Pose estimation and passivity-based control of over-actuated flexible robots.

Abstract

Flexible robotic manipulators often operate under a high degree of model and parameter uncertainty. This uncertainty complicates the synthesis of robust controllers and accurate estimators for these systems. The first part of this dissertation seeks to address these challenges as they appear in the pose estimation for cable driven parallel robots (CDPRs), which are parallel manipulators that use deformable elastic cables. This part is split into three chapters. The first derives forward kinematic (FK) pose estimators assuming rigidly taut cables, along with an error covariance estimate for sensor fusion. The next chapter builds on this work by deriving extended Kalman filters (EKFs) that fuse these FK algorithms with IMU measurements that directly measure the payload motion subject to the elastic dynamics of the cables, improving the estimation accuracy. The next chapter presents experimental results with another EKF framework that uses direct measurements of the rigid-assumed cable lengths instead of FK computations. This allows for a more computationally-efficient and accurate pose estimation algorithm in the presence of elastic cable deformations. The uncertainty inherent to many flexible manipulator models also complicates the synthesis of robust flexible manipulator controllers. Passivity-based control is a method that can be robust to parameter uncertainty. While flexible manipulator systems are typically not passive, µ-tip control can be used to slightly augment the output of these systems to make them passive under the assumption that the payload is massive. This allows for the use of passivity-based control on flexible manipulators with relatively massive payloads. The second part of this dissertation generalizes prior work in µ-tip control to be applicable to over-actuated systems, which is something that previous µ-tip formulations were not capable of. Discussions on the application of the proposed controller to parallel manipulators, serial manipulators, and collaborative manipulators are presented, and numerical results for manipulators in each of these subclasses are also included.