Unsupervised Neural Network Based Forward Kinematics for Cable-Driven Parallel Robots with Elastic Cables

Abstract

Forward Kinematic (FK) analysis of under-constrained Cable-Driven Parallel Robots (CDPRs) deals with the inherent coupling between the loopclosure and static equilibrium equations. The non-linearity of the problem is magnified with the addition of the coupling between the cable lengths and their tensions based on the elastic cable model. The paper proposes an unsupervised neural network algorithm to perform fast forward geometrico-static analysis for CD-PRs in a suspended configuration with elastic cables. The formulation determines a non-linear function approximation to model the FK and proves to be efficient in solving for consecutive and close waypoints in a path. The methodology is applied on a simulated six-degree-of-freedom (6-DOF) spatial under-constrained suspended cable-driven parallel robot. Specific comparison results to show the effectiveness of the proposed method in tracking a given path are presented against the results obtained from least-square non-linear optimization.

Forward Kinematics for Suspended Under-Actuated Cable-Driven Parallel Robots With Elastic Cables: A Neural Network Approach

“…The overall performance for the proposed algorithm and lsqnonlin is compared based on the time efficiency and accuracy. Extending from our previous contributions [22,23], we extend the approach to validate on hardware with the setup as shown in Fig. 1 for a particular set of poses from the pre-planned simulated trajectory.…”