The need to identify the best treatment options for upper limb (UL) recovery is included in the top 10 priorities relating to life after stroke. Although robotic devices and Functional Electrical Stimulation (FES) have shown to be valid training modalities per se, their combination in hybrid systems can overcome the limitations of every single approach. Training assisted by hybrid devices in lifelike scenarios, allowing the subject to freely interact with the environment, may optimize the transfer of functional gains to activities of daily living (ADL).

To this aim, it is fundamental to detect the user's intention to guide the therapy appropriately. The electromyographic (EMG) signal can be exploited as a non-invasive source of information for detecting the intention to move before its actual execution, either by using multi-muscle EMG features or by taking advantage of the muscle synergies theory. In this context, HYBR-ID aims at developing a multi-modal robotic platform to support three-dimensional (3D) UL reaching, targeted to enhance functional arm recovery and ADL gains in stroke survivors with some residual motor capabilities.

The HYBR-ID (upper limb reHabilitation sYstem integrating exoskeleton assistance and functional electrical stimulation driven By a user tailoRed synergy-based Intention Detection framework) platform will consist of a cooperative control module, which combines the user's residual volitional contribution, FES-induced assistance and robot assistance, and an intention detection module, able to identify the user's intention to move towards a set of specific targets, both integrated with a previously developed compliant-controlled arm exoskeleton. The cooperative control module will include:

1. a feedforward anti-gravity robot assistance to compensate for device and arm weight;
2. a multi-muscle synergy-based FES controller to guide the subject's hand towards the desired reaching target;
3. a feedback corrective robot assistance to facilitate target reaching.

Both FES and robot assistance will be implemented according to an assist-as-needed paradigm to maximize user’s involvement. The intention detection module will consist of a machine-learning-based classifier fed by EMG-based and synergy-based features, which are extracted within an initial observation window to detect both the intention to move and the desired reaching target. The output of the intention detection module will serve as a trigger for the cooperative control module, and it will be optimized in terms of time needed for calibration and training, target classification speed, and robustness to undesired events in EMG recordings (e.g. contact loss).

HYBR-ID encompasses different technological advances, as it will represent the first attempt to integrate voluntary contribution, FES action and robot assistance to perform 3D UL reaching tasks, directly guided by the user. From a clinical perspective, we envision that training with HYBR-ID may enhance arm functional recovery in stroke survivors, improving the capability to transfer functional gains to ADL.