Small, inexpensive, and capable of managing its own energy resources, Camaroptera was developed as a collaboration between Carnegie Mellon University and the University of Trento in Italy to collect images and transmit interesting data. The remote image sensor, named after a tiny, industrious bird is batteryless and powered entirely by solar panels. Even in a crowded city environment, Camaroptera can relay images over miles after processing them itself. Using edge computing to distinguish interesting images and disregard uninteresting data, the device could be used to find and reroute traffic jams, mark unfilled parking spaces in lots, and even identify dangerous situations in public areas.
Improvements in energy-harvesting systems have led to wireless IoT units that are entirely energy-autonomous, like Camaroptera. These systems collect energy from their environment, buffering the energy in a battery or capacitor, then activating and performing some number of tasks after sufficient energy is collected. To ensure that every application — data collection, computing, and sending data over radio — the device almost never runs out of energy, even on a cloudy day. Brandon Lucia of Carnegie Mellon and Ph.D. candidate Harsh Desai developed a new algorithm for scheduling how an energy-harvesting device uses its time and energy. If less energy is available, the scheduling algorithm is able to correct and have the device perform a simpler, less energy-intensive task.
The hardware platform, comprised entirely of COTS components, is designed for sensing, computing, and long-range communication. Three small boards make up the Camaroptera: the sensor board, the power system board, and the solar board. The software system is centered around its use of near-sensor processing pipelines and on the newly developed algorithm, an adaptive control mechanism that ensures the minimum quality of service requirements by varying operations based on the amount of incoming energy — I love this feature. Fine work!
The ability of Camaroptera to distinguish interesting data is centered on its processing pipeline, which functions in four stages: difference filtering, inference, pre-transmission transformation, and transmission. In the inference stage, Camaroptera performs an application-specific routine on an image to determine whether it is interesting to the application. Given the low-cost components and ability to manufacture them in large quantities, the device would be an inexpensive option to monitor civil infrastructure. In addition to other benefits, the energy-harvesting aspect also serves to eliminate device maintenance costs related to battery replacements. Camaroptera can manage its own time and energy and is cheap to build, cheap to deploy, and cheap to maintain.
Coll and all, but I loathe the idea of these little cameras spying everywhere.