Identification of Human Shoulder-Arm Kinematic and Muscular Synergies during Daily-Life Manipulation Tasks

Understanding the inherent synergistic nature of human neuromuscular control has already successfully contributed to the development of human-centered robotic systems. To further extend this line of research, a thorough understanding of the synergy spaces in which human movements are planed and executed is necessary. In this paper, the shoulder-arm kinematic and muscular synergies for typical 30 daily-life tasks are identified. For this, an experimental dataset of multi-joint motion trajectories and surface electromyograms of 6 healthy male subjects was created. Based on this, the synergies are identified by applying state-of-the-art machine learning techniques. The identification results suggest 1) synergy space dimensionality correlates with task complexity, 2) 3-D synergy spaces are sufficient to explain ≈95% variance, 3) presumably, each task is mainly executed in its own kinematic and muscular synergy space, and 4) the similarity in synergy coordinates is indicated to correlate to similarity in joint space. Subsequently, these synergy spaces shall be integrated into human-inspired controller design of robotic systems for improved rehabilitation, assistance, and more human-like prosthetic devices.