Session:

LBP - Late Breaking Papers

Title:

Using GP to Model Contextual Human Behavior - Competitive with Human Modeling Performance

Authors:

Hans Fernlund
Avelino Gonzalez

Abstract:

To create a realistic environment, some simulations require simulated agents with human behavior pattern. Creating such agents with realistic behavior can be a tedious and time consuming work. This paper describes a new approach that automatically builds human behavior models for simulated agents by observing human performance. With an automatic tool that builds human behavioral agents, the development cost and effort could be dramatically reduced. This research synergistically combines Context-Based Reasoning (CxBR), a paradigm especially developed to model tactical human performance within simulated agents, with the Genetic Programming machine learning algorithm able to construct the behavior knowledge in accordance to the CxBR paradigm. This synergistic combination of AI methodologies has resulted in a new algorithm that automatically builds simulated agents with human behavior. This algorithm was exhaustively tested with five different simulated agents created by observing the performance of five humans driving an automobile simulator. The agents show, not only the capabilities to automatically learn and generalize the behavior of the human observed, but they also exhibited a performance that was at least as good as that of agents developed manually by a knowledge engineer.