In this study, we designed and implemented a pneumatic-driven translational pyramidal manipulator (PTPM) to provide reliable and accurate motion control. The designed PTPM is compact, has a high weight-to-force ratio, and does not exhibit an air leakage problem. An observer-based intelligent controller, namely the linear active disturbance rejection controller (LADRC)-reduced-order linear extended state observer (RLESO)-radial basis function neural network optimizer (RFNNO), was developed for the designed PTPM. This controller has the following advantages: (1) it rejects total disturbances and coupling effect, (2) it reduces the order of the extended state observer, (3) it selects observer gains according to system frequencies, and (4) it optimizes controller gains in real time. The experimental results indicated that the PTPM controlled using the LADRC-RLESO-RFNNO increased the robustness in external disturbance rejection and provided accurate trajectory tracking.