Chances are you’ve used Google Maps or a similar navigation application to help you get from one place to another.
But what if you had twenty spots to visit and you wanted your mapping tool to recommend the most efficient route? It would take a modern computer hours to solve the problem, if it was solvable at all. These are the kinds of issues facing organizations that work in supply chain and logistics each and every day.
This challenge is known as the Travelling Salesperson problem in computing and is used to highlight how we’re reaching the limits of what is possible with the computing technology we’ve relied on for decades.
In response to this challenge, researchers like Bhavin Shastri are working on a photonics-based computer chip which emulates the human brain and uses lasers, instead of ones and zeroes, to rapidly communicate information.
An Assistant Professor in the Department of Physics, Engineering Physics, and Astronomy in the Faculty of Arts and Science, Shastri is cross appointed to Smith Engineering.
“Every time the human race has gone forward, computing has actually taken us to that next level,” he says. “What my research is all about is trying to figure out how we can make this computing sustainable and how we can enable new forms of computing.”
Solving this problem is not just a ‘nice-to-have’ technology that will only matter to delivery companies. It will help average users of platforms like mapping and artificial intelligence (AI) technologies to get the information they need much more quickly and more efficiently from an energy perspective. Current chat-based AI platforms, for instance, require tens of thousands of computers and months of time to be trained to respond to user questions.
While modern computers tend to have separate parts like a processor core or a stick of memory, the neutral network chip is a proof of concept that is modeled on the human brain with dense connections between the different nodes of its ‘brain’.
“That interconnectedness between the nodes helps us with pattern matching, that part of the brain that helps me recognize you even if you’re wearing a scarf or glasses,” he says. “Our neural network chip replicates 100,000 nodes of the human brain and it can learn to recognize things a billion times faster than the human brain.”
Shastri’s research also connects to advances in quantum computing, which can also be used for computer-based problem solving but relies on probabilities. His aim is to bridge the gap between these different technologies using photonics-based computing to efficiently return definitive answers to problems like the Travelling Salesman.
Shastri was recently named a Tier 2 Canada Research Chair in Neuromorphic Photonic Computing, building on his existing research which unites him with researchers at Princeton University and has been backed by the Canada Foundation for Innovation, the Ontario Research Fund, and the Natural Sciences and Engineering Research Council of Canada. Research enabling faster and more efficient computing will be critical to realizing the goals of the federal government’s Pan-Canadian Artificial Intelligence Strategy and Canada’s National Quantum Strategy.
Shastri is one of 16 Canada Research Chairs announced for Queen’s in Fall 2024, including three professors affiliated with Smith Engineering: Cao Thang Dinh, Amir Fam, and Lidan You.