Neural Network-Based Inverse Kineto-Static Analysis of Cable-Driven Parallel Robot Considering Cable Mass and Elasticity

Ishan Chawla,Pushparaj Mani Pathak,Leila Notash,Arun Kumar Samantaray,Qingguo Li,Umesh Kumar Sharma

DOI: https://doi.org/10.1007/978-3-030-75789-2_5

2021-01-01

Abstract:Inverse kinematic problem of a cable-driven parallel robot (CDPR) is a straightforward problem for massless inextensible cables. However, the problem becomes a non-linear kineto-static problem with the consideration of cable mass and elasticity. This problem is conventionally solved using numerical methods. These methods are iterative in nature, have slow convergence speed, and are highly dependent on initial values, which makes them difficult to be used in real-time applications. Therefore, this paper proposes a deep neural network-based approach to obtain the inverse kineto-static (IKS) solution of CDPR considering cable mass and elasticity. The training dataset has been obtained from IKS solutions attained by using numerical methods. Then, appropriate parameters as well as training algorithm is utilized for off-line training of the network. The performance of the proposed approach is validated by the simulation results for a redundantly constrained planar CDPR. From the simulation results, it has been observed that the proposed approach shows an 83% reduction in computational time in comparison to conventional numerical methods. In addition, the proposed neural network-based approach has performed satisfactorily with a minor error in the IKS solution. Therefore, the proposed approach can be employed for real-time applications due to its low and bounded computational time.