Batteryless, Ambiently-Powered Internet of Things That Think: An Asynchronous Message Passing Approach
- Funded by: National Funds
- Project Acronym: AMPERE
- Funded under: Hellenic Foundation for Research & Innovation (H.F.R.I.)
- Budget: € (Overall: 187887.20 €)
- Start Date: 1st March 2020
- Duration: 36 months
- Website(s): HFRI
Powerful message passing algorithms (e.g., sum-product, max-product) have offered concrete examples on how decision making and inference can be facilitated through communication, at carefully crafted probabilistic graphs. More importantly, recent advances on scatter radio sensor networks by the principal investigator (PI) have demonstrated feasibility of ultra-low power (in the order of 20 microWatts) and cost (in the order of some Euros), joint sensing and wireless networking, through single-transistor radio frequency (RF) front-ends and reflection radio principles. Furthermore, the PIhas demonstrated energy harvesting circuits from ambient RF or bioelectric sources (plants) with record-breaking sensitivity, able to harvest ambient power, as small as 1 microWatt.
AMPERE is inspired by the fact that ambient energy, e.g., solar, kinetic, thermal, bioelectric or RF, has a common characteristic: fixed (on average) density per squared (or cubic)centimetre and thus, wireless sensor networks (WSN) over an extended area (or volume) could in principle harvest a large amount of energy. Thus, autonomous, in-network decisions should be possible, solely using ambient power,
- by exploiting ultra-low power wireless communication principles (e.g., scatter radio) and novel energy harvesting circuits, and more importantly,
- by balancing the WSN computation and communication load of (inherently parallel and distributed) asynchronous message passing algorithms (for inference), across various (distributed in space) WSN nodes.
AMPERE offers a methodology framework for reliable inference from unreliable, ambiently-powered WSNs, with bounded execution time, quantified convergence/correctness tradeoffs, careful modifications of message passing for efficient communication, exploitation of powerful asynchronous message passing algorithms (e.g., for clustering, signal de-noising/reconstruction), as well as hidden links between message passing algorithms and iterative numerical methods. Case studies in environmental sensing / agriculture and home automation will be examined with tremendous socioeconomic impact, while the design principles should accommodate other applications.
AMPERE attempts a concrete step from coming Internet-of-Things to future Internet-of-Things-that-Think with ambient energy.