Energy harvesting is a key technology for the Internet of Things, specifically wireless sensor nodes. However, it presents challenging requirements engineers need to understand and address for optimal designs.
The technology lets devices work autonomously in locations where human maintenance is impossible. Solar, thermal, and vibration energy can be harvested easily, but those energy sources may deliver varying energy levels or may not be present at all times.
So, tomorrow's wireless sensor nodes will have to recover energy from a variety of energy sources if they are going to be fully autonomous. In addition, they will require micro-batteries to serve as backup energy sources, recharged as soon as the nodes have harvested enough energy.
Some of the most efficient nodes will be fully energy driven. They will transmit sensor information only when the energy they have harvested is high enough to deliver the needed power supply to the node to ensure a safe communication.
If sensor information is requested at fixed times -- including times when not enough energy is being harvested -- the nodes will need to use their batteries. This second scheme is an event-driven operating mode. Power-management algorithms similar to a best-effort scheduling or quality-of-service behavior are required to handle both energy- and event-driven modes.
The low energy levels at which wireless sensor nodes operate and their voltage-level variations over time are two big challenges in building autonomous systems. Designers have to think about a new way to supply their circuits with very low voltages and/or voltage levels that could vary in a wide operating range.
The easiest way to face those issues is to use robust design techniques fully adaptive to voltage variations. In addition, designers need to consider how to handle degraded communications or less-accurate sensing when the nodes are down to very low energy levels
Asynchronous quasi-delay-insensitive logic is useful for creating adaptive digital circuits dedicated to data treatment and power management. This kind of logic is insensitive to delay variations caused by process, voltage, or temperature changes. Moreover, the logic consumes power only where and when activity is detected -- inactive parts of the circuit are automatically on standby without any control.
Using such techniques, future autonomous Internet of Things nodes can work in a way in which they are woken up by external events. Such events can either be energy events indicating some environmental changes, like the arrival of sunlight in the morning, or data events such as when another node is requesting sensor information.
— Edith Beigné is a senior scientist in the Architecture, ID Design, and Embedded Software Department at CEA-Leti, a microelectronics research institute in France. Cyril Condemine and Jean-Frédéric Christmann also contributed to this article.