Computationally Efficient IMU-Based Modeling of a Flexible Manipulator for Subsystem-Based Control

Abstract

Flexible manipulators offer high power-to-weight ratios and are well suited for challenging environments, but their control is complicated by structural flexibility. This paper presents a practical modeling and control framework that approximates the flexible link as an open chain of rigid segments, using data from an inertial measurement unit (IMU) sensor network. This replaces complex partial differential equation-based models with a tractable set of ordinary differential equations, enabling real-time implementation. A complementary filter recovers virtual system states from IMU data, despite noise and drift effects. Based on this model, we propose a robust subsystem-based model predictive control strategy for task space control. The controller is computationally efficient and avoids the complexity of controlling a high degree-of-freedom open chain based on the system’s modular structure. Experimental results using a long-reach flexible manipulator demonstrate improved tracking accuracy and smoother input generation, validated against laser-based position and force cell sensor measurements.