Related Experiment Video
Updated: Jul 7, 2026
11:06
Design and Fabrication of an Elastomeric Unit for Soft Modular Robots in Minimally Invasive Surgery
Published on: November 14, 2015
Identification of a two-link flexible manipulator using adaptive time delay neural networks
A Yazdizadeh1, K Khorasani, R V Patel
1Dept. of Electr. & Comput. Eng., Concordia Univ., Montreal, Que.
Summary
This study introduces adaptive time delay neural networks (ATDNNs) for identifying complex two-link flexible manipulators. These neuro-dynamic models successfully map nonlinear systems without prior knowledge or offline training.
Area of Science:
- Robotics
- Control Systems Engineering
- Artificial Intelligence
Background:
- Flexible manipulators are complex nonlinear systems crucial in robotics.
- Accurate system identification is essential for effective control.
- Traditional methods often require significant a priori information.
Purpose of the Study:
- To develop novel neuro-dynamic identifiers for a two-link flexible manipulator.
- To demonstrate the capability of adaptive time delay neural networks (ATDNNs) in nonlinear system identification.
- To provide methods for selecting network parameters and adaptation laws.
Main Methods:
- Proposed two neuro-dynamic identifier structures based on ATDNNs.
- Analytical demonstration of the networks' capability to represent nonlinear input-output maps.
- Generation of input-output data from a two-link flexible manipulator under nonlinear control for various trajectories.
Main Results:
- The proposed ATDNN structures successfully identified the highly nonlinear system.
- Identification was achieved without prior knowledge of system nonlinearities.
- No offline training was required, indicating efficient online learning capabilities.
Conclusions:
- ATDNNs offer a powerful tool for identifying complex, nonlinear robotic systems.
- The proposed neuro-dynamic identifiers are effective for real-time or adaptive control applications.
- This approach simplifies the identification process by eliminating the need for offline training and prior system information.
