Optimization Based Control

Professor Scott Bortoff
United Technologies Research Center
(on sabbatical leave from the University of Toronto)

Control systems based on real-time on-line optimization, such as Model Predictive Control (MPC), can provide several advantages over other model-based methods. Perhaps the most important of these is that constraints on the control, state and output can be incorporated directly into the design. Constraints are often the dominant characteristic in an industrial control problem. i

This talk will focus on two applications of OBC on United Technologies products. First, we will study the problem of path planning for an Unmanned Air Vehicle (UAV), such as the Sikorsky Cipher. We will frame this as an optimal control problem which is constrained by the UAV dynamics and control rates and magnitudes. The objective function is a trade-off between flight distance to a known target, and radar reflectivity to enemy radar sites (stealth). This problem can be solved in closed form for some simple cases. The UAV path planning problem will also be posed and solved using potential field theory. The second application of OBC is for gas turbine engines. This problem is dominated by constraints on the state and inputs, and optimization is important for purposes of efficiency and wear reduction.