Computational Intelligence as an Object Tracking Paradigm for Game Agents
Leigh Carey · 2008
Particle Swarm Optimisation is a stochastic, population-based evolutionary algorithm for optimising continuous non-linear functions, based on a model of social optimisation. A swarm of particles iteratively evaluates the fitness of candidate solutions to a given optimisation problem, with each particle remembering the location where it had its best success. Each particle makes its best solution available to its neighbours, while also examining where they have had success. Movements through the search space are guided by these successes, with the population converging on a problem solution. PSO has been applied to the training of artificial neural networks, finite element updating and grammatical evolution, providing robust and efficient optimisation algorithms in environments with incomplete or noisy data. It is therefore a promising object tracking paradigm for computer game agents that are required to make decisions using visual information rather than a more conventional data model. In this paper, I test the hypothesis that a Particle Swarm Optimisation-based