One way to do long computations with short bursts of power
In the 2000 movie Memento , the main character tries to solve the mystery of his wife’s murder, despite suffering from amnesia that causes his brain to effectively ”reboot” every 5 minutes. In the world of computing, ”passive” radio-frequency identification (RFID) chips have a similar problem. Dependent for power on infrequent, scavenged RF energy from a reading device, RFID chips may reboot more than once per second and then lie dormant indefinitely, waiting for the next reader to come along.
But computer scientists in Massachusetts are working on software, aptly named Mementos, that could allow an RFID to perform computations that span many power losses and reboots. The software may enable the chip to compute cryptographic protocols, leading to more secure signals. And it might allow RFID chips to be more than just data collectors. They could analyze and possibly take action based on changes to the stress on a ”smart” bridge or to trends in a person’s vital signs, for instance. Such computational RFIDs could play a role in the transformation of the Internet from a network of computers to a network of things--appliances, cars, smart clothes, and so on.
”We’re working on software to make it possible to actually compute, given that our power is going to be disappearing and reappearing,” says Kevin Fu, assistant professor of computer science at the University of Massachusetts Amherst.
Mementos does two things: It makes sure that the RFID keeps working toward finishing a computation, and it also keeps the chip in a state such that if it loses power, it can quickly resume work when the power returns. One way the software does that is to have the chip perform energy-intensive tasks, such as writing data to flash memory, only when ample power (more than 2.2 milliwatts) is available.
Ravi Pappu, cofounder of the RFID company ThingMagic, based in Cambridge, Mass., says that the work of Fu’s team is very important.
”We have millions of computers everywhere, but computers have been chained to desks and server racks and other kinds of infrastructure,” Pappu says. ”Rather than being constrained by dragging computers of various shapes and sizes into the real world, could we equip the real world with computing? I think what Kevin and his guys have done is an advance in that direction.”
Fu and his colleagues developed their RFID computing software on Intel’s prototype Wireless Identification and Sensing Platform (WISP), a postage-stamp-size RFID chip with specs suited for a 1980s home-brew computing enthusiast: a 16â¿¿megahertz microprocessor, 512 bytes of RAM, and 8 kilobytes of storage (in the form of flash memory).
The amount of RF power WISP picks up, Fu says, can vary drastically. And Mementos must make programs run on WISP perform under each scenariosleep (0.2 to 2.5 microwatts); midrange (1.8 to 3.6 mW), which allows it to read from memory and compute; and active (2.2 to 25.2 mW) in which it can even write to flash memory.
”As much as possible we’d like to protect programmers from the underlying problems” of fluctuating power and regular reboots, says Benjamin Ransford, a graduate student in Fu’s lab.
Jason Flinn, associate professor of computer science and engineering at the University of Michigan, says he’s impressed by Mementos but thinks that it still has a long way to go.
The University of Massachusetts work ”asks some very interesting questions and has some preliminary ideas,” he says. ”But I don’t know that they’ve solved those problems or fully validated it yet.”