The neuronal mechanisms of adaptive behavior in insect lead us to expect how to possess a simple adaptation mechanism. This research focuses on behavior selection of crickets. Male crickets start fighting when they come across others. The cuticular pheromone is a trigger stimulus of this behavior. Once a cricket is beaten by the opponent, the loser wouldn’t fight anymore against the opponent. It is demonstrated that stimulation of male cuticular substances on antenna increase generation of nitric oxide (NO) in the antennal lobe of the cricket. Inhibition of NO cascade increases aggressiveness in loser cricket. Based on available biological results, we propose network models. From recent researches, multi-systems have been revealed to possess various structures with regard to functionality, for example, a small-world structure realizes effective communication in actual phenomena. Thus, network transformation causes a transition in the system functionality. We here develop a model of an antennal lobe by using an oscillator network that changes behavior according to its structure, and realize the model by employing an analog circuit comprising coupled van der Pol oscillators. Subsequently, we propose a robot that autonomously selects its behavior. The model is realized using an analog circuit and it is applied to a simple robot model that displays the behavior of a real insect.