Modeling Situated Cogniton in Reactive Robotic Architecture

Abstract

Abstract Robots based on conventional AI architectures often failed to act in environ- ments where the location of the objects in the environment is unknown. This is due to the unpredictability of interactions that are made by the robots within the en- vironment. As an alternative to this approach, situated robotic architectures were introduced. Situated robotic architectures are used in the process of implementing robots in complex and dynamically changing environments. Subsumption architecture, a kind of situated robotic architectures introduced by Rodney A. Brook succeeded in such environments and was well-received by the situated AI research community. This research focuses on enhancing the situated- ness of `subsumption architecture'. Situatedness refers to the fact that to which extent agent's cognitive process is a ected by its environment. In this research, enhancement is based on how the robot's behavior is a ected by its internal and external environmental condition. Fuzzy controller based approach was used to interface selected internal and external environmental variables with the existing subsumption architecture. As a further enhancement, behaviours of subsumption architecture was implemented using neural dynamics. Results of the experiments conclude that fuzzy controller based approach can be used to interface the internal and external environmental parameters to the subsumption architecture. Furthermore, there is signi cant improvement of the reactiveness of robot when behaviours of the robot are implemented using neural dynamics. Keywords: Situatedness, Subsumption Architecture, Neural Dynamics, Fuzzy Controller