abstract
- The central aim of this study was to develop a virtual robotic trainer platform for teaching Mexican Sign Language (MSL) to deaf and non-deaf people. The MSL-Virtual Robotic Trainer (MSL-VRT) was designed to mimic a human upper limb (hand, forearm, and wrist). It has 25 degrees of freedom (DOF) to perform finger-spelling/dactylology of words, considering one-handed static signs (21 signs) of the MSL alphabet. The design of MSL-VRT uses Solidworks¿, and the assembly was then exported to Simulink¿/Simscape¿ included in Matlab for numerical evaluation of the control action that regulates the MSL-VRT motion. The controller design uses adaptive proportional¿integral¿derivative (PID) gains, with the derivatives of the tracking errors estimated by the super-twisting algorithm (STA). The gains of these controllers were adjusted to achieve the designed reference trajectories, which were composed of a superposition of sigmoid functions. The gains are adjusted adaptively to ensure convergence of the tracking errors and provide an accurate reproduction of the signs in the MSL-VRT. A numerical evaluation confirms the tutoring device's ability to reproduce the character sequence from texts. A comparison with the traditional non-adaptive PID was included to verify that the adaptive form performs better in the tracking trajectory task, leading to a better tutoring outcome.