*** NOTE ***
By Dawn Wiseman
The Governor General’s Gold Medal, awarded each year at fall convocation to Concordia’s most outstanding graduate student, usually goes to a PhD graduate. However, for the first time, this year’s winner, Javad Lavaei (Electrical and Computer Engineering), is being honoured for his master’s work. That alone should provide some sense of the importance of what he has demonstrated.
When Lavaei arrived at Concordia in 2005, he and his supervisor, Amir Aghdam, sat down to discuss his research. The conversation turned to a long-standing problem in control research: Developing a robust means of controlling multi-component systems that do not rely on a central control point. As an example, Lavaei offered the engineer’s dream of an automated highway system where cars are equipped with onboard electronic controls that act as drivers.
In a centralized control system, every car would communicate key data (position, velocity, etc.) with a central station somewhere on the highway. The station would then send back instructions about what to do. Lavaei pointed out this strategy is impractical for several reasons, “First, the communications overhead would be huge. Moreover, it introduces a single point of failure. So, if the station does not work properly, all cars would malfunction.”
In a decentralized system, the cars’ onboard controllers would communicate not with a central station but with each other. The best driving strategy would then be determined based on a simple set of rules such as: maintain a minimum distance of X metres between vehicles; if the car in front slows down, slow down too; move in the same general direction as surrounding traffic. This is how birds flock, so how hard can it be?
As it turns out, pretty hard. “Decentralized control has been a long-lasting intractable area in the sense that not much progress has been made in it.” So he and Aghdam decided there must be new and different ways to approach the problem.
What Lavaei did was break his research into eight sub-problems, each in itself was an important and difficult problem that had been explored for a number of years with no efficient solution. He came up with solutions for each, then brought them together to develop a comprehensive means to synthesize and analyze decentralized controllers.
More importantly, he demonstrated the work. “We successfully applied some of our ideas to the practical example of flight formation, where a set of unmanned aerial vehicles accomplished a mission cooperatively.”
As Aghdam underlined, the achievement is impressive. “This problem has been studied by the control community for about 25 years, and the method proposed by Javad is one of the most effective techniques so far. He is also the first researcher in the world to obtain necessary and sufficient conditions for robust stability of a class of uncertain control systems.”
With this success, Lavaei has moved on to doctoral work at the California Institute of Technology. There he is collaborating with researchers John Doyle and Steven Low trying to figure out how to optimize virtual traffic on the electronic highway.
The way information flows through the Internet is actually based on decentralized control, but as Lavaei underlined, “It is by no means optimal and its performance can be improved significantly. The reason is that the current architecture of the internet is built and engineered by computer science people, while the problem should really be tackled by control guys.”
So, if one day it becomes much quicker for you to upload photos to your Flickr or Kodak Gallery, you’ll know who you have to thank.