
Submission Server: https://comp.fossgalaxy.com/competitions/t/11
StarCraft AI Competition
- Kyung-Joong Kim (Sejong University, Republic of Korea; email@example.com),
- Seonghun Yoon (Sejong University, Republic of Korea; firstname.lastname@example.org)
Description: IEEE CIG StarCraft competitions have seen quite some progress in the development and evolution of new StarCraft bots. For the evolution of the bots, participants used various approaches for making AI bots and it has fertilized game AI and methods such as HMM, Bayesian model, CBR, Potential fields, and reinforcement learning. However, it is still quite challenging to develop AI for the game because it should handle a number of units and buildings while considering resource management and high-level tactics. The purpose of this competition is developing RTS game AI and solving challenging issues on RTS game AI such as uncertainty, real-time process, managing units. Participants are submitting the bots using BWAPI to play 1v1 StarCraft matches.
Competition Webpage: http://cilab.sejong.ac.kr/sc_competition2018
The General Video Game AI Competition – Learning Track
- Jialin Liu (email@example.com), Southern University of Science and Technology, China; Queen Mary University of London, UK
- Ruben Rodriguez Torrado (firstname.lastname@example.org), New York University, US
- Philip Bontrager (email@example.com), New York University, US
- Julian Togelius (firstname.lastname@example.org), New York University, US
- Diego Perez-Liebana (email@example.com), Queen Mary University of London, UK; and
- Simon M. Lucas (firstname.lastname@example.org), Queen Mary University of London, UK
Description: The GVG-AI Competition explores the problem of creating controllers for general video game playing. How would you create a single agent that is able to play any game it is given? Could you program an agent that is able to play a wide variety of games, without knowing which games are to be played and without a forward model? Participants are also encouraged to submit papers about this competition to the main conference.
Competition Webpage: www.gvgai.net
3rd Angry Birds Level Generation Competition
- Matthew Stephenson (Australian National University; email@example.com),
- Jochen Renz (Australian National University, firstname.lastname@example.org),
- Lucas Ferreira (UC Santa Cruz, email@example.com)
- and Julian Togelius (New York University, US; firstname.lastname@example.org)
Description: This year we will run our third Angry Birds Level Generation Competition. The goal of this competition is to build computer programs that can automatically create fun and challenging Angry Birds levels. The difficulty of this competition compared to similar competitions is that the generated levels must be stable under gravity, robust in the sense that a single action should not destroy large parts of the generated structure, and most importantly, the levels should be fun to play and challenging, that is, difficult but solvable. Participants will be able to ensure solvability and difficulty of their levels by using open source Angry Birds AI agents that were developed for the Angry Birds AI competition. This competition will evaluate each level generator based on the overall fun or enjoyment factor of the levels it creates. Aside from the main prize for “most enjoyable levels”, two additional prizes for “most creative levels” and “most challenging levels” will also be awarded. This evaluation will be done by an impartial panel of judges. restrictions will be placed on what objects can be used in the generated levels (in order to prevent pre-generation of levels). We will generate 100 levels for each submitted generator and randomly select a fraction of those for the competition. There will be a penalty if levels are too similar. Each entrant will be evaluated for all prizes. More details on the competition rules and can be found on the competition website aibirds.org. The competition will be based on the physics game implementation “Science Birds” by Lucas Ferreira using Unity3D.
Competition Webpage: https://aibirds.org (last year: https://aibirds.org/other-events/level-generation-competition.html)
The Text-Based Adventure AI Competition
- Timothy Atkinson (University of York, UK; email@example.com),
- Hendrik Baier (University of York, UK; firstname.lastname@example.org),
- Jerry Swan (University of York, UK; email@example.com)
Description: Before the widespread availability of graphics displays, text-based games such as Colossal Cave Adventure and Zork were popular in the adventure and role-playing gaming community. Due to the richness of natural language text, such games with text-only interfaces offer a useful testbed for AI research. Building a fully autonomous agent for an arbitrary text adventure game is AI complete. However, we test on a variety of different games with graded numerical scores, allowing competitors to gradually increase the sophistication of their approach to handle increasingly complex challenges. We believe that our competition can foster research into fields such as natural language processing and automatic model acquisition, as well as shed light on the relative merits of model-based and model-free approaches Entrants will submit agents to play games for the classic text adventuring engine, the Z-Machine. The competition will be scored according to two independent criteria:
C1: Score on an unseen game instance (objective, built-in to the instance)
C2: Freedom from a priori bias (subjective decision by the judges)
C1 is the dominant criterion, with C2 deciding in the event of a tie. C2 is intended to favour agents that have no (or less) knowledge of the problem domain built in.
Visual Doom AI Competition
- Marek Wydmuch (Poznan Univ. of Technology, Poland; firstname.lastname@example.org),
- Michał Kempka (Poznan Univ. of Technology, Poland; email@example.com), and
- Wojciech Jaśkowski (NNAISENSE, Switzerland)
- Sharada P. Mohanty (École polytechnique fédérale de Lausanne, Switzerland)
Description: The participants of the Visual Doom AI competition are supposed to prepare a controller that plays Doom from pixels. Our ViZDoom framework gives a real-time access to the screen buffer as the only information the agent can base its decision on. The winner of the competition will be determined by a multiplayer deathmatch tournament. Although the participants are allowed to use any technique to develop a controller, the design and efficiency of ViZDoom allows and encourages to use machine learning methods such as reinforcement deep learning.
- Track 1: https://www.crowdai.org/challenges/visual-doom-ai-competition-2018-track-1
- Track 2: https://www.crowdai.org/challenges/visual-doom-ai-competition-2018-track-2
Competition Webpage: http://vizdoom.cs.put.edu.pl/