DescriptionSmart cities include a diverse of set of infrastructures that are woven into the fabric of the lives of its residents. These infrastructures range from communications to transportation to utilities such as power, water and gas. The interactions of these ecosystems of machines and people is facilitated by the presence of a large collection of devices that form the Internet of Things (IoT). This thesis addresses two subcomponents that are integral parts of the smart city ecosystem, namely the smart grid and IoT. We model end-user (prosumer) behavior in the smart grid in the first part of the thesis relying on both Expected Utility Theory (EUT) and Prospect Theory (PT). In the second part of the thesis, we implement privacy mechanisms in IoT on a testbed realized using the ORBIT framework at WINLAB.
Most studies of prosumer decision making in the smart grid have focused on single, temporally discrete decisions within the framework of expected utility theory (EUT) and behavioral theories such as prospect theory. In this thesis, we study prosumer decision making in a more natural, ongoing market situation in which a prosumer has to decide every day whether to sell any surplus energy units generated by the solar panels on her roof or hold (store) the energy units in anticipation of a future sale at a better price.
Within this context, we first propose a new behavioral model that extends EUT to take into account the notion of a bounded temporal horizon over which various decision parameters are considered. Specifically, we introduce the notion of a bounded time window (the number of upcoming days over which a prosumer evaluates the probability that each possible price will be the highest) that prosumers implicitly impose on their decision making in arriving at "hold" or "sell" decisions. The new behavioral model assumes that humans make decisions that will affect their lives within a bounded time window regardless of how far into the future their units may be sold. Modeling the utility of the prosumer using parameters such as the offered price on a day, the number of energy units the prosumer has available for sale on a day, and the probabilities of the forecast prices. We fit both traditional EUT and the proposed behavioral model with bounded time windows to data collected from 57 homeowners over 68 days in a simulated energy market. Each prosumer generated surplus units of solar power and had the opportunity to sell those units to the local utility at the price set that day by the utility or hold the units for sale in the future. For most participants, a bounded horizon in the range of 4--5 days provided a much better fit to their responses than was found for the traditional (unbounded) EUT model, thus validating the need to model bounded horizons imposed in prosumer decision making.
In consideration of the facts that EUT is guided strictly by objective notions of gains and losses, and to capture more realistic behavioral patterns of consumers who own energy storage and renewable energy units, we use Nobel-prize-winning prospect theory (PT) as a tool which provides a psychologically accurate description of the decision making under uncertainty and risk. We enhance a PT based behavioral model by taking into account bounded horizons. This new behavioral model for prosumers assumes that in addition to the framing and probability weighting effects imposed by classical PT, humans make decisions within a bounded horizon. For most participants, the PT based model with a bounded horizon in the range of 1--6 days provided a much better fit to their responses than was found for the traditional EUT based model, thus validating the need to model PT effects (framing and probability weighting) and bounded horizons imposed in prosumer decision making.
By design, IoT devices collect copious amounts of information some public and some personal, and as such pose unique challenges to privacy which stem from integrating connected devices to a larger network such as the internet. To overcome such compromising of private information, we develop an experimental testbed using the ORBIT framework at WINLAB to implement differential privacy measures via traffic shaping mechanisms designed for different optimal criteria. We validate the feasibility of implementations of the traffic shaping mechanisms in a real-world testbed along with evaluating their performance, and also projecting the issues that can be faced in practical setup. The implemented mechanisms are designed to shape packet send times and sizes to make traffic rates uncorrelated with user activities. We utilize the network traffic of three commercial off-the-shelf IoT devices to evaluate the performance and effectiveness of these shaping mechanisms. We present performance comparison of the shaping mechanisms in terms of relative byte overhead and relative excess delay per packet.