- 20 views
- 21 downloads
Control of Sensorless Articulated Manipulators using Vision-based Estimation Techniques
-
- Author / Creator
- Singh, Charanjot
-
Conventional control of an articulated manipulator such as robot arms involves the use of sensor measurements of joint values to calculate the position and orientation of the end-effector and to perform motion control. Conversely, in cases where direct sensing is not available, a vision-based method can prove to be advantageous. This thesis deals with one such system, an excavator used regularly at construction sites, with the long-term goal of enabling fully autonomous operation. For our hardware experiments, we used a scale excavator model inside a lab setting, using a single external monocular camera to estimate the joint states of the vehicle in real-time and using the resulting information to run motion planning.
Our solution employs a computer vision algorithm integrated with numerical optimization to perform 2D and 3D pose estimation, respectively, from single RGB image frames in real-time. We use a limited collection of real images for training, in contrast to other approaches presented in the literature which require a CAD model to generate a large set of synthetic data. We leverage geometric constraints of the manipulator and the camera's optical parameters within our iterative optimization step. We also build a custom microcontroller-based system to drive the scale excavator from a desktop computer. The experimental results are benchmarked against a high-fidelity optical motion capture system installed in our lab to quantify the accuracy of our estimates. Since our proposed method relies only on a small training dataset and geometric information about the manipulator, it can readily be ported onto larger manipulator systems, such as large industrial excavators or cranes found on real construction sites.
-
- Subjects / Keywords
-
- Graduation date
- Fall 2023
-
- Type of Item
- Thesis
-
- Degree
- Master of Science
-
- License
- This thesis is made available by the University of Alberta Libraries with permission of the copyright owner solely for non-commercial purposes. This thesis, or any portion thereof, may not otherwise be copied or reproduced without the written consent of the copyright owner, except to the extent permitted by Canadian copyright law.