Games

Kynogon markets Kynapse for games the most widely used A.I. middleware within the game development community with clients such as A2M, Activision, Atari, Bethesda Softworks, Blitz, Digital Illusion CE, Electronic Arts, Lionhead Studios, Real Time World, Spark Unlimited, SEGA, Sony Online, THQ, Turbine, etc.

Kynapse has been used in the development of more than 60 of some of the best known game titles including Alone in the Dark 5, Crackdown, Fable 2, Medal of Honor: Airborne, Sacred 2 and The Lord of the Rings Online™: Shadows of Angmar™.

Kynapse for games also exists as integrated into the Unreal Engine 3 developed by Epic Games. Kynogon is a member of Epic's Integrated Partner Program.

Kynapse for games can be licensed for PLAYSTATION®3, PSP™, PlayStation®2, Xbox 360™, Xbox™, Wii™, and GameCube™ consoles as well as PC (Windows and Linux).

It brings innovation (3D dynamic topology analysis, hierarchical 3D dynamic pathfinding, ...) as well as efficient production tools (automatic pathfinding and perception data generator, ...).

Kynapse generic A.I. architecture structures code development, facilitates re-usability and team communication. Kynapse is also fully open for customization with high level performance.

Technical documents

White paper #3 : Large Scale A.I.

New generation consoles (PS3, XBox 360) and recent PCs fundamentally reassess the way A.I. fits into the whole game development process, from initial game designing to final tests. Large Scale A.I. becomes key: A.I. must often simultaneously handle several hundreds or even thousands of NPCs (Non Player Characters) living in complex 3D environments of several dozens of square kilometers. It is obviously not possible to handle such huge worlds and such huge amounts of NPCs with existing technical solutions. Following questions arise:

How could A.I. navigation and 3D perception assets (Topological Data) be produced in a reasonable amount of time ? Due to the huge amount of A.I. assets to be generated, any manual process would be totally irrelevant.

How is it possible to fit megabytes of Topological Data in memory at runtime ? Game developers usually do not dedicate more than 10% of their memory budget to A.I. and this is probably not going to change.

As it is impossible to fit all Topological Data in memory at runtime, how could long paths (joining one corner of a huge environment to the opposite corner for example) be computed ?

AStar (the most popular pathfinding algorithm) is not efficient enough when dealing with huge graphs including hundreds of thousands of vertices and edges). How is it possible to overcome this fundamental limitation ?
How to ensure a good perception for thousands of NPCs at run time ? Thousands of NPCs requesting perception updates on other NPCs will cause massive CPU consumption peaks.

How to take advantage of next generation hardware architecture (PS3, XBox 360) ?

With Kynapse 4.0, Kynogon, the leading A.I. Middleware developer, introduces the “Large Scale A.I.” solution to answer all these problems.

White paper #2: 3D Pathfinding

A game without Non Player Characters (NPCs) moving around is an interesting challenge for a designer ! Moving is the foundation of action: an NPC that cannot move will not pick up objects, attack, hide, implement tactics, etc. Pathfinding, the technology required to move characters, is therefore a must for almost every game. This is probably the reason why game developers have been working on pathfinding for years and why it’s extensively discussed in many articles and conferences. It is unusual to see a development team without pathfinding experience.

On the other hand, it is not unusual to see NPCs still having troubles with their pathfinding. We often see characters blocked, lost, running against a wall, unable to go through a door, bumping into each others, falling into holes, etc.

The paradox probably comes from the fact that pathfinding is not A*. A* or equivalent algorithms are widely used throughout the game development community to solve the key question : how to go from one point of the map to another one ? A* is an old and mature technology, extensively presented in most pathfinding articles and you can download code examples on the web: A* is not complex to implement. One might conclude that pathfinding is therefore easily solved.

A* is the trivial part of a pathfinder however - there is much more to pathfinding than A* : pathfinding is not path-planning. A lot of pending issues outside of path-planning remain :
How to take into account constraints such as furtiveness when computing a path?
How to follow a path computed by the path planner ?
How to take into account other NPCs when following a path ?
How to deal with dynamic evolutions of the world ?
How to ensure pathfinding overall performance

This white paper’s objective is to stress the challenges of traditional pathfinding outside of A* and explains Kynapse’s ability to meet them. These challenges become critical as traditional workarounds will not suffice for next generation games. In order to remain focused, herd pathfinding (flocking), formation, etc. will not be considered.

White paper #1 : Perception

Game developers often perceive A.I. as high level decision making processes with autonomous Non Player Characters (NPC). As a consequence, they are frightened to loose control of their game: NPCs making autonomous decisions will not stay in line with their scenario. Simple decision processes like scripts or Finite State Machines are usually more relevant for games.

On the other hand, 3D perception is not considered as critical. In today’s video games, NPCs have very limited 3D perception capabilities. Ray casting is very often the only way a NPC can understand its environment: NPCs are blind men with a stick.

This white paper’s objective is to stress the importance of perception, its challenges and explains Kynapse’s ability to meet them.