When we imagine an “interconnected world,” we may think of the world the internet has already delivered: interconnected people and ideas. The next step forward is the interconnection of things: organizing communication not just between you and your dentist, but between your phone and your toothbrush.
How we would organize that world is still a rich question for researchers, corporations, and the people who use these products. The Internet of Things promises to revolutionize industries and our daily life, but how would a fully decentralized system — independent devices working as a collective system, capable of adapting to new additions or subtractions from the network — actually work?
That brings us to the world of “spatial edge services,” which is focused on linking collective, decentralized controlling devices (actuators) and sensors in such a way that they can gather, share and aggregate information between them. Right now, this can be sliced up a few ways. You can send the data back to a central service — the “cloud” for example — or you can rely on the device to sort the information itself.
For many, sending information you gather from your phone to Iowa to return to your television seems inefficient — security leaks, server lag, and other problems can creep in. That’s why researchers are turning to decentralized networks.
Houssem Ben Mahfoudh is a researcher at the University of Geneva, working with Dr. Giovanna Di Marzo Serugendo, head of the Institute of Information Service Science. He is exploring novel ways of distributing information amongst many devices that can work together as a collective system in collaborative ways. Ideally, these systems are decentralized in such a way that they can understand when a new sensor is added to the network. He sees this kind of network as applicable to a wide variety of use cases.
“They exploit spatially distributed data and enable smart environments,” writes Mahfoudh. “These services are provided as the result of a collective interactions among multiple agent-based entities, involving processes and calculations. It takes place across several geographically distributed computational nodes, using chemical, bio-inspired mechanisms.”
In the simplest terms, these devices are smart enough to talk to each other without needing to report back to a central organizing computer. Think of it as passing a note to a colleague working beside you, rather than walking to the post office to mail them a letter. Why send your data to a server somewhere miles away when you’ll only ever use that data locally?
The Future of Energy
The energy landscape is ripe for this kind of technology. The market is moving from one where “consumers” purchase energy to one where customers participate in an energy ecosystem, contributing excess power from sources such as solar panels. Future scenarios involve people’s choices of energy type (mostly green) and how to share energy among neighbors. To manage this changing infrastructure, sensors have to be adaptable and responsive.
Mahfoudh’s research points to the potential for the spatial services in a green energy ecosystem. Imagine that your neighbor has created excess energy through solar panels, and you’re looking to complement your own energy supply. A mesh network could see the local (in this case, your home’s) energy needs, determine excess energy in the nearby radius (your neighborhood) and intelligently redistribute that energy through a smart grid system.
To do this, the machines are looking at a variety of information sources — not only your energy surplus or demand, but also the asking price for excess energy. Mesh computing can do this elegantly, moving energy while determining a fair price amongst the suppliers, all with minimal engagement of the seller or buyer.
That’s beneficial for people who have solar panels and don’t want to waste excess energy, but don’t exactly want to monitor or track requests or sales of the extra energy they produce. By circumventing a mediator, such as a power company, and allowing the decentralized network to handle the workload, it can encourage energy redistribution on a small scale. Spread broadly, that could have an extreme impact on the future of clean energy.
The next step for these systems is helping them determine how to effectively manage decisions made from diverse data from a variety of sources. Mahfoudh has been looking at composing services on-demand using reinforcement learning, a technique for machine learning based, in part, on behavioral psychology.
It’s relatively easy to understand in principle, at least for humans: as we’re rewarded for actions, we learn to repeat those actions. Training a decentralized network is similar in principle, though obviously you can’t throw a piece of candy to an algorithm.
“Designing such a system would mean it could intelligently balance the exploration of new paths of composition with those that have historically provide the most reliable answers,” says Mahfoudh.
In other words, the system could try novel ways of finding answers, but not at the expense of reliability. That would mean new additions could plug into the network, which would easily adapt by integrating data in a way that makes its overall performance more precise. Such ease of interoperability could encourage more people to participate in producing and sharing the power they don’t use — creating a more efficient energy landscape, especially for green energy.
This article is part of swissnex San Francisco’s continuing investigation into the emergent future of energy and the environment. To stay up to date on the latest news, ideas, events, and insights from Switzerland and Silicon Valley, sign up for our monthly newsletter.
Photo: Houssem ben Mahfoudh, by Ray Potes for swissnex SF.