GA-tuning of Nonlinear Observers for Sensorless Control of Automotive Power Steering IPMSMs

Abstract

The paper considers two observer-based rotor position estimation schemes for sensorless control of interior permanent magnet synchronous motors (IPMSMs) for use in future automotive power steering systems. Specifically, emphasis is given to techniques based on feedback-linearisation followed by classical Luenberger observer design, and direct design of non-linear observers. Genetic algorithms (GAs), using the principles of evolution, natural selection and genetic mutation, are introduced to address difficulties in selecting correction gains for the observers, since no analytical tuning mechanisms yet exist. Experimental measurements from an automotive power steering test-facility are included, to demonstrate the enhanced performance attributes offered by tuning the proposed observer schemes, online, in this manner.