Intention-detection strategies for upper limb exosuits: model-based myoelectric vs dynamic-based control

Abstract

The cognitive human-robot interaction between an exosuit and its wearer plays a key role in determining both the biomechanical effects of the device on movements and its perceived effectiveness. There is a lack of evidence, however, on the comparative performance of different control methods, implemented on the same device. Here, we compare two different control approaches on the same robotic suit: a model-based myoelectric control (myoprocessor), which estimates the joint torque from the activation of target muscles, and a dynamic-based control that provides support against gravity using an inverse dynamic model. Tested on a cohort of four healthy participants, assistance from the exosuit results in a marked reduction in the effort of muscles working against gravity with both control approaches (peak reduction of 68.6±18.8%, for the dynamic arm model and 62.4±25.1% for the myoprocessor), when compared to an unpowered condition. Neither of the two controllers had an affect on the performance of their users in a joint-angle tracking task (peak errors of 15.4° and 16.4° for the dynamic arm model and myoprocessor, respectively, compared to 13.1o in the unpowered condition). However, our results highlight the remarkable adaptability of the myoprocessor to seamlessly adapt to changing external dynamics.