Evolutionary Hidden Markov Modeling for Dynamic Agent Systems

An evolutionary approach is proposed to model dynamic agent systems using Hidden Markov Models(HMMs).An enhanced genetic algorithm is used to automatically learn the structure and parameters of the HMM,and the final HMM can represent the agent's behaviors by segmenting its environment with an appropriate manner.Experiments indicate that the new method is good at searching the global model parameter space of HMMs,it outperforms conventional optimal HMM topology design methods which has a tendency to stagnate on local optima and usually requires a priori knowledge from a field expert.