Environmental Energy Harvesting for Wireless Sensor Networks

Professor Mani Srivastava, Director
The Networked and Embedded Systems Laboratory (NESL)

Wireless networked sensors and actuators deeply embedded in the physical environment around us promise to revolutionize many facets of human society including sciences, engineering, arts, entertainment, education, the military, and day-to-day life. But limited battery energy in size-constrained sensor nodes has proven to be a major difficulty in most application contexts.

Traditionally researchers have sought to address this by developing energy-efficient circuits and software so as to make the batteries last longer, thus increasing the useful lifetime of the network before batteries need to be replaces. However, true autonomy of wireless sensor networks depends on their reliable operation for extended times without human intervention for tasks such as energy resupply.

A possible solution to this problem is to "harvest", or "scavenge" energy from the physical environment at run time. Sources of energy scavenging opportunities include solar, vibration, wind, thermal, etc. While hardware technologies to harvest many forms of environmental energy exist, there is currently little understanding of how a distributed embedded system should exploit this capability. In distributed systems, not only does the energy source vary in time, but also the energy available at different locations, and thus at different nodes of the sensor network, differs. In this situation, the performance can be improved by scheduling tasks according to the spatiotemporal characteristics of energy availability.

The challenge, then, is to find mechanisms that can adapt the system behavior to the spatiotemporal profile of available environmental energy. For example, consider the task of routing a packet through the ad hoc wireless network formed by the nodes in a sensor network. Classical routing protocols either ignore energy, or at best take into account remaining battery energy or the energy spent in wireless communication, when selecting the route that the packet should take. However, a routing protocol that is harvesting-aware may choose a route which is higher in cost, or whose nodes have depleted batteries, if it knows that the nodes along this route have more opportunity for harvesting energy from the environment in the near future.

Researchers at NESL are developing wirelesssensor networks where the nodes use distributed online algorithms to learn the spatiotemporal profile of the available environmental energy. The nodes then use this knowledge to predict future availability and adapt network level performance and resource allocation for tasks such as routing, network topology control, and sensor tasking.

Initial results with a prototype network of HelioMote sensor nodes developed at NESL with solar energy scavenging and storage capabilities show that harvesting-aware operation yields more than 2x improvement in useful network lifetime relative to a network of HelioMotes operated without such intelligence. NESL Researchers have also developed a "Harvesting Theory" to provide a mathematical framework to model and reason about the energy harvesting process. Using this, designers can determine bounds on achievable performance and lifetime, and judiciously design and operate the system so as to achieve the "holy grail" of energy neutral wireless sensor networks that can last forever.

Researchers at NESL are also addressing the issue of optimal design of the harvesting power supply system that is crucial to design of harvesting-based sensor node platforms. For harvesting sources such as solar cells, key characteristics such as source impedance are a strong function of the environmental condition. Moreover, rechargeable batteries also exhibit various non-idealities. A proper addressing of these issues requires smart and properly designed harvesting circuits that maximize energy extraction efficiency over a range of environmental conditions and battery nonidealities; otherwise, the advantages of harvesting may get overwhelmed by the inefficiencies induced by the circuit.