CafkNet: GNN-Empowered Forward Kinematic Modeling for Cable-Driven Parallel Robots

When deploying Cable-Driven Parallel Robots (CDPRs) in practice, one of the challenges is kinematic modeling. Unlike serial mechanisms, CDPRs have a simple inverse kinematics problem but a complex forward kinematics (FK) issue. Therefore, the development of accurate and efficient FK solvers has been a prominent research focus in CDPR applications. By observing the topology within CDPRs, in this letter, we propose a graph-based representation to model CDPRs and introduce CafkNet, a fast and general FK solver, leveraging Graph Neural Network (GNN). Extensive experiments are conducted on 3D and 2D CDPRs across various configurations, including under-constrained, fully-constrained, and over-constrained cases, in both simulation environments and real-world scenarios. The experimental results showcase that CafkNet can learn the internal topological information of CDPRs and accurately solve the FK problem as an FK solver. Furthermore, training the CafkNet model on partial configurations enables zero-shot generalization to other configurations. Lastly, CafkNet effectively bridges the sim2real gap by using both simulation data and part of real-world data. To the best of our knowledge, it is the first study that employs the GNN to solve the FK problem for CDPRs.