Swarm reasoning

This paper presents a new technique, the hypothesis swarm problem-solving technique (HSPT), that allows swarms to operate on more complex problems in a general way that produces solutions faster than traditional swarms by emerging compatible sub-solutions cooperatively. Instead of agents simply reacting to their own stimuli, they produce hypotheses about their environment to resolve discovered conflicts. As they gather supporting or refuting evidence for their hypothesis, we force a higher level of social interaction - requiring them to compare their evidence with other agents. They reward compatible hypotheses and devalue conflicting ones, encouraging acceptance of mutually supporting hypotheses. We demonstrate the effectiveness of this approach on graph-coloring and round-robin scheduling problems and also compare its efficiency with other swarm techniques including brute force, reactive swarms and ant colony optimization algorithms.