Three-Tiered, Delay-Tolerant Architecture for Wildlife Tracking: How Researchers Track Fruit Bats and Stop the Spread of Deadly Viruses

By Lori Cameron
Published 04/13/2018
Share this on:

Flying Lyle's flying fox (Pteropus lylei)

Several years ago, over 28,000 people contracted the Ebola virus and nearly half of them died from it. Although they can’t be sure, health officials believe the outbreak started when a young child in West Africa came in contact with a fruit bat.

Also known as “flying foxes,” fruit bats are of particular interest to data scientists and ecologists because they are known to carry the Nipah, Hendra, and Ebola viruses.

“The significance of wildlife tracking is further amplified for species that can carry virulent and potentially deadly diseases, as the movement of animals correlates highly with the likelihood that the disease spreads across the landscape, as was the case with Ebola spread by fruit bats,” write Philipp Sommer, Branislav Kusy, Philip Valencia, Ross Dungavell, and Raja Jurdak, authors of “Delay-Tolerant Networking for Long-Term Animal Tracking.” (login may be required for full text)

The researchers, who hail from ABB Corporate Research Switzerland and CSIRO’s DATA61—Australia’s leading digital research network—teamed up to test a new hardware and software animal tracking framework on flying foxes. They hope their system can be used to prevent the spread of deadly diseases such as Ebola and to support wildlife conservation in general.

Challenges to wildlife tracking

In studying the technology used in animal tracking, the researchers faced three significant challenges:

A Constrained and Variable Energy Budget

Since animal ethics considerations limit the form and weight of electronic devices attached to animals, mobile nodes need to rely on tiny batteries for energy storage. Therefore, the “software needs to gracefully adapt to varying energy budgets and schedule sampling of sensors in accordance with the current battery state of charge, daily energy budget, and predicted activity of the mobile object,” say the authors.

Intermittent Network Connectivity

Since wildlife can roam vast areas each day, the authors say “software framework must offer seamless communication with local data buffering to support situations when animals leave a known area for weeks and return to a different area, possibly hundreds of kilometers away.”

Lack of Physical Access

Because it might be impossible to recapture tagged animals again if their collars malfunction, the authors say it is important that the software framework provide for remote, wireless debugging and task configuration.

In response to these challenges, the researchers developed a three-tiered solution: mobile nodes (animal collars), gateways, and the cloud.

“We based our system architecture around a sparse network of gateways that communicate directly with mobile nodes to download the most recent data and use the Internet to connect to a cloud-based service to deliver sensor data and to synchronize the metadata among gateways,” say the authors.

A three-tier device architecture for delay-tolerant animal tracking and monitoring. Bottom left: A mobile node integrated into a collar for flying foxes (fruit bats). Bottom right: A gateway node built around a low-cost embedded Linux platform, which is mounted on the back of a solar panel.

Proposing an architecture: Mobile Nodes

Ethics regulations require that collars attached to animal weigh no more than 5% of the animal’s body weight. Fruit bats usually weigh about a quarter of a pound, so that’s 20 grams.

That means the collar must be tiny, yet energy efficient.

“The energy budget, computational power, communication bandwidth, and storage capabilities of such devices are highly constrained. Mobile nodes feature multiple sensing modalities, persistent data storage, and a short-range wireless transceiver,” the authors say.

Energy and time dynamics from flying-fox trackers. Top: Battery voltage measurement and GPS starts as reported by a mobile node attached to a free-living flying fox. The task scheduler adapts the sampling rate of the GPS receiver based on the battery voltage. Bottom: Distribution of the time interval between successive contacts with a wireless gateway for 73 flying foxes collared with tracking devices.


Next, gateway nodes must be placed strategically where animals tend to congregate. For fruit bats, those places are generally roosting camps where they can mate and raise their young. Gateways are connected to cellular networks or wi-fi links.

According to the authors, they “implemented a framework for Remote Procedure Calls (RPC) with low overhead using bidirectional radio packets between a gateway and a mobile node. [Their] framework provides the flexibility to implement high-level communication protocols for data downloads, remote configuration, or reprogramming on top of basic RPC commands.”

Top left: A network of gateways deployed across Australia’s East Coast. Top right: Movement tracking for an individual flying fox based on GPS samples at up to 1 Hz. Bottom: Data acquisition and download at the gateway node, and variations in the mobile node’s battery voltage during the deployment period. (Map: ©Mapbox, Data ODbL ©OpenStreetMap contributors)

Cloud Services

The third step involves connection to a cloud network. The bats fly where they will, the collars send data to the gateways, and the gateways transfer the data, via the cloud, to its final destination.

“The third tier in our system consists of a cloud-based web service. As bandwidth is less restricted between gateway nodes and the cloud services, we employ standard Internet protocols such as HTTP and encode data into JavaScript Object Notation objects,” say the authors.

The software architecture of our framework. Sensing tasks can be remotely configured using a set of simple rules that are evaluated at runtime by the energy- and context-aware task scheduler. Sensor readings from different sources are encoded into a Tagged Data Format (TDF) stream on the mobile node and decoded on the gateways before the data is forwarded to the cloud service. Gateway nodes can update the task configuration and program image remotely using RPC commands.

While the current research focuses on animal tracking, the authors believe it has potential in other areas as well.

“While animal monitoring has helped motivate and mature our network architecture, we expect it to be useful more broadly in logistics to provide traceability of unpowered mobile assets across heterogeneous and spatially spread landscapes, such as food products in transit. An interesting direction for future research is how to leverage and direct mobile gateways, such as smart-phone-based gateways, to download data from nodes that never come within range of an infrastructure gateway,” they say.

Related research on wildlife, animal tracking, and conservation in the Computer Society Digital Library

Login may be required for full text.




About Lori Cameron

Lori Cameron is a Senior Writer for the IEEE Computer Society and currently writes regular features for Computer magazine, Computing Edge, and the Computing Now and Magazine Roundup websites. Contact her at Follow her on LinkedIn.