Date of Award
Doctor of Philosophy (PhD)
Modeling Simul & Visual Engineering
Ryland Gaskins III
While psychology has shown that perception is very important for the human decision process, agent perception has not been covered in sufficient detail within the agent directed simulation field. To contribute to such a solution, an open challenge lies in capturing the knowledge of human sciences, such as psychology, and making this knowledge usable for engineers. This dissertation addresses perception by describing an experimental method where agent perception simulates human perception. In particular, it presents engineering methods based on accepted psychological approaches resulting in a proof of concept. To prove the feasibility, an Artificial Perception (AP) meta-model is presented using logical assumptions, generalized Validation and Verification (V&V) approaches, cognitive testing, and experimental comparisons with similar state-of-the-art agents.
Current publications show that Human Behavior Representation (HBR) in agent simulations remains a top priority for the modeling and simulation community. Agents are a promising way to represent human behavior. However, there are limited methods for measuring and validating perception within simulated environments. Within a simulated environment, there are no clearly established methods for understanding surrogate HBR perception relationships with their authentic human behavioral representations. These present obstacles for advancing simulation of human behavior. The approach presented in this dissertation provides an engineering solution that allows bridging this gap and configuring agents based on human behavior evaluations using accepted psychological approaches, in particular, applied cognitive task analysis.
In many cases, the foundation for these engineering methods and the effects of agent perception are similar to “experimental frames” with some known potential for accomplishing a task. Effects of these tasks require accurate assessments beyond face validation. Agents play a critical role in HBR's capacity to accomplish specific tasks. An agent, in this sense, becomes less a matter of size or complexity, and more a matter of “viewpoint.” This viewpoint makes it possible to infer an agent-oriented approach for conceptual modeling of AP, thus implying an agent-object relationship.
An experimental approach is provided using proven psychology perception measurements to help answer the question whether there is a correlation between human perception in the real world and its same representation in an artificial world. The results presented in this dissertation show that the method applied in the experimental approach can be generalized. Furthermore, the example used builds a proof of concept, as the resulting agent with AP reproduced the human behavior more closely than did the state-of-the-art approaches.
Garrett, Randall B..
"A Method for Introducing Artificial Perception (AP) to Improve Human Behavior Representation (HBR) Using Agents in Synthetic Environments"
(2009). Doctor of Philosophy (PhD), Dissertation, Modeling Simul & Visual Engineering, Old Dominion University, DOI: 10.25777/nrtd-5d26