The Internet of Things (IoT) has emerged as a major paradigm, enabling applications in industries from smart cities to health care and dietary assessment. IoT devices embed sensors and low power processors into the physical world, allowing a rich set of information to be gathered by a diverse set of devices. IoT devices are becoming so common that they are expected to be over 50 billion devices worldwide by 2020. The widespread deployment of IoT devices necessitates that we understand the challenges of designing system-level solutions for these platforms. IoT platforms have traditionally been passive devices, sourcing information from edge endpoints and streaming the data up to big beefy cloud servers that process the "big data" and provide end-users valuable insights. It was a one-way flow of information. But the landscape of IoT is changing. IoT is moving from being a passive sensing environment to a more agile and active sensing environment where information is flowing back and forth between the cloud services and the edge endpoints (as shown in the picture below).
|
|
| Passive IoT | Active IoT |
We focus on Active IoT specifically in the context of autonomous machines (e.g., drones, and thus the project name "Flying IoT"). Since IoT devices are often power and performance constrained, our interests lie in understanding how to compute and offload the data processing intelligently between the edge, its endpoints, and the remotely available cloud computing resources, so that there can be a balance between performance and power efficiency. In the context of autonomous vehicles, we are also interested in understanding how safety and reliability are affected when offloading. Furthermore, the group focuses on understand how these IoT devices can communicate with one another to offload computation amongst one another in a swarm like fashion so that there can be collaborative processing taking place.