Hamid Haidarian Shahri - Intelligent Systems Laboratory Faculty of Computer Engineering and Information Technology Amirkabir University of Technology (Tehran Polytechnic), Tehran, Iran
Abdolahzadeh Barforush - Intelligent Systems Laboratory Faculty of Computer Engineering and Information Technology Amirkabir University of Technology (Tehran Polytechnic), Tehran, Iran

Distributed search in inaccessible environments is an important problem in robotics and multi-agent systems (MASs). Vacuum cleaner agent is a classic example, which is particularly interesting, since it encompasses many aspects of search problems in MAS. In this paper three different architectures that are complimentary to one another have been studied to determine their suitability for distributed searches. Various interaction protocols for coordinating an agent society were simulated and interestingly competition with blackboard produced the better results. The effect of environment accessibility on performance was also analyzed. It was deduced that in complex environments or preset societies, where it is impossible to change the agent interaction protocols, the same performance can be gained by increasing the environment accessibility for individual agents. The effects of utilizing memory to keep track of agent’s internal state and introducing complexity into the environment have also been studied.