A computational model for building modular animals: Design and configuration of the decision network

The factors behind the ability of animals to perform complex tasks in a dynamic environment are a fascinating area of research. In order to effectively build adaptable and robust artificial intelligent systems, it is important to fully understand how complex behavior emerges through self-organization and embodiment in the actual animals. Biologically-inspired modeling has emerged as the most effective way of building artificial animals capable of complex behaviors in the recent times, and is the primary research focus in this thesis.;The following hypotheses are the fundamental basis of this work: 1) Artificial intelligent systems built using developmental and/or evolutionary learning techniques are more efficient in performing complex tasks in dynamic environment; 2) Embodiment plays a significant role in emergence of intelligent behavior in complex animals; 3) Modularity is one of the major factors that favors the 'evolvability' of systems; 4) Complex animals arise through systematic reconfiguration and duplication, composition, reorganization and reallocation of modules in simpler modular animals through evolution.;This thesis extends a previously developed framework for configuring and simulating modular animals [1] to add a control structure that receives sensory input and generates neural control signals for the motor system. The framework also allows the possibility of simulating developmental processes to systematically build complex animals. The ultimate objective is to iii develop an environment for evolving animals that would be capable of highly complex and intelligent behavior.