Thursday, October 03, 2013

Would you implement adaptive control on your process?

Today I attended a half-day seminar at the Technical University of Denmark on L1 Adaptive Control Theory and Applications. The feature keynote speaker was professor Naira Hovakimyan from University of Illinois at Urbana-Champaign (In my younger days I was actually accepted into the Ph.D.  program at that institution, but a week earlier I had already accepted an invitation to study at University of Alberta in Edmonton, Canada. However, I do recall that the applications required me to described the content of each course I had been taking at the Technical University of Denmark). Dr. Hovakimyan's keynote was titled "L1 Adaptive Control and It's Transition to Practice", and she started by mentioning the crash of the X15 in 1967, which was found to be caused by a fault in the planes adaptive controller according to Dr. Hovakimyan. Then she mentioned the successful flight of X-36 exactly 30 years later also using an adaptive controller. Dr. Hovakimyan's group have successfully used  L1 adaptive controllers on 5.5% scaled airplane models, and plans are in place to go to 15% scaled models next year. From the Q&A after the talk it became very clear, that the  success was to a large degree due to the practical talent of the pilots and students working on the system. There is a large difference between theory of adaptive control and the practice of adaptive control. This is something, I also experienced when implementing a Kalman filter on an industrial polyethylene reactor in the early eighties on my first industry job.

After the keynote Jussi Hermansen gave a practical demonstration of the capabilities of the L1 adaptive flight controller on a small quad-copter. The L1 controller was able to stabilize the quad-copter during rapid (full trost) accent and decent even under the windy conditions at DTU this afternoon. Jussi even balanced one arm of the quad-copter on a bench while the other three was in ear, and showed the plane hovering just a centimeter above ground. The win conditions today could be considered equivalent to large disturbances to process plants, so I feel sure we will see L1 adaptive control applications in the process industries. Appearantly collaboration with both  Statoil and Schlumberger are already ongoing.

The seminar finished by work the Automation and Control group at DTU had been doing with the Danish Navy on control of un-maned power boat for towing shooting targets. Dr. Hovakimyan recommends the book she has written for learning L1 adaptive control, but my fealing is that there is more to practice than is in the book. So to be successful with L1 adaptive control the best path is properly to hire one of Dr. Hovakimyan's students.