2014: Sensor nodes in the savannah to combat wildlife poaching
The Tsavo Sensor Nodes project sought to determine the viability of placing smart sensors in the Wildlife Works project area in order to gather rich information about wildlife and poacher movement, and to track environmental changes over time. Additionally, the project aimed at working with local rangers and state officials. Elephant poaching for ivory is a very real, very immediate problem in Kenya. Director’s fellow David Kobia is involved in an initiative to see if technological solutions can be applied to the issues that wildlife rangers are facing on a daily basis in the savannah. During the May 2014 Director’s Fellow event in Kenya, he brought the Media Lab team to Tsavo East National Park in southern Kenya to brainstorm with the rangers and identify feasible, potential solutions to deterring and detecting poaching initiatives in real time.
The Wildlife Works project area is a migratory corridor that connects Tsavo East and Tsavo West in Southern Kenya. This area, which is 200,000 hectares, experiences high poaching rates due to its main connection between Tsavo East and West. In order to learn more about the movement patterns of wildlife and poachers, as well as the environmental state of the project area, we propose the use of low-power, low-cost sensors that are distributed throughout the project area. Our research centers on the development of these sensor nodes, the wireless protocols that govern their communication, and the mechanical requirements for chronic installation in the Tsavo wilderness.
What worked well in the Tsavo Sensor Node track
The team successfully deployed a sensor node in Tsavo and were able to collect data. The data collected was sound-level data from a microphone and motion data from a PIR sensor. We used an Arduino and an SD Arduino shield to log the data to an SD card. The Arduino was powered by an external battery pack to ensure that it lasted the entire duration of data collection.
What didn’t work well in the Tsavo Sensor Node track
The team did not have the sensor node in an out-of-the-box state when we arrived in Tsavo. In fact, the sensor node was developed from scratch in the field, which caused some delay and created challenges in in 1) quality of data recorded, and 2) types of data recorded. Originally the team had intended to use the Tidmarsh node developed by Brian Mayton (Media Lab Responsive Environments group) to collect temperature, pressure, and acceleration data in addition to the microphone and motion data. There were some issues communicating with the Tidmarsh node, and were consequently unable to use the node; we lacked the AVR programmer for the Tidmarsh nodes, so we could not reprogram it to address the communication problems. Working in the field with very limited power and Internet also caused problems. By the end of the sensor node build session in Tsavo, two of three laptops were out of power and the third was at 3% battery. The sound data was not captured fast enough for audio quality information (it could not be played back as audio) since the readings were taken at 28Hz (typical audio requires a minimum of 8000Hz). The motion sensor was never activated during the night. This could be due to infrared-blocking coating on the sensor enclosure.