Compared to conventional rigid robotic manipulators and unlike traditional industrial robotic manipulators, flexible manipulator can be more efficient and safer to operate due to their low power consumption and lightness. Moreover, the next generation of robots will require compliance to be able to perform safe and close interactions with human beings. However, due to the flexibility in arms and joints, flexible manipulator control and operation suffers from severe structural vibration, leading to the inaccuracy in end-effector trajectory tracking and visual servoing.
This research work aims to address the problem by developing a systematic motion estimation methodology for flexible robotic manipulators by using vibration and visual sensing. The aim was achieved by addressing several research objectives that include the investigation of the dynamics characteristics of a flexible robotic manipulator for the end-effector’s motion estimation, the study of decentralized vibration control of a multi-link flexible manipulator with piezoelectric sensor, the development of motion estimation framework experimentally through a single-link flexible manipulator test-rig and also the development the computer vision based motion estimation method for a flexible robotic manipulator with an eye-in-hand configuration.
The dynamic characteristics of the flexible manipulator were investigated for development of the end-effector’s motion estimation. This was obtained by incorporating the modelling of flexible manipulator with embedded smart piezoelectric transducer based on co-rotational finite element method. This modelling method allows the simulation of non-linear dynamic motion of a flexible multi-link manipulator undergoing not only large rotations and translations of the manipulator, but also potentially large non-linear structural deformations of flexible links. Numerical simulations were performed on single-link and two-links flexible manipulator embedded with smart piezoelectric transducers. The results have demonstrated that this model could be effectively used to study the motion of end-effector for a multi-link flexible robotic manipulator undergoing complex large motions, while allowing the use of embedded piezoelectric transducers for measuring its structural vibration. In order to suppress the structural vibration of flexible manipulator, a decentralized control scheme based on the voltage rate feedback using pairs of collocated piezoelectric actuator/sensor has been utilized. The numerical case studies of single-link and two-links manipulator have been performed respectively. A special consideration is given for finding the best location for piezoelectric transducers over the multi-link flexible system.
Generally speaking, motor encoder measures rigid body motion of the manipulator while vibration signals measure the flexible body motion of flexible manipulator. Thus a Kalman filter is developed as to estimate the motion of end-effector of flexible manipulator with the signals of piezoelectric sensor and motor encoder. The effectiveness of the proposed estimation method is evaluated and demonstrated by the developed co-rotational finite element model. The benefit of this method is that although the motion of flexible manipulator can be highly non-linear due to large rotations, the use of a linear Kalman filter, combined with the encoder information, may be sufficient for the estimation purpose for determining the motion of end-effector.
On the other hand, robotics vision sensing has been intensively studied from various aspects in the last decade due to the increasing computational power available for processing images. This work proposes a visual-based motion estimation method for flexible robotic manipulator with an eye-in-hand configuration. The object tracking is proposed based on the autocorrelation method that measures the similarity of features between consecutive frames thus the estimation of corresponding apparent motion. The apparent motion is then combined with the geometry of the calibrated camera allowing deriving the angular position of the flexible arm at high frequency with high accuracy. This kind of method requires minimum computation and suitable for fast vision computation applications.
Experimental results performed on the single-link flexible manipulator test rig demonstrated the feasibility of its application in flexible manipulator with eye-in-hand system. The developed vision sensing method obtains more precise information of the end-effector location and it is easy to extend to multiple flexible-link problems. The benefit of using this vision sensing method is that it can provide a direct measurement method without the need of any complex transformation procedures and also can be easily extended to multiple-link flexible manipulator cases without additional measurement devices or sensors.
|Date of Award||1 Jul 2017|
- Univerisity of Nottingham
|Supervisor||Dunant Halim (Supervisor), David Siu-Yeung Cho (Supervisor) & David Branson (Supervisor)|
- motion estimation
- flexible manipulator
- structural vibration
- computer vision