A vision that starts with a bridge: Introducing Hypermedia Multi-agent Systems
Imagine this: A worldwide hypermedia fabric stretched across interconnected physical and digital things, where people and autonomous software agents coexist and interact with one another.
As humans and machines (so-called ‘agents’) observe, reason about and act on these physical and digital things via physical, digital or holographic interfaces, they can create high-level services by mashing-up what they find in their environment –orchestrating the affair in a flexible, open, and scalable Web (of Things).
The system is undergirded by the World Wide Web, which users can navigate with the ease it takes to input a query into a Google search engine. It brings value to a breadth of domains and improves the way humans work, live, learn and play. Users can operate intuitive systems in which a multitude of interconnected devices tick along and take up very little of their attention.
Fields such as manufacturing, healthcare, agriculture and logistics stand to benefit from this increasingly autonomous technology. But its impact stretches even further –it could help users more astutely navigate the social web, for example, warning them of misinformation and emotional contagion as and when it arises.
This is the futuristic vision outlined in a paper produced in 2018 by a team of researchers at the University of St.Gallen that includes computer scientist Dr. Simon Mayer, which describes how creating a bridge between research in Web-based multiagent systems will help create a world in which work that is arduous, dull or dangerous is performed entirely by machines.
“We want to combine technologies from the Web architecture field with a domain called multiagent systems,” he says. “The problem is that these fields have operated for the last 15 to 20 years without talking too much to one another,” he adds.
“This means that the advancement we have seen in Web architecture has not precipitated in the field of multiagent systems, and vice versa.”
Designing the future human-machine collaboration
The energetic and engaging Dr. Mayer serves as Chair of Interaction and Computer-based systems at Switzerland’s University of St.Gallen, one of IntellIoT’s thirteen partners.
St.Gallen’s chapter will coordinate tasks around Hypermedia Multi-agent Systems (HyperMAS)– while also supporting human-machine collaboration across the three use cases: healthcare, manufacturing, and agriculture.
The department, and Dr. Mayer, were excited to jump onboard the IntellIoT mission to enable IoT technologies that will power the future of these use cases, not least because its goals chimed those of Dr. Mayer’s research group.
“We are now investigating interactions between many machines, and between machines and humans,” he tells Next Generation IoT, explaining the group’s motivation for joining and contributing to IntellIoT.
This fostering of intricate ‘machine-to-machine’ collaboration is a key objective for IntellIoT, as is its ‘human-in-the-loop’ dimension. That is, the work of creating a system that appeals for human support and direction as and where needed, and which learns from that process.
“Autonomous collaboration of devices is just fascinating,” says Dr. Mayer. “And it’s terribly hard to do,” he adds, careful to qualify that statement with the caveat that the challenge lies in orchestrating such collaborations at scale.
When many agents work together, a tremendous number of options presents that Dr. Mayer’s team believes can be managed with the help of a HyperMAS that coordinates actions within the system. And as such, this work hinges on the creation of this bridge between hypermedia and multiagent system technologies spearheaded by Dr. Mayer’s research group at the University of St.Gallen.
“This is what drives me personally,” Dr. Mayer says. “How can one device broadcast its capabilities in a way that can be understood by other devices? And how can this device be matched automatically with the requirements of other services? And this extends to humans. Maybe the system knows about their health state, or about their agenda today. And can then infer what they require and make the match.”
When Hyper met MAS
For the uninitiated, the term ‘multiagent system’ (MAS) refers to a computerised system made up of multiple interacting agents designed to solve tasks that cannot be completed alone or without an element of autonomy on the part of the agent.
The theory behind it stretches back to the fifties, when mathematicians and scientists first started dreaming up visions of how a world powered by ‘Artificial Intelligence’ might look, explains Dr. Andrei Ciortea, who works alongside Dr. Mayer on the project.
Dr. Ciortea describes how, in 1956, academics gathered at Dartmouth College in New Hampshire to discuss ideas around “thinking machines.” The Dartmouth workshop is largely considered the founding event for artificial intelligence (AI) as a field. The concept and conversation around AI has since then grown far more complex, convoluted by increasingly broad conceptualisations of what counts as AI — especially since the recent popularisation of the term.
“Us computer scientists have obviously lost the battle of who creates the definitions around here. And that’s the way it is,” says Dr. Mayer, with a shrug and caveat that he still sees the global prevalence of the subject in public discourse as a blessing for computer science, despite the confusing misuses of the terminology involved.
Dr. Mayer adds that while many understand the terminology of Artificial Intelligence to refer to the topic of machine learning, or human-like intelligence simulated by a machine, the roots of the field lie strongly in the work of exploring the opportunities that come from multiagent systems ‘that we might call autonomous’ –combining technologies together such that the boundaries between research fields blur.
“We have seen computer science enter many, many fields, over the last twenty years,” says Dr. Mayer. “And through these fields — and because we are using computers to enable increasingly complex behaviors, no individual can anymore explain every aspect of an end user application that is based on these technologies.”
This complexity touches on some areas of contention around the rise of these increasingly sophisticated autonomous systems, notes Danai Vachtsevanou, MSc., who is a PhD candidate in Dr. Mayer’s team, and explains that there will be queries around the legitimacy when machines are bestowed with capabilities so intricate humans can no longer fathom them.
What’s more, the greater the number of devices collaborating with one another, the greater the challenge of intelligibility. And, as an increasing number of machines are empowered to autonomously operate in unison and at scale, finding and agreeing upon a shared language among all these devices is proving a difficult exercise to manage.
Managing these challenges requires extensive work in making sure there exists sufficient transparency and regulation around such technologies. All the more reason for expansive research projects like IntellIoT, that foster the cross-pollination of ideas and standardisation across an ecosystem championing open-source sharing and utmost transparency.
Demonstrators, glorious demonstrators
Through the course of our interview, Dr. Mayer’s passion for the subject at hand takes expression as much in the sprawling diatribes that touch on several interlocking themes, from philosophy and social sciences and economics, to lively deep dives into the more nitty-gritty side of the technology, pulling out smartphones, and other devices to demonstrate his concepts.
His enthusiasm very much stretches to the role St.Gallen’s research can play in helping transform healthcare, manufacturing and agriculture for the better. Research that will help engineer workplaces where tasks that are boring, tedious and dangerous are performed by robots, and where the human’s needs and comforts come first. Or, as Dr. Mayer puts it; “Machines as tools for humans, not the other way around”.
Of the various projects that make IntellIoT’s expansive programme, he is very excited about creating three demonstrators for the manufacturing use case. There will, for instance, be collaborative manufacturing lines in which robots can be swapped in and out without affecting the behavior of the overall system. He’s especially thrilled about what it will mean to achieve such feats and put them on show.
“Such systems have been made on a small scale. Already in the past, my team has been demonstrating such systems for a handful of participating devices. But what we are setting out to do here is supposed to work for a much larger number of devices,” he says, explaining that accomplishing this means harnessing the power of the Web architecture to create something unprecedented.
“This will be a masterpiece. I mean, just think of the demonstration –once we can show that we can have a partially autonomous manufacturing line that can flexibly adjust to changing context and even request human help when required, and then learn from the human so that it won’t run into similar trouble in the future. It will be glorious, and it will be an excellent showcase of what we are doing.”
The University of St.Gallen boasts a pioneering Masters programme in Computer Science. Learn more about it here.
Dr. Simon Mayer’s team at the University of St.Gallen is one of 13 partners in IntellIoT, a three-year-long European Commission-backed research and innovation project enabling AI and IoT technologies in manufacturing, healthcare and agriculture. The team will coordinate around the Hypermedia Multi-agent Systems (HyperMAS) and support the project’s target use case of human-machine collaboration in manufacturing, agriculture and healthcare.
For more materials and updates regarding the IntellIoT project, subscribe to our newsletter.
