Abstract
This thesis proposes a non-hierarchical hybrid cognitive agent architecture, as an alternative to largely used hierarchically layered architectures which assume a predefined hierarchy of cognitive processes and, consequently, a fixed hierarchical control framework. As opposite to hierarchical view, the non-hierarchical approach considers that all cognitive processes coexist at the same level, but carry different weights in different environmental contexts; hence, a dynamic control framework.
The non-hierarchical agent architecture is based on “Society of Mind” approach on human mind and cognition, and is intended to demonstrate that such an approach can be successfully used for modelling complex cognitive processes and a wide range of behaviours.
In order to demonstrate its viability, the proposed architecture is evaluated in a cognitively demanding environment as a driver agent in traffic behaviour context. The thesis demonstrates that the resultant SoM driver agent can be successfully used as an investigation tool for a variety of road transportation issues. First, it is used in an individual setup for modelling microscopic behaviour-enabled car-following situations. Then, it is used in a multi-agent setup for modelling populations of drivers with various driving habits in order to understand how their behavioural pattern influences the traffic performance. Also, in a wider view, the thesis demonstrates that behavioural capabilities of the SoM driver agent make possible the investigation of more general transportation issues, such as the influence of human behaviour on resilience of transport systems in certain geographical areas. In a field in which cognitive-affective and behavioural issues are still treated separately from the infrastructure related aspects, a tool that is able to treat them simultaneously offers a great deal of possible benefits to the field. This thesis demonstrates that a driver agent of SoM type can become such a tool by providing the requested capabilities.