Adaptive multi-agent system development by quantifying contribution of agents

Contribution of agents with respect to teamwork has been introduced in the paper for solving problems, when very little or almost no information is available about the environment and the agents must therefore act under uncertainty. The agents are implemented as objects and completely specified by their behavior. A semi-autonomous modular approach is proposed by developing a simulation program that interacts with the environment and with the domain experts to formalize the problem into a hierarchical tree structure. An agent's role to solve the problem has been determined considering its specialization while its claim to contribute is quantified using probability and belief measures. The agents represented as objects try to improve their qualification by adapting knowledge from others or environment and thus, contribute more towards finding solution of the problem. Execution of tasks and adaptation of knowledge exploits the essence of pipeline processing, so provide better system performance with less computational time.