An energy-harvesting sensor architecture and toolkit for building monitoring and event detection

Abstract

Understanding building usage patterns and resource consumption, particularly for existing buildings, requires a sensing infrastructure for the building. Often, deploying these sensors and obtaining real-time information is hindered by installation and maintenance difficulties resulting from scaling down and powering these devices. Devices that rely on batteries are limited by the scale of the batteries and the maintenance cost of replacing them while AC mains powered sensors incur high upfront installation costs. To mitigate these burdens, we present a new architecture for designing building-monitoring focused energy-harvesting sensors.The key to this architecture is masking the inevitable intermittency provided by energy-harvesting with a trigger abstraction that activates the device only when there is useful work to be done. In this paper, we describe our architecture and demonstrate how it supports existing energy-harvesting sensor designs. Further, we realize three additional design points within the architecture and demonstrate how the sensors are effective at building monitoring and event detection. The sensors, however, are classically disruptive: they improve ease of installation and maintenance, but to do so, they sacrifice some fidelity and reliability. Whether this tradeoff is acceptable remains to be explored, but the technology needed to do so is now here.

Capacity over capacitance for reliable energy harvesting sensors

“…The Gecko and Monjolo platforms ignore the difficulties associated with completing longer workloads and instead allocate just enough capacitance to turn on and perform a simple task. Sometimes, the rate of harvesting is the sensor itself [8,12,45]. However, this approach can require tedious and non-standard optimization of the cold start process and is severely limited in its simplicity.…”

“…Occupied indoor environments are the focus of a significant amount of prior work, and for good reason: most applications aim to improve the lives of people and are necessarily present in the spaces they occupy. Even the example applications of intermittent, energy harvesting systems are nearly all centered around monitoring indoor and human-centric phenomena [8,10,16,18]. We expect our environment to be lit, and that it may occasionally get direct or indirect sunlight.…”

Slocalization: Sub-uW Ultra Wideband Backscatter Localization

“…To that end, this paper introduces Slocalization, a new localization system that can localize static tags in both static and non-static environments with decimeter-level accuracy for less than one microwatt. At this power level, Slocalization is suitable for use with the burgeoning array of batteryless, energy harvesting systems [4,22]. A standalone Slocalization tag will well outlast the self-discharge lifetime of a standard coin cell battery [11,32].…”

Correlative monitoring for detection of false data injection attacks in smart grids

“…This paper proposes a framework for tackling the problem of detecting false data injection attacks. Motivated by recent advances in home and building monitoring (e.g., see [19], [20] and energy-harvesting metering [21]) we study a behavioralbased model that integrates sensor measurements from homearea networks and aims to "learn" normal electricity usage patterns 2 . Given the power state (i.e., the one reported by the smart-meter to the utility) and our forecasted usage, one can formulate a sequential hypothesis testing problem that reflects whether the system remains "in-control" or it is operating at Fig.…”

“…For example, alternative forecasting modules could be employed, mutatis mutandis, assuming they offer a predictive distribution (e.g., Gaussian processes regression) and computational restrictions are not in play; iii) a lightweight (see computing times in Figure 2), elegant and adaptive solution to the problem at hand that can be easily transitioned to practice and implemented using inexpensive sensor devices (see, e.g., [19], [21]) and computing nodes.…”