The nervous system of animals transforms dynamically changing sensory information from the environment into appropriate behavioral responses. In particular, olfactory information plays critical roles in adaptive behaviors in the form of long- and short-range chemical cues that encode spatiotemporal information and chemical identity. To elucidate the neuronal mechanisms underlying olfactory behavior, it is desirable to quantify behaviors and neural circuit activities under realistic olfactory stimulus. However, reproducing realistic spatiotemporal patterns in odor concentrations is challenging due to diffusion, turbulent flow, and measurability of odor signals. We have developed an integrated microscope system that produces a virtual odor environment to quantify behaviors and neural circuit activities of the nematode C elegans. In this system, C elegans is maintained in the view field of a calcium imaging microscope by an auto-tracking stage using a pattern-matching algorithm. Simultaneously, odor stimulus is controlled with sub-second and sub-µM precision to reproduce realistic temporal patterns. Using this system, we have found that two types of sensory neurons, ASH and AWB play significant roles in decision making for navigation in a repulsive odor gradient. Calcium imaging and optogenetic analysis revealed that temporal increments of repulsive odor activate ASH to trigger turns that randomize the migratory direction, while temporal decrements of the odor activate AWB to choose proper direction for straight migration. Further mathematical analysis indicated that these sensory neurons are not only antagonizing, but also cooperate with each other by responding to stimulation changes at different time scales. Using this method will lead to comprehensive understanding of cellular mechanisms of decision making in a simple neural circuit.