It’s the age old question: what do you do when you’ve sold your technology upstart to a bigger company? Do you live the life of the nouveau riche just off the coast of Monte Carlo, or do you do it all again: put your (new) money where your mouth is and start from the bottom once more?
For Dr. Paul Kruszewski, it was the latter course of action. Having set up BioGraphic and nurtured the growth of the AI.implant middleware technology, he sold the firm to Engenuity in 2005.
“I’ve been working on game AI since 2000 and in that time have seen a lot of progress in the field of AI, but after finishing AI.implant, it became clear to me that pathfinding was only the tip of the iceberg and there was so much more to do,” he says.
“One of the key reasons for the success of AI.implant was the working relationship that co-founder Aaron Davey and myself created while at Engenuity, the company to whom I sold BioGraphic. While there, I was in charge of all outside stuff – sales, relationships – and Aaron was in charge of all inside stuff, such as the dev team, product management; and that worked well. Grip was another opportunity to work with Aaron – I bring in crazy ideas, razor sharp programmers and AAA game contracts, while Aaron and his team leverage our technology to blow away our customers.”
The new offering is, in fact, two separate products: the Grip Digital Extra System and Character Control System.
The Digital Extra System is, as the name somewhat suggests, dedicated to introducing the movie and television concept of ‘extras’ into games. It takes the two primary benefits of film extras – their cheapness compared to full actors, and their ability to be directed as a group rather than as individuals – and transposes those into the game space; meaning that designers can quickly create crowds that are controllable as a whole, and that take up a fraction of resources compared to fully-fledged NPCs.
“Numerous attempts have been made to dumb down NPCs and make them work for crowds,” explains Kruszewski, “but this is pretty much impossible because the sub-systems of an NPC –physics, animation, navigation, and so on – can’t be easily extricated and optimised. The Grip approach was to build a new character type from scratch only including the key elements of background characters. The result was the Digital Extra System.
“Of course, there are games that have already invested in creating secondary characters, but my question is: how many great games were never designed or produced because creating a living world was either too hard or expensive? Most of us live in cities but games are rarely set in them.”
The Character Control System, on the other hand, is a visual way of authoring behaviour trees that aims to bridge the divide between designer and programmer when working on AI.
“Traditionally, the game designer will write up a description of an NPC which the AI programmer then reads and implements in code,” Kruszewski says.
“Once done, he gives it back to the designer who plays the NPC which inevitably isn’t what he envisioned, so the designer and programmer meet, pictures are drawn, and they try again. And again, and again. The issue is that they lack a common AI language to define behavior.
“With CCS, the programmer uses the behaviour tree system as a way to organise the flow of the NPC’s AI. Because the tree can be visualised and its control flow is easy to read, the designer can sit down with the programmer and visually walk through the tree. This allows the designer to catch his own inevitable gaps in logic long before any code is laid down. Also because a tree is made up of small atomic units, the large problem of developing an entire NPC AI is made more manageable.”
As with all of Grip’s products, the system was designed within the context of a game’s development in order to ensure that it fulfils a need. With the Digital Extra System it was EA Montreal’s Army of Two: The 40th Day – set in a virtual Shanghai that needed populating. For the Character Control System, it’s Intel’s Project Offset and Eidos Montreal’s Deus Ex 3 that is proving the incubator for the tech.
“Every technical decision Grip makes is based on how game developers actually build games and the kind of AI they are trying to achieve,” asserts Kruszewski, “and so working directly on these games helps us incredibly.”