An extended agent model embedded with "intelligent-deliberation" for improved performances in a container terminal application

Abstract

The lack of learning competencies and difficulties in dealing with vague or imprecise data sets in the environment are the main obstacles to finding an optimal solution in the present belief-desire-intention (BDI) model. We present a new "intelligent-Deliberation" process in the hybrid belief-desire-intention (h-BD[I]) architecture that enables improved decision making features in a dynamic, and complex environment. Observation and prediction of future effects and the results of the previous plan executions are analyzed in the intention reconsideration process of our model. The forward thinking ability of the agent is improved with the introduction of Temporal Difference (TD) learning in reinforcement learning. An Adaptive Neuro Fuzzy Inference System (ANFIS) is proposed for improved decision making in the intention reconsideration process. A modified version called the Shared Learning Vector Quantization (SLVQ) of the existing neural network based learning vector quantification algorithm has been proposed to handle the agent deliberation called. The paper also discusses the improved behaviour of the agent deliberation process with the introduction of the SLVQ.