Article
files/splash/LINEUP

Game AI Roundup Week #4 2008: 8 Stories, 1 Source Code, 1 Video

Alex J. Champandard on January 26, 2008

This Saturday at AiGameDev.com, there are quite a few insightful Smart Links from the blogosphere, in particular about upcoming games. Be sure to contact me if you have any news or tips for next week.

Remember there’s a mini-blog over at news.AiGameDev.com (RSS) with game AI news from the web as it happens.

Game AI Roundup

When Complex Physics Turn Evil…

The developers behind Penumbra: Black Plague discuss the development of the game and its in particular its physics. Here’s the part about the AI:

“As if things where not complicated enough, once we started to add AI to the world, it got even more problematic. The problem with normal AI routines is that they assume the world is static, which is certainly not the case in Penumbra: Black Plague where almost every object can be moved. This meant that a route the AI had picked out could be blocked at any time and become invalid. To solve this we made the AI constantly check if the path was clear and if it was not, have them update their paths to go around the obstacle. Also, the AI needed to have routines to see if a blocked path could perhaps be cleared (by pushing or kicking and breaking) instead of going around, which took a lot of tweaking to get right. In the end we had to tweak all world objects (again) and make sure that the AI would handle them correctly. Countless of hours have been put into checking so that the enemies behave intelligently in the many situations that can arise.”

Developers Discuss Turok’s AI

There’s another video from inside Propaganda Games, the developers of Turok. This one covers the sandbox AI for both the dinosaurs and the human soldiers.

Reinforcement Learning in Ms. Pacman

RL is becoming increasingly widespread in academia, and this technology is slowly starting to spread into game AI sand middleware. Here’s another example with Pac Man:

“The researchers had agents play 50 games using different RL methods. They found that methods utilizing the cross-entropy policies performed better than methods that were hand-crafted. As they explained, the basic idea of cross-entropy is that it selects the most successful actions, and modifies the distribution of actions to become more peaked around these selected actions.”

Protothreads - Lightweight, Stackless Threads in C

Here’s a library that hit the limelight this week that’s ideal for games. If you need the benefits of a cooperative scheduled game but don’t want the complexity of a scripting language, here’s just what you nead. Microthreads in C with only 2 bytes of overhead per hit! Here’s what the code looks like for I/O:

Discussion 0 Comments

If you'd like to add a comment or question on this page, simply log-in to the site. You can create an account from the sign-up page if necessary... It takes less than a minute!