Lecture
Design and Integration of Agent-Based Partial Brain Models for
Robotic Systems, by means of Hierarchical Cooperative CoEvolution
Speaker: Michalis Maniadakis (PhD Student)Date: Wednesday, 17nd May 2006,
Time: 16:00 - 18:00
Mediterranean Studies Seminar Room - FORTH
Host: Prof. Panos Trahanias (Head of Computational Vision and Robotics Laboratory)
Abstract:
The current talk addresses the development of cognitive abilities in
artificial organisms, a topic that has attracted many research efforts
recently. We introduce a novel computational framework for modeling
partial brain areas, following a coevolutionary agent-based approach.
Specifically, self-organized agent structures are employed to represent
distinct brain areas. In order to facilitate the design of agents, we
introduce a Hierarchical Collaborative CoEvolutionary (HCCE) approach
that specifies the structural details of autonomous, but cooperating
system components. By utilizing a distributed model and a distributed
design methodology, we are able to explicitly address the special characteristics
of substructures representing brain areas, and additionally integrate
them effectively formulating composite systems.
Overall, the proposed computational framework facilitates the design of brain-inspired cognitive systems because it:
- assigns distinct roles to the agents representing brain-areas,
- supports computational modeling of biological lesion findings, addressing explicitly the pre- and post- lesion performance of the model,
- alleviates the integration of partial models developing gradually more complex ones,
- re-designs (if necessary) existing substructures in order to support forthcoming modeling efforts.
The effectiveness of the proposed computational framework is demonstrated
on a number of different experiments. The implemented models are successfully
embedded in a simulated robotic platform, developing artificial organisms
with advanced behavioral and cognitive abilities.
