Distributed Energy Harvesting for Energy Neutral Sensor Networks

Works in Progress Editor: Anthony D. Joseph UC Berkeley adja@cs.berkeley.edu Energy Harvesting Projects EDITOR’S INTRODUCTION This month’s Works in Progress column has four contributions. The first examines how harvesting environmental energy in sensor networks changes the way an application developer views energy management, and discusses prototype devices. The second proposes devices that combine energy harvesting and data acquisition. The third explores novel approaches for opti- mizing the power extracted using piezoelectric materials. The final one explores kinetic and ther- mal energy harvesting from human users’ activities. —Anthony D. Joseph DISTRIBUTED ENERGY HARVESTING FOR ENERGY- NEUTRAL SENSOR NETWORKS Aman Kansal and Mani B. Srivastava, University of California, Los Angeles Embedded deployments, such as sen- sor-actuator networks, constitute a large class of pervasive computing devices. Unlike cell phones or laptops, which users can periodically recharge, pervasive de- vices must operate on their initial batter- ies. The highest reported energy densities for current battery technologies range around 3.78 kJ/cm 3 , which implies that for a low-power device operating at an average consumption of 1 mW to have a 10-year lifespan, it needs a large 100 cm 3 battery. Thus, energy supply is a major bottle neck for system lifetime, and har- vesting energy from the deployment envi- rons can help alleviate this. UCLA’s energy harvesting project (http://nesl.ee.ucla. edu/projects/heliomote) is making a two- pronged effort to address the various chal- lenges in building practical energy har- vesting sensor networks. First, we’ve started developing a theory for energy-neutral systems. These systems strive to meet application performance requirements using only environmentally available energy and can thus sustain themselves infinitely (until the hardware or application becomes outdated). In con- trast to battery-operated systems, power management in energy harvesting systems differs fundamentally in that it’s the avail- able power that’s limited and not the total energy. Also, power availability varies in time and might be different at different nodes in the network. While most appli- cations result in energy consumption at multiple nodes, some flexibility in deter- mining which specific nodes are used is usually possible. For instance, when a sen- sor network records an event in the envi- ronment, there’s a choice in determining the exact route used for routing the data to the user. Clearly, the achievable system lifetime depends on how the relay nodes make these choices with respect to the environmental energy’s spatio-temporal profile. Notably, the network as a whole must make the task allocation decisions in a distributed manner because no single node has complete knowledge of the entire network’s energy opportunity. We developed analytical models for the har- vesting and consuming entities and derived theorems that characterize their achievable performance. 1 We also demon- strated distributed protocols to schedule tasks in accordance with the environ- mental harvesting opportunity available at different network nodes. 2 The second part of the effort involves developing working prototypes, which has also revealed several interesting issues in the harvesting systems’ hard- ware and software design. Our first pro- totype consists of a network of solar energy harvesting sensor nodes called heliomotes. Each heliomote (shown in Figure 1) consists of a solar energy har- Figure 1. A heliomote, an energy-harvesting sensor node. PERVASIVE computing 1536-1268/05/$20.00 © 2005 IEEE ■ Published by the IEEE CS and IEEE ComSoc