We analyzed properties of a brain responding to unnatural sensory feedback by using a brain-machine hybrid system. The brain-machine hybrid system is a two-wheeled mobile robot with an insect's head as a controller. On the robot, we recorded neural signals relating to steering behavior and translated them into angular velocity of the robot. Because the hybrid system has sensors on an insect, we can give disturbances to control sensory feedback by manipulating motor output of the robot. By calculating neural activities responding to disturbances, we could observe that an insect cancelled out rotational disturbances. Moreover an insect kept motor output of the hybrid system by modifying neural activities according to conversion rule during odor source localization behavior.