Computational Science Technical Note CSTN-098


Auto-Training Animated Character Motion: A Rule-Base Tuning Hybrid Fuzzy-Genetic Algorithm

A. P. Gerdelan

Archived June 2009


The last ten years have seen animated film and computer game technology evolve to a point where controlling astronomical numbers of animated characters at once is part and parcel of a modern production. These urban crowds, armies, flotillas, and road traffic are no longer pre-calculated or hand-animated. Movement of these characters is dynamic, reactive, intelligent, and in the case of many complex systems occurs in real-time; each character is controlled by an intelligent agent. Because of its fast, efficient, and reactive properties, and its capacity for intelligence, fuzzy logic is an excellent choice of reactive motion control technology for these systems. However, there is a major drawback; each new character that a fuzzy controller is designed for has different specifications - size, speed, and acceleration - and therefore each new character requires a very large amount of manual tweaking of fuzzy set parameters, rules, thresholds, and other connected system parameters before it operates effectively (and realistically). We are interested in developing a generalised fuzzy navigation algorithm for 3D animated characters. To approach this goal we are developing a self-calibrating fuzzy controller using a hybrid fuzzy-genetic algorithm. The new algorithm that we present here is stage 2 of this target algorithm - a genetic fuzzy rule-based system that takes a new character and trains its controlling agent to operate in its intended 3D environment dynamically, in real-time, and in a parallel fashion on one CPU+GPU for very fast fuzzy rule-base tuning.

Keywords: fuzzy navigation; animated character; genetic algorithm; parallel algorithm; GPU.

Full Document Text: PDF version.

Citation Information: BiBTeX database for CSTN Notes.

BiBTeX reference:

  author = {Anton P. Gerdelan},
  title = {Auto-Training Animated Character Motion: A Rule-Base Tuning Hybrid
	Fuzzy-Genetic Algorithm},
  institution = {Computer Science, Massey University},
  year = {2009},
  number = {CSTN-098},
  address = {Albany, North Shore 102-904, Auckland, New Zealand},
  month = {June},
  timestamp = {2009.12.16},
  url = {}

[ CSTN Index | CSTN BiBTeX ]