Game09 at Imperial College

An interesting day looking at some new techniques in games software, pushing the angle of how a clever department at Imperial College in London, can help game studios with their research and production. It was a smaller group than I was expecting but it had an interesting group of talks, with some stimulating conversations around some of the research areas that the college students are involved with. I’ve been talking with contacts at other universities and they do seem to have the time to look into areas whereas a commercial company has to keep more tightly to their production plan. The computer department here has a lot of AI studies, which reflected in some of the talks.

Mark Morris of Introversion led off with a talk about how to use students from Imperial or indeed, any other college. There are many clever guys around, but they need to be given something that is tuned in with their area of study so choose a department that has areas that may bering something new to your project; to make it bigger or better quality. They could speed up development, do some research, add some tools and they could allow you to bring in some government or sponsored funding from areas that you may not have thought about. It’s about collaboration, but not reliance on the outcome. There may be some great code or theory comes from the project, but sometimes academics need more time to think so don’t rely on having a particular solution. Have a backup plan as well. If you do get something useful into the game, have alook at what could be done for a sequel, by extending the research for another period. Introversion themselves have been looking into auto generation of cityscapes, taking the ideas from automatic evolution of art to the modelling world with generation of 3D buildings and cities.

Introversion also gave a talk/demo about their api for their Defcon game. They’re having a competition to build bots for the game but only seem to have a couple of players so far. Interesting for someone into bots and AI – see Robin Baumgarten’s page or the mailing list for details.

Simon Colton, a professor at Imperial talked through some of the AI areas that they are looking at; the Computational Creativity group have a good site showing their research areas. A lot of it seems to around the AI aspects and they are looking at how to collect data that will allow the games to adapt to the player. This could make it easier for beginners and harder for more experienced gamers. They are looking at the different styles of players, some bloodthirsty, some tourists, some with other desires. By adapting to these styles that game can become more of what that player would like. The department have also been writng software to create new board games based on what the AI thinks the players enjoy in the current games. Some of these theories will fall away, some will be interesting to follow up on, but all of them seemed to generate further ideas and good areas of possible investigation.

Mark Baker from the Blackrock Studio in Brighton (there seems to be an awful lot of creative people there at the moment, Flash especially!) gave some insights into the complexities of developing a major title. They developed Pure, the off road racer, and are due to finish SplitSecond which looks to be a cool track racing game with a twist. Their main problems are the build process for each revision can take hours. This is not just a code build though. The code takes an hour or so. The graphics take the rest. There are so many grpahics in these huge titles that they need to develop special tools to create the assets and variations for each platform. It’s an immense task and getting bigger. Players want more realism (or Hollywood’s version of it), more variations, more video. They needed help with the sort of techniques where they compile the graphics, but also looking at ideas where they could procedurally create some of the graphics. Evolution from simple procedural textures to procedural techniques elsewhere. A few laughs from the fact that they use Jam and Scon for builds (c and python), but they’re also looking at older business tools to help. (Perforce was mentioned – haven’t looked at that)

Another professor from Imperial, Murray Shanahan then gave a view of their Cognitive Robotics developments. They have a very cool, but extremely expensive robot to play with, from a European funding programme. They seem to be trying to create a thinking maching by using GPUs to create neural networks. The GPU chips seem to give the easies way to parallelize this and they have looked at up to 40,000 neurons in real time, with 1000 synapses each. Impressive looking, but it will take some time to find some results. They are not halfway to the brain power of a bee, but again there can be spin offs in other areas. An interesting aspect of one example was that they were using PhysX to rehearse the actions that might occur for many robot actions. By looking at all the theoretical outcomes, the system can then tell the robot to do the action that should bring the best result and the result in the real world can then be measured.

Fred Hasson from RedBedlam (another team from Brighton) talked about the gaming market place. The figures are now showing the Western Europe is overtaking the size of the US market. It increased by 42% in 2008; somewhat less this year no doubt. The problem to solve is that of getting someone to pay for the developments. Three out of ten games make money and they have to pay for the others. They are waitng for the new round of consoles for the next wave of big funding. Mobile is getting bigger as ar virtual worlds. Social aspects are everywhere and having a game that covers more than one platform could be a real advantage. There are increasing amounts of in game economics, whether this is advertising, or buying stuff to increase status, buying services from other players (cf SecondLife or Forza racing). RedBedlam is providing their Vertiverse system free so that people can build this type of virtual world. They make money from their Roman Britain world and are now hoping to spread their system more widely and make money from the business engine behind it.

Dave Taylor showed us the medical side of media at Imperial. They are using Second Life to create virtual doctor’s surgeries and have seen an inreasing satisfaction of people visiting them. Home surgery coming soon…?

Kristof Beets from Imagination talked about the latest in their chip designs. The iPhone has their chip in it although Apple is reportedly buying another chip company to have more production control. Their PowerVR chips are now OpenGL 2.0 compliant and have achieved the Khronos conformance for Collada models etc. As well as openGL, they quoted openVG and now openCL on the slides. I should look at these; I’m not up with these latest terms for vector graphics and general computing. openCL or open computing language is what perople are using to do non graphic processing on these chips. They can be immensely powerful for tasks that are able to be split into parallel processes. Image processing where each pixel can be done independently is the one that I am interested in, or ray tracing even. Kristof gave a couple of developer links for the arm cortex chip. The site is where the SDK information can be found. If you’d like a test board to play on, the BeagleBoard is the one that most people are using. He showed the software running a 3D scene on the latest Samsung phone. It looked impressive with all the latest types of filers going on, blurs, warps, refractions etc. Very interesting.

Chris Sweetman from SplashDamage gave an talk on the audio side of games and how you can use sound to aid the players interaction. They’ve done a lot with id software and are currently working on Brink, a title for all platforms.

Finally Bruno Nicolletti was there from The Foundry to show the special effects from some of their recent work. Originally producing plugin software, they took over the Nuke compositing software from Digital Domain a number of years ago. It is much more robust now. Perhaps I’ll give it another test, rather than Fusion/AfterEffects. The output was tremendous. If you’re into compositing, they have some good videos from the Masterclass that they held in January. I wish I’d seen that at the time.

it was an interesting day, showing that there is some talent around that could easily (maybe cheaply) used to enhance a gaming team and the final game. I sat next to a PhD student who was sponsored by an F1 team to create an AI racing driver. Fantastic project for both student and company, I’d imagine. Well worth the visit.


~ by ianm on June 1, 2009.

Leave a Reply

Please log in using one of these methods to post your comment: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: