High Performance Computing for Agent-Based Cognitive Modeling.

摘要

The aim of this project is to evaluate a High Performance Computing (HPC) environment for running simulations involving a large number of intelligent agents. Simulations may benefit from separating the environment from the intelligent agents. This could allow for larger scale simulations, and different environments may alter the results of the simulations. A simulation environment was developed for agents to interact within. This environment and agents were tested using a standard computer server, and an attempt was made to use cluster computing resources to run the environment and agents on a larger scale. Difficulties in the account setup process, and technological limitations of the existing cluster environment led to an unsuccessful test on the cluster. The communication method chosen (sockets) for the client-server interactions was not available on the cluster. However, tests in two single-server environments were successful. The outcome of this scenario suggests that more development is needed to address the portability of the communication model used for the client and server. Additionally, account creation procedures for the HPC environment may benefit from a streamlined process that addresses the rapid academic life cycle of student researchers.