To realize an autonomous odor source localization robot, we focused on the adaptability of an insect's brain to compensate for rotational disturbances during odor source searching behavior. We manipulated motor outputs to control the sensory feedback of an insect using a brain-machine hybrid system. This system is composed of an insect's head and a two-wheeled mobile robot. The velocity of the robot is proportional to neural activities descending from an insect brain. We successfully manipulated the behavior of the robot. In disturbance experiments, insects responded to given rotational disturbances by modifying their neural activities to make compensative angular velocity. We assumed this control system of the compensation as an output-error model. We calculated the parameters under different motor gains to reveal it as an adaptive controller. We propose that an insect has its appropriate angular velocity during odor source localization, and performed simulation experiments involving an odor source searching agent and odor distribution environment. We calculated the cost for odor source localization by changing the angular velocity of the agent, and found that it had the minimum value.