Improved Stewart platform state estimation using inertial and actuator position measurements. (May 2017)

• Record Type: Journal Article
• Authors: Miletović, I.; Pool, D.M.; Stroosma, O.; van Paassen, M.M.; Chu, Q.P.
• Is Part Of: Control engineering practice. Volume 62(2017)
• Page Start: 102
• Page End: 115
• Publication Date: 2017-05
• Subjects: Sensor integration -- State estimation -- Unscented Kalman Filter -- Parallel robotics -- Stewart platform; Automatic control -- Periodicals; 629.89
• DOI: 10.1016/j.conengprac.2017.03.006
• Languages: English
• ISSNs: 0967-0661

Abstract

Abstract: Accurate and reliable estimation of the kinematic state of a six degrees-of-freedom Stewart platform is a problem of interest in various engineering disciplines. Particularly so in the area of flight simulation, where the Stewart platform is in widespread use for the generation of motion similar to that experienced in actual flight. Accurate measurements of Stewart platform kinematic states are crucial for the application of advanced motion control algorithms and are highly valued in quantitative assessments of simulator motion fidelity. In the current work, a novel method for the reconstruction of the kinematic state of a Stewart platform is proposed. This method relies on an Unscented Kalman Filter (UKF) for a tight coupling of on-platform inertial sensors with measurements of the six actuator positions. The proposed algorithm is shown to be superior to conventional iterative methods in two main areas. First, more accurate estimates of motion platform velocity are obtained and, second, the algorithm is robust to inherent measurement uncertainties like noise and bias. The results were validated on the SIMONA Research Simulator (SRS) at TU Delft. To this end, an efficient implementation of the algorithm was driven, in real time, by actual sensor measurements from two representative motion profiles. Abstract : Highlights: A novel kinematic state reconstruction algorithm for Stewart platforms is proposed. Inertial and actuator position sensors are fused using the Unscented Kalman Filter. A real-time validation of the method on an operational Stewart platform is presented. The method allows inference of platform speed without numerical differentiation. The method is robust to measurement uncertainties like noise and bias.