Learning to Locate an Odour Source with a Mobile Robot
T. Duckett, M. Axelsson, and A. Saffiotti
We address the problem of enabling a mobile robot to locate a stationary
odour source using an electronic nose constructed from gas sensors. On
the hardware side, we use a stereo nose architecture consisting of two
parallel chambers, each containing an identical set of sensors. On the
software side, we use a recurrent artificial neural network to learn the
direction to a stationary source from a time series of sensor readings.
This contrasts with previous approaches, that rely on the existence of a
model of the sensor's dynamics. The complete system is able to orient and
turn towards the source. An experimental validation was carried out to
evaluate the performance of the system.