Maintaining and renewing our infrastructure will require a considerable investment in the next decades. For example, most of Quebec bridges were built in the sixties and have already exceeded half of their expected service life. Environmental concerns are also imposing major changes to our energy or transportation systems. At the same time, modern sensing, computing and communication technologies provide a tremendous opportunity to enhance the capabilities of our infrastructure rather than merely repair it. GERAD researchers are playing a key role in fundamentally rethinking how to operate our future infrastructure with increased efficiency and reliability.
Infrastructures as Large-Scale Decentralized Control Systems
The power grid is equipped with feedback control loops that match generation and demand in real-time. A major current challenge is to integrate an increasingly large portion of renewable energy sources, which are much more unpredictable than generation from fossil fuel power plants. However, new sensing and control points are becoming available through smart grid programs. Smart meters measure demand more precisely and demand response schemes could help reduce peaks in consumption by directly controlling individual devices in homes, e.g., to spread the charging of a fleet of electric vehicles during the night.
To address such challenges, we are designing decentralized control schemes that can integrate millions of sensing and actuation points without overwhelming the communication infrastructure. Similar large-scale control issues arise if one attempts to optimize the congestion on a road network by coordinating traffic lights based on real-time traffic data measured from a variety of sensors, from those embedded in the pavement to smartphones sending drivers’ location data.
The Emergence of Autonomous Robots
Advances in robotics and machine perception in the past decade are now allowing robots to operate outside of the tightly controlled environments of laboratories and factory floors. Self-driving cars represent the most visible aspect of this revolution. They will lead to deep transformations of our transportation system and our relationship to car ownership and play an important part of mobility-on-demand systems. Another example is the use of mobile robots such as drones to autonomously inspect infrastructures such as power lines, bridges or buildings. GERAD develops mathematical methods for decision making under uncertainty, which are required to develop real-time planning capabilities for these robots.
Finally, one critical aspect of an intelligent infrastructure is that it supports the activities of a population of users, which imposes strong constraints on the type of sensing and decision-making algorithms that we can develop. First, predicting how these users will react to a given system being put in place, e.g., predicting the impact of a new traffic information or regulation scheme on overall congestion, requires tools from game theory. Second, there are concerns about the invasiveness of the data collection practices envisioned as the foundation of intelligent infrastructures, and the detrimental effects of the resulting loss of privacy. Here again, mathematical methods developed at GERAD can be used for the privacy-preserving analysis of real-time data, providing formal privacy guarantees to the users.