Multi-muscle synergy-based control of a robotic device: a promising approach for helping people with movement disorders

Human–robot interfaces may play a key role in helping people with motor impairments, especially children, to achieve mobility, manipulation, and functional communication. Therefore, the implementation of an accurate, smooth, and flexible control interface can positively impact the quality of life of people with movement disorders. Electromyographic activity (EMG) has often been used as a control interface for robotic devices (myocontrol). However, it has proven difficult even for healthy adults to learn to use EMG to control multiple degrees of freedom. One common hypothesis in the field of motor control is that movements are produced through muscle synergies, coordinated activations of groups of muscles with specific balances of muscle activations. Whether muscle synergies are indeed a simplifying control strategy actually implemented by the CNS is still a debated issue. Nevertheless, many recent studies indicate that muscle activations can be represented into a low-dimensional space as a linear combination of muscle synergies. The aim of this work is to achieve real-time multi-muscle control of a robotic device by extraction of controllable synergies. We have developed and tested, in healthy subjects, a real-time controller of an external actuator, based on synergies extracted from 8 muscles of the upper limb. An ad-hoc algorithm was designed to process and analyze EMG and to perform torque-based control of one degree of freedom of a desktop robot at 1000 Hz (Phantom Omni, SensAble TM ). To tailor the interface to the specific motor patterns of each subject, synergies are identified using nonnegative matrix factorization (NMF) for each subject individually. For each time frame, the driving signal is based on the similarity between the vector of EMG and the vector of each direction-related synergy. We expect muscle synergies to capture regularities in the noisy sensorimotor mapping, thus allowing an easier and smoother real- time control even in subjects affected by movement disorders. The next step is to test this control interface on children with movement disorders, and to compare synergy-based with single-muscle control