Very fast action selection for parameterized behaviors

In this paper, I will discuss a set of techniques for supporting limited variable binding in behavior-based systems. This adds additional useful expressivity while preserving the property of selecting actions in 1 time and space. An unusual feature of the method is that it allows inference operations to be reduced to bit-parallel and short-vector operations, making it particularly well suited to real-time operation on modern superscalar and SIMD architectures. I also describe an implementation of the technique in the Twig procedural animation system. Categories and Subject Descriptors I.2.m [Artificial Intelligence]: Miscellaneous General Terms Algorithms Keywords Behavior-based control; agent architectures; reactive planning.

• Type: Conference
• Year: 2009
• Where: FDG
• Authors: Ian Horswill
• Keywords: Additional Useful Expressivity, Computer Graphics, FDG 2009, Keywords Behavior-based Control, Limited Variable Binding