Walking and scuba diving assisted amphibious exoskeleton robots: the designing of power assist control and myoelectricity based wearers' fatigue evaluation

Exoskeleton robots have the potential to augment human motor capabilities. however, current control strategies often require task-specific control laws tailored for different scenarios, which limits the applicability of exoskeletons. In this study, we propose a control strategy for exoskeleton robots that is adaptable across various scenarios. We employ adaptive oscillators (AO) with feedback control to rapidly estimate the wearer's motion phase and subsequently provide torque assistance to the wearer's hip joint based on a TCN-LSTM model. During experiments, we collected surface electromyographic (sEMG) signals from the tibialis anterior, gastrocnemius, and rectus muscles of seven groups of subjects performing treadmill walking and inclined treadmill exercises. We utilized the short-time Fourier transform to extract frequency characteristics of the signals and statistically analyzed the rate of frequency change in each muscle group under different strategies. The results indicate that when wearing the exoskeleton, the overall muscle frequency changes more slowly, suggesting that subjects can maintain activity for a longer duration before fatigue sets in. This control strategy effectively reduces the energetic cost of lower limb work for the wearer and enhances the exoskeleton's versatility in various applications.