Robotic Experiments in Cricket Phonotaxis

Research Staff:

The LEGO Lab, University of Aarhus (CIT):

Henrik Hautop Lund

Department of Artificial Intelligence, University of Edinburgh, UK:

John Hallam

Department of Psychology, University of Nottingham, UK:

Barbara Webb

Funding by:

UK Engineering and Physical Science Research Council (GR/K78942)

National Danish Research Councils

By making a robot implementation of how female crickets find male crickets by walking (or flying) towards a species specific song that the males produce, we find evidence for our neuroethological hypothesis that the control system required for obtaining such a phonotaxis behaviour can be much simpler than that traditionally hypothesised by biologists. The robot model implements a mechanism for this task that is consistent with neuroethological evidence but substantially simpler than the mechanisms hypothesised by neuroethologists. In particular, ``recognition'' of the species song occurs as a side-effect of the ``location'' device, rather than requiring additional processing.

Initial studies to verify our hypothesis were made with a LEGO robot. To get more control over experimental settings, we are now using the Khepera miniature mobile robot that measures only 55 mm in diameter and has its own on-board controller. We have constructed ears (i.e. microphones with programmable pre-amplifiers, delays and mixers) for the Khepera robot, so that we can perform the robotic experiments in cricket phonotaxis much more precisely with real cricket sounds that we have recorded and are playing through a host computer. In order to study female choice behaviour in crickets we put more sound sources in the environment in which the Khepera robot moves around. The Khepera robot will move towards just one of these sound sources, as is evident when studying real crickets.

The robot has been tested with real male crickets of the species Gryllus bimaculatus. When the crickets sing the calling song, the robot (that models a female cricket) responds by turning and moving directly toward the crickets.

Further, we have shown that the robot will respond to songs with the right pattern and frequency (e.g., the Gryllus bimaculatus song is built up by chirps with 200ms intervals. In each chirp, there are 3 syllables of 20ms each with 20ms in between them. The carrier frequency is 4.7kHz). The robot will not respond to a constant tone, and neither will it respond to song with carrier frequencies much different than the ideal frequency (4.7kHz in the case of Gryllus bimaculatus).

In order to be able to compare robot and cricket behaviours, we built a video-tracking system for the Khepera robot. Using this video-tracking system, we provide behavioural evidence that is comparable to the the evidence provided by biologists. At a later stage, we are going to implement the firing structure of of the cricket's neural processing in a spiking neural network for controlling the Khepera robot. The methodology of showing sufficient control systems for specific behaviours is believed to be extendible to other animal behaviours.

In general terms, we use an artificial organism (i.e. robot) to verify a hypothesis by using it together with biological data. We implement a hypothesised control mechanism in a robot, put the robot under the same conditions as the living animal that we hypothesise something about, and we show how the hypothesised control mechanism can account for the animal behaviour under these (biologically true) conditions.

With this work, we show that our hypothesised control mechanism can account for phonotaxis behaviour not only under artificial experimental settings, but also when tested under biologically true conditions: the robot with the newly developed auditory circuit does phonotaxis to real cricket song. Hence, we can conclude that the hypothesised control mechanism can account for the female cricket's phonotaxis behaviour, even though that these experiments cannot tell whether this is the cricket control mechanism --- only neurophysiological evidence can tell that. However, this simple control mechanism has been verified to be a possibility.

Research Articles:

Henrik Hautop Lund, Barbara Webb, and John Hallam. A Robot Attracted to the Cricket Species Gryllus bimaculatus. In P. Husbands and I. Harvey (eds.) Proceedings of Fourth European Conference on Artificial Life, pp. 246-255, MIT Press/Bradford Books, MA, 1997.

Henrik Hautop Lund, Barbara Webb, and John Hallam. Physical and Temporal Scaling Considerations in a Robot Model of Cricket Calling Song Preference. To appear in Artificial Life 4:1, 1998. Also to appear in Proceedings of Artificial Life VI, 1998. Plenary talk.