Pneumatic artificial muscle(PAM) H2-based metal hydride (MH) is considered a compact inherent soft actuator. To aim at a high-performance manipulated-PAM based MH actuator, the bottleneck of improving the performance lies in the parametric and nonlinear uncertainties occurred by an unpredictable environment in addition to an inherent nonlinear dynamics of PAM and a large scale dimension of MH. We develop the parameter-based adaptive robust control framework to cope the various operations, where a data-driven learning-based approach is dealt with the parameter adaptation. The effectiveness of our proposed approach is demonstrated through extensive experiments in terms of position and tracking control.