RWTH Aachen
Ratpack: A study on nest structures and social interactions of subterranean rodents with self-organizing sensor networks
Uni Tübingen
Project description
Regardless of the fact that the Norway rat (Rattus norvegicus) is a well-established laboratory animal, very few work has been done concerning the ecology and sociobiology of the rat in its natural habitat. Apart from the comprehensive work by Calhoun in 1963 [1], no further work has been done on issues like the complex structures of the subterranean burrow systems built by these animals or the social interactions between a group of animals sharing the same burrow.
However, Calhoun's way of disclosing these structures was to dig them out, thereby destroying them. In contrast, our approach of reconstructing the burrow systems of free ranging rats will be non-destructive: Individual animals will be equipped with small microcontroller units, so-called sensor motes. Such motes typically carry a variety of sensors and are able to communicate via radio using highly flexible communication protocols. They are cheap and designed to build up autonomous, self-organized sensor networks. We equip rats with a mote-type that carries a 2-axis accelerometer, thereby enabling us to record the temporal and spatial activities of the animals while they move through their burrow. Further sensors are a microphone and a light sensor which produce additional data about the behavior of the animals. The data obtained by the mote network will enable us to reconstruct the spatial structure of the burrow as well as the social interactions of the rats.
Reconstruction of burrow structures
Recording and analysis of rat vocalizations
Our approach
to disclose the structures of subterranean burrows in an
non-destructive manner measures the steps and the directions of
the rats' movements while they walk about the burrow [2]. In order
to detect walking activity, we equip our sensor motes with
inertial sensors that measure twists of the animals vertebral
column when the rat is walking.
Extensive experiments [3] with rats walking on a treadmill at various speeds showed a coherence between the time period of a step and its physical length, allowing an easy estimation of the length of a displacement just by counting step events. Further, a magnetic compass sensor supplies us with the animals' actual heading at any time. Both components add up to a displacement vector that is used to update the rats' estimated position. Pilot studies with animals walking in artificial burrows show that it is indeed possible to reconstruct the pathway of a rat using exclusively the informations from our sensors.
Another sensor that is attached on the sensor mote is an ultrasonic microphone to record the acoustic communications between rats, when two or more animals meet. Vocalizations of rats occur in the frequency band between 10 and 90 kHz, they are very variable but in many cases the signal structure relates to a specific behavior, allowing to draw conclusions about the kind of relation between the involved individuals [4,5].
However,
as battery and computing power are short in our setup, we can't
afford to use sophisticated sound analysis techniques like fft. We
decided to use zero crossing analysis (ZCA) instead and measure
the pulse width of the signal after it was extensively filtered
and converted into a 1-bit digital stream.
The results of our ZCA-processing is promising: Despite an enormous data reduction, the amplitude and spectral properties of a signal are represented almost as good as with a fft, although they were calculated only by a battery-driven 8-bit microcontroller. We suppose that it will possible to develop a classifier that can do the characterization of most of the recorded rat vocalizations in real-time too.
However, like all path integration algorithms the method suffers from inaccuracies. But as a rats' burrow consists basically of pipe-shaped pathways, repeated walks of the same animals will increase the certitude of burrow shapes and finally allow better estimations.
The red dots in the graph show the reconstruction of a rats' pathways through a burrow system based on simulated inertial and magnetic compass data.
Gaining social informations from network connectivity
Beyond the social information we gained by vocalization analysis, a meeting of two individuals by itself is a data point that can be used to reconstruct the social network of the animals sharing a burrow. Due to the fact that the radio waves of our motes only propagate over short distances in the soil, motes will only transmit data, when the animals meet.
As a consequence, the sensor motes network is sparsely connected and quite variable. Therefore each meeting event is not only critical for data transmission through the network but it also reflects a precious information that can be used for the reconstruction of the social structure in the burrows rat population [6].
Routing algorithms have to be found that, on the one hand, transfer the data packets through the network quick and reliable, while keeping redundancies as low as possible. Moreover, the actual routing strategy itself will map the social network relations of the rat population in the burrow and is a significant component of the behavioral results in the research project.
This work was kindly supported by Microsoft Research, Cambridge, UK
References
[1] Calhoun JB (1963) Sociology and Ecology of the Norway Rat US Health Serv. Publications No. 1008. Bethesda, Maryland.
[2] Zeiß, M (2009) Rekonstruktion von natürlichen Laufbewegungen der Ratte mit Hilfe von Magnet- undInertialsensoren. Diplomarbeit, Fakultät für Biologie der Universität Tübingen.
[3] Schulte, T (2010) Ganganalyse und Pfadrekonstruktion mittels Dead Reckoning bei Ratten. Diplomarbeit, Fakultät für Biologie der Universität Tübingen.
[4] Kaltwasser, M-T, (1990) Acoustic signaling in the black rat (Rattus rattus), Journal of Comparative Psychology. 104(3):227-232.
[5] Voipio, H (1997) How do rats react to sound? Scandinavian Journal of Laboratory Animal Science. Supplement.
[6] Wey T, Blumstein DT, Shen W, Jordan F (2008) Social network analysis of animal behaviour: a promising tool for the study of sociality. Animal Behaviour,75: 333-344.
