Exploiting Redundancy for Robust Sensing

Опубликовано 6 сентября 2016, 5:28

Over the last few years, we have seen a number of real-world systems using live data from sensors (e.g., sensor motes and webcams). A crucial requirement of these systems is high availability. However, achieving high availability is extremely challenging due to three factors. First, due to their harsh deployment environments, sensor and communication failures are common. Second, sensors are often resource constrained. Third, the systems must run mostly unattended. In this talk, I will address the challenge of robustly collecting and storing sensor data in a sensing system. I will show that several domain-specific properties can be exploited to address this challenge. In particular, for wireless sensors (e.g., motes), I will show how the underlying broadcast medium can be exploited to robustly collect aggregate sensor data. For archiving sensor data in the Internet, I will show how the weak data consistency requirements of a sensing system can be exploited to tolerate many large failures typical in today's Internet. I will describe my solutions in the context of IrisNet, a wide-area sensing infrastructure that I have designed and implemented as part of my dissertation.