On a Wildlife Tracking and Telemetry System: A Wireless Network Approach Andrew Markham BSc(Eng) in Electrical Engineering, 2004 A thesis submitted to the Department of Electrical Engineering, University of Cape Town, in fulfilment of the requirements for the degree of Doctor of Philosophy. August 2008 Declaration I declare that this dissertation is my own, unaided work. It is being submitted for the degree of Doctor of Philosophy at the University of Cape Town. It has not been submitted before for any degree or examination in any other university. Signature of Author ........................................................... Department of Electrical Engineering University of Cape Town August 2008 iii Abstract Existing approaches to monitoring wildlife with wireless networks have not taken into account the vast heterogeneity inherent in the Animal Kingdom, especially with respect to bodyweight (and hence tag carrying capacity). This has resulted in a single design of tag which is only suitable for placement on larger animals. Thus, with existing technol- ogy, small animals cannot be monitored using the wireless network system. Motivated by the diversity of animals, a hybrid wildlife tracking system, EcoLocate, is proposed, with lightweight VHF-like tags and high performance GPS enabled tags, bound by a common wireless network design. Tags transfer information amongst one another in a multi-hop store-and-forward fashion, and can also monitor the presence of one another, enabling social behaviour studies to be conducted. Information can be gath- ered from any sensor variable of interest (such as temperature, water level, activity and so on) and forwarded through the network, thus leading to more effective game reserve monitoring. Six classes of tracking tags are presented, varying in weight and function- ality, but derived from a common set of code, which facilitates modular tag design and deployment. The link between the tags means that tags can dynamically choose their class based on their remaining energy, prolonging lifetime in the network at the cost of a reduction in function. Lightweight, low functionality tags (that can be placed on small animals) use the capabilities of heavier, high functionality devices (placed on larger ani- mals) to transfer their information. EcoLocate is a modular approach to animal tracking and sensing and it is shown how the same common technology can be used for diverse studies, from simple VHF-like activity research to full social and behavioural research using wireless networks to relay data to the end user. The network is not restricted to only tracking animals – environmental variables, people and vehicles can all be moni- tored, allowing for rich wildlife tracking studies. To transfer the obtained data effectively through resource diverse nodes, a network protocol, termed the Adaptive Social Hierarchy (ASH) was designed that ranks nodes according to their resources, such as energy or connectivity. ASH provides a scalable and adaptable method for nodes to discover the role within the network, inspired by the way animals form linear dominance hierarchies through dyadic (pairwise) interac- tions. Three different methods of forming the social hierarchy are presented. In the first method, pairwise ASH, pairs of nodes exchange their attributes and their estimates of rank in a two-way exchange. Although this is a simple method of forming the hierarchy, it does not take advantage of the broadcast nature of the radio channel. In light of this, a one-way method of updating ranks is proposed and shown to be able to estimate the node ranks faster than pairwise ASH, due to multiple nodes receiving the same beacon. However, both methods are unable to form an accurate social hierarchy in a stationary network, due to a limited visibility horizon. It is shown how to extend the horizon by cre- ating pseudo-connections between unconnected nodes, using an agent based approach. Simulation results are presented that demonstrate how the ASH concept can be used as a network underlay to enhance existing protocols or how it can form a cross-layer proto- col in its own right. ASH is simple, scalable and has a negligible load on the network as ASH data piggybacks on top of existing network discovery packets. The focus is shifted from the network design to considering how to better schedule lo- cation fixes for power hungry GPS receivers. Existing wildlife tracking collars acquire v fixes at constant time intervals. This leads to undersampling of high speed motion, and multiple redundant fixes when the animal is stationary. Uniform distance sampling of GPS locations is thus proposed. A low power, neck mounted, accelerometer is used to capture brief acceleration snapshots. An adaptive model, relating the snapshots to host speed is trained when the GPS unit is active. When the GPS receiver is powered down, the model is used to predict the speed of the host, and thus schedule GPS fixes at uniform distance intervals. The proposed scheme is implemented on low power 8 bit mi- crocontrollers, and demonstrates that it is able to automatically learn the habits of the host animal or person. This technique can reduce collar power consumption by as much as 50%, whilst generating more accurate traces than uniform time sampling, as well as creating a detailed speed-time profile as a byproduct of the speed estimation process. This work has considered the problem of tracking and monitoring wild animals and their interaction and dependency on their environment. A scalable, modular and adaptable solution has been proposed, that allows small and large animals to be monitored using the same system and which automatically sends data through the network to the end user. Thus, this work has the potential to greatly enhance the understanding of animal behaviour, by providing large amounts of inter-related sensor data with minimal human input. Acknowledgements I am indebted to a number of people for their guidance, help and advice during the course of my PhD. First and foremost, I would like to thank my supervisor, Dr Andrew Wilkinson, for his support and friendship over these years, from when I first knocked on his door and asked him to supervise my undergraduate thesis. Dr Wilkinson has the ability to cut past the fluff and go to the core of the problem, and this has benefited me immensely. He has also given me academic freedom to pursue whatever path I thought best, with nudges now and then when I was getting seriously off course. Dr Ken Stratford of Ongava Research Centre has been a wonderful sounding board, allowing me to bounce ideas off someone who is actually working in the field. Without his input, I doubt whether my work would be physically realisable and would be some blue sky abstraction. He also generously sponsored the purchase of GPS receivers and microcontrollers to build and test more toys. I would also like to acknowledge a debt of gratitude to Prof. Braae for donating his ‘paperclip’ fund together with Andrew so I could attend a conference in Italy and for reading through some of my papers. Prof. Braae also put me in touch with Ken, which helped immensely. Sam Ginsberg has been a great help too – it is helpful having the guru of microcontrollers just round the corner. He is always ready to shed some light on peculiar problems (such as MOSFETs misbehaving on remote islands). Sam also showed me how to use the milling machine, so I could make toys to my heart’s content. My lab mates, Kush, Kent and Rachel have been a great help in keeping disasters at bay and I am going to miss the deliberation over where to go for lunch every day. My friends have been very patient with me through this all, and I thank them for helping keeping me sane. In particular, thanks to Shannon for reading through the draft and giving me useful comments from the other side of the fence. I would like to thank my family, both old and new for their help during the course of my research. Phil showed me the black art of aluminium welding and he welded up the aluminium frames used for the penguin logger – for this I am very grateful. The Cunninghams have welcomed me into their home and fed and watered me and I would like to thank them for their generosity. I would like to thank Tony for drilling home modularity many years ago and for his constant stream of jokes into my inbox. Thanks to my little sister for reading through my dissertation – it always helps having feedback from MIT. David has kept me going with his flippant comments about the state of the dissertation. I would like to thank my parents from the bottom of my heart for their support and love and for giving me the greatest gift of all – a passion for learning and discovery. Finally, this would not have been possible without the light of my life by my side. Kathy has helped so much in getting this all done, from emergency missions to Dassen to cups of tea to keep me going. She has listened to me rattle on about hierarchies, ranks and nodes with infinite patience. Thank you for your unwavering love and support through this all. vii Contents 1 Introduction 1 1.1 Background . .1 1.2 Research Objectives and Solutions . .3 1.2.1 Ecolocate: A heterogeneous wireless network system for wildlife tracking . .4 1.2.2 The Adaptive Social Hierarchy (ASH): A Network Ranking System5 1.2.3 Uniform distance GPS sampling . .7 1.3 Contributions of this work . .8 1.4 Guide to the thesis . .9 2 Wildlife Tracking 11 2.1 Introduction . 11 2.2 Real world constraints . 12 2.2.1 Size, weight and shape of device .