Abstract
We present a network-augmented inverse-kinematics and control framework for a two-actuated, five-bar spherical parallel manipulator. Using spherical geometry, we derive compact forward and inverse kinematic relations and show that the mechanism admits four discrete IK branches for any target orientation. A Deep Operator Network (DeepONet) is trained to regress branch-independent primitives across the workspace and combined with a branch-aware selection policy that enforces continuity, respects joint limits, and avoids singular neighborhoods. Deployed in closed-loop control, the learned model generates joint trajectories from orientation commands. Simulation results demonstrate accurate joint prediction (mean/max 0.0002/0.0007 rad) and precise spherical tracking (mean/max 0.0708/0.2994 mm), while maintaining stable branch behavior.
| Original language | English |
|---|---|
| Title of host publication | 2025 7th International Conference on Control and Robotics, ICCR 2025 |
| Publisher | Institute of Electrical and Electronics Engineers Inc. |
| Pages | 67-71 |
| Number of pages | 5 |
| ISBN (Electronic) | 9798331558765 |
| DOIs | |
| State | Published - 2025 |
| Event | 7th International Conference on Control and Robotics, ICCR 2025 - Kyoto, Japan Duration: 04 12 2025 → 06 12 2025 |
Publication series
| Name | 2025 7th International Conference on Control and Robotics, ICCR 2025 |
|---|
Conference
| Conference | 7th International Conference on Control and Robotics, ICCR 2025 |
|---|---|
| Country/Territory | Japan |
| City | Kyoto |
| Period | 04/12/25 → 06/12/25 |
Bibliographical note
Publisher Copyright:©2025 IEEE.
Keywords
- DeepONet
- inverse kinematics
- redundancy resolution
- spherical parallel manipulator
Fingerprint
Dive into the research topics of 'DeepONet-Assisted Inverse Kinematics and Branch-Aware Redundancy Resolution for a Five-Bar Spherical Parallel Manipulator'. Together they form a unique fingerprint.Cite this
- APA
- Author
- BIBTEX
- Harvard
- Standard
- RIS
- Vancouver