Playable Experiences
Indie developers, industry developers, and researchers who are developing innovative AI-based games or other interactive media (“playable experiences”) are invited to submit their work to the playable experiences track. We welcome playable experiences that involve some articulable innovation in the use of AI that directly affects the user's experience. This includes novel game designs that leverage existing AI techniques, as well as innovations in the techniques themselves that lead to new kinds of playable experiences. Playable experience submissions should be sufficiently complete and polished enough for naïve users to play them. Authors will have the opportunity to show their games during the evening poster/demo session of the AIIDE conference.
Accepted systems
A top-down sneaking game by Justus Robertson in which the world is generated from a plan-based experience manager, where the player directly manipulates the planner’s underlying state via in-game mechanisms
An interactive narrative by Sergio Poo Hernandez and Vadim Bulitko in which the player influences an AI experience manager and branching narrative choices by assuming dance positions perceived by the Microsoft Kinect
A unique commercial turn-based digital strategy card game in development by Lunarch Studios, submitted by lead AI programmer Dave Churchill, featuring sophisticated AI opponents navigating a large state-space
A puzzle-platformer by Martin Cerny with a cooperative AI sidekick that telegraphs its search state in-game, set in a problem space designed to highlight and simplify AI search while creating perceived complexity for the human player
An infinite runner building on rhythm-based approaches to real-time procedural level design
Justus Robertson, R. Michael Young
Base Case is a top-down sneaking game in which the player must free captured comrades from a military compound while avoiding detection by guard patrols. The game is guided by a plan-based experience manager’s declarative state transition system that drives the configuration of the procedurally-generated game world as well as the behavior of the game’s NPCs. The planner creates each game session from one of several pre-authored initial problems described using the Planning Domain Description Language (PDDL). The player then navigates the base, looking for a computer console while avoiding guards. Once the console is discovered, it can be used to directly inspect and modify the underlying declarative state and open the final exit through which the player and comrade may escape. The game repeats with variations on a given session until the player achieves this objective.
Download: http://justusrobertson.com/BaseCase.zip
Sergio Poo Hernandez, Vadim Bulitko
iGiselle is an AI-managed narrative inspired by the Romantic ballet. In iGiselle, the player takes control of Giselle, a young ballerina, and experiences an interactive narrative through a series of still images, voice-overs, and music. To further immerse the player in the world of ballet, a traditional game controller was eschewed in favor of having the player indicate his/her narrative choices by assuming dance positions which are perceived by a Microsoft Kinect connected to a PC. The game utilizes PACE (Player Appraisal Controlling Emotions) as its AI experience manager. PACE determines the next narrative event to show to the player in an attempt to keep him/her on an author-specified target emotional curve. PACE models the player's emotions by determining the player's playstyle from the actions he/she has taken in the game so far. Candidate narrative events that PACE selects from are computed automatically by the Fast Downward AI planner as the narrative domain is encoded in the Planning Domain Description Language (PDDL).
Download: https://www.dropbox.com/sh/k1gm9u2uruqu6n2/AAAzgtNcF3ZIlKm3ZZv1N5roa?dl=0
David Churchill
Prismata is a fast-paced hybrid strategy game from Lunarch Studios which combines elements from real-time strategy games, collectible card games, and tabletop games. In Prismata players build up an economy, spend resources to buy armies, and unleash attacks on their enemies while simultaneously defending incoming enemy barrages. Think ''turn-based StarCraft'', but without a map, or Hearthstone with workers and build orders instead of decks. Games in Prismata last just a few minutes, and have infinite replay value due to the randomly selected units up for purchase at the start of a game. Each game there are new units to construct, new build orders to discover, and entirely new strategies to unleash on your opponent. Several challenges are faced when creating AI systems for modern online strategy games like Prismata:
* Strategy games often have enormous action spaces, with millions of possible action combinations to consider on any given turn.
* Different difficulty settings must be offered so that players of all skill levels can play against and enjoy their experience with the AI.
* As new units are frequently added to the game, the AI system must be robust enough to handle design and balance changes made to the game over time.
To take on these challenges, the Prismata AI system uses Hierarchical Portfolio Search (HPS), a new search technique developed by the presenter, David Churchill, which is described in the accompanying AIIDE 2015 publication: ''Hierarchical Portfolio Search: Prismata's robust AI architecture for games with large search space''. HPS offers a modular and generic approach to dealing with abstract games with large action spaces, and is especially powerful in strategy games like Prismata or real-time strategy games like Starcraft, where turns have multiple actions which can be broken down into tactical categories. Another benefit of HPS to game designers is that the modular nature of the search system allows for complex strategic behaviours to be constructed from smaller tactical pieces. These modular configurations allow the AI to have multiple difficulty settings and strategic play styles, which are easily created and tuned within seconds. The AI bots in Prismata have many difficulty settings, ranging from the virtual punching bag all the way up to experienced player level, so that anyone can play the game and have fun.
Play: http://play.prismata.net
Martin Cerny
Sarah & Sally is an experiment in designing around the problems inherent in cooperative AI development: it is a cooperative 2D puzzle-platformer that looks similar to mainstream examples of the genre, but allows for an easy implementation of a quality sidekick AI. The design of the game allows the AI to find optimal solutions using a straightforward search-based approach while the problem remains relatively hard for a human player. In this cooperative puzzle-platformer, the player switches between moving and allowing the AI sidekick to move, with the objective of navigating both characters to their end goal points. This task is further complicated by the fact that both characters are affected by gravity, and that the game world can be reoriented by the activation of in-game levers.
The result is a puzzle game in which the problem space has a much higher apparent complexity to the human player yet remains solvable to the AI. Additionally, the system’s dynamic planning is exposed to the human player via in-fiction speech bubbles from the AI sidekick, which acts as the game’s built-in (light-handed) hint system while providing rationale for the sidekick’s actions, making her seem both helpful and intelligent.
Tommy Thompson, James Tatum, Neall Dewsbury
Sure Footing is an infinite runner game currently in development by Table Flip Games Ltd and is programmed by Neall Dewsbury, James Tatum and Tommy Thompson. Players take the role of one of several characters known as 'pixellites', each with its own unique skills, who are attempting to escape from an evil force chasing them known as 'The Deletion Wave'. Players are tasked with navigating a series of increasing more challenging platforming environments that are procedurally generated at runtime. The game is broken up into segments known as sprints, separated by prefabricated chunks of environment (or “prefabs”) that are designed to look like empty streets. When the player reaches a new street segment, the next batch of platforms is built before it comes within the player's view. The generation process is largely inspired by existing work in the adoption of rhythm for the creation of gameplay segments. Each sprint is built in two phases: a grammar-driven action generation, followed by the construction of geometry. Each of these phases is reliant upon a budget, which allows for designers to constrain the expressiveness of the system at a given point: the former constrains the number of activities the player will expect to complete in a sprint, while the latter dictates the difficulty of said activity given how it is built in the game world. Constraints are in place that prevent particular activities appearing within close proximity to one another, or with a given frequency. Furthermore, each activity carries a cost, with the action generator responsible for minimizing costs of activities to fit within the assigned budget.
With a complete action sequence prepared for a given sprint, the geometry generator selects from one of multiple prefabs that embody this particular activity. The selected platform types, their length and the potential difficulty they represent, dictate the cost of the prefab. Given a particular budget, the geometry generator will aim to utilize it as best it can. In addition, upon placing particular prefabs, the game adds obstacles or collectables onto platforms.
Play: http://1drv.ms/1CQcA6h
Justus Robertson, R. Michael Young
North Carolina State University
fjjrobert@ncsu.edu, youngg@ncsu.edu
Sergio Poo Hernandez, Vadim Bulitko
University of Alberta
pooherna@ualberta.ca
David Churchill
Lunarch Studios / University of Alberta
dave.churchill@gmail.com
Martin Cerny
Charles University in Prague
cerny.m@gmail.com
Tommy Thompson, James Tatum, Neall Dewsbury
Table Flip Games / University of Derby
tommy@t2thompson.com