Research on morphology deformation prediction and compensation of fusion lightweight manipulator

This paper proposes a fusion lightweight manipulator (FLM) morphology deformation prediction and compensation algorithm for the precise maintenance of the fusion reactor. Firstly, the FLM links are flexibly modified using nylon material, and the total load deformation error of FLM is close to 30 mm through static analysis. The FLM fiber grating sensing system is designed to measure the deformation error of the FLM prototype. Then, the FLM morphology deformation prediction control system framework is constructed. Through the decoupling calculation and virtual-real fusion of FLM deformation data collected by sensors, the predictive control of FLM morphology deformation is realized. To compensate for the positional error of the FLM, the deformation prediction neural network (DPNN) model of the FLM is proposed. The joint error parameters of FLM can be obtained by training the morphological deformation dataset collected by the FLM sensing system through the coupling of multilayer perceptron, transformer, and multi-attention mechanism. Finally, the flexible FLM experimental platform and data collection system are established, and the experimental results show that the predicted position errors of the flexible FLM end after compensation are less than 0.5 mm, which proves the effectiveness of the morphological deformation prediction control and compensation algorithms. • Proposing a fusion lightweight manipulator morphology data collection method. • Proposing a morphology deformation prediction control framework. • Proposing an error compensation algorithm based on a deformation prediction neural network. • The effectiveness of the algorithm is verified by prototype experiments.