Wildlife monitoring with the help of distributed artificial intelligence

Environmental changes on our planet have reached historic proportions in both intensity and speed. The massive decline in biodiversity and climate change unbalance the ecosystem.

In the course of evolution, nature has developed countless solutions to almost every conceivable challenge. One remarkable approach is the swarm intelligence of many animal species, both within species and across species. Synergies are evolved that make a swarm more intelligent than the sum of its members. However, digitalization also creates technologies and developments to intelligently meet concrete challenges.

The aim of the “SyNaKI” project is to map a natural swarm intelligence virtually in a network of microprocessors (digital swarm) in order to enable data analysis by artificial intelligence (AI). This AI is to be developed using the example of a concrete application scenario of tagged vultures in a cross-species association with tagged scavenging terrestrial mammals. The collected and analyzed data will then be sent to the experts for wildlife observation via satellite. With the help of the data generated by the tagged animals, these scientists can gain deep insights into the ecosystem of the living creatures. In this way, the behavior of the animals can be researched in the long term, any changes in the animals' behavior can be detected at an early stage and negative developments can be counteracted.

“SyNaKI” is a sub-project of the GAIA initiative, for which the Fraunhofer Institute for Integrated Circuits IIS and the Leibniz Institute for Zoo and Wildlife Research (IZW) have joined forces to form an interdisciplinary consortium.

The challenge

Network coverage

Animal swarms tend to occur in remote areas beyond the reach of terrestrial communications infrastructure. But large amounts of data can often only be transmitted within days or even weeks in poorly connected areas. Therefore, it is necessary to connect existing networks via satellite to enable global connectivity.

Data rate

Sensors make it possible to measure data on animals and in their environment, from which scientists can derive important information on the state of ecosystems. The amount of data produced is problematic, because in principle more data can be collected than transmitted. Instead of risking incomplete data sets, local preprocessing of sensor data can reduce the amount of data to be transmitted. This means that less data rate is required for the transmission.

Latency

The limited data rate and the lack of terrestrial network coverage result in high latency, a delay in the transmission of data. This in turn can result in delayed data evaluation. In situations where fast action is required, such a delay can be problematic. Local preprocessing by AI enables a reduction in the amount of data to be transmitted and thus lower latency. Therefore, the connection of local networks to the satellite also plays an important role here in order to transmit the information generated in the local AI network quickly and efficiently.

Hardware limitation

In order to be able to process data directly on local hardware, the computing power available there must be sufficiently dimensioned. This can become a challenge, especially for mobile data acquisition by transmitters attached to small animals. Since vultures can only carry a small amount of their own body weight without being affected in their behavior, a strong limitation of the hardware is necessary. This is because computational power is always correlated to the resulting energy consumption and therefore battery size. The larger the battery, the heavier it becomes. The execution of complex AI algorithms directly on the animal transmitter is consequently only feasible to a limited extent, as the hardware required for this would be too heavy and too large.

Synergy of natural and artificial intelligence in a swarm

Within the scope of this project, solution approaches for the aforementioned challenges are being developed. To this end, Fraunhofer IIS is developing a distributed and sensor-based evaluation of large data series using a dynamic swarm of microprocessors, mirroring the evolutionary intelligence of a community of scavengers. For this purpose, an artificial intelligence is designed that is able to specifically classify swarm behavior patterns. Real biometric measurement data of individual animals obtained from distributed sensors within a community of species will serve as a basis. For this purpose, an extreme edge or extreme edge-like network will be designed, which is able to act autonomously and analyze the collected measurement data. This extracted information can then be sent via a future satellite-based mioty® net.

Sensor-based evaluation of the data

The data collected with the help of the animal transmitters is difficult to interpret on the one hand, and on the other hand, the full amount of available data cannot be transmitted via existing radio technology. One solution to this is the implementation of sensor-related AI. The measured data can thus be reduced to the most relevant information, which minimizes the amount of data to be transmitted.

Extreme edge computing

Due to their design, the small, lightweight animal transmitters carried by the vultures can only provide limited computing power. Therefore, the use of extreme edge computing is particularly promising in this use case. In the extreme edge concept developed by Fraunhofer IIS, complex algorithms are not processed centrally on a single device, but distributed across multiple devices in the vicinity of the extreme edge. For this purpose, ad hoc networks are set up that connect the individual radio nodes in a network. In concrete terms, the complex algorithms in the project are not processed centrally on a single device, but distributed on several animal transmitters available in a swarm. 

Data transmission via satellite

Area-wide IoT connectivity will only be possible if the networks in remote areas are completed by satellites. For direct transmission from the transmitter node to the satellite, Fraunhofer IIS is developing a communication system based on terrestrial mioty® technology. By using multiple communication links, bidirectional communication is to be enabled to ensure not only robust and energy-efficient data transmission to the satellite, but also parameterization and control of the ground station at the same time.

You may also be interested in

GAIA-Sat-IoT

Sensor-based AI, energy-efficient electronics and satellite-based IoT technology for wildlife research and conservation

 

Artificial intelligence

Go back to business area “Artificial intelligence“