Robot-assisted rehabilitation therapy has been proven
to effectively improve the patients motor function and it is demand is
increasing annually. However, one of its major limitations are their complexity
in operation, robustness, and difficulty of therapies personalization that
sometimes explain the concerns of clinicians to use this technology in their
daily practice. Automating the rehabilitation cycle, by introducing robotic
assistance will help to and homogenize protocols, performing automatic
recording of outcomes and rationale processes to increase sample size in
research studies. Moreover, a common problem remains if perform the therapies
in an open-loop manner without getting patients in the control loop and
considering them an homogenous entity. If there is no adjustment of the system
to each particular individuals , then patients are forced to adapt themselves
to the system capabilities. In contrast, in the same way that a skilled
physiotherapist does intuitively, a smart robotic assistance should modulate
their behaviour according to the user’s intention, action, state, as well as
emotions, providing feedback to perform a bidirectional adaptation. An
automatic administration of robotic therapies will
need this this smart skill to successfully deal with
complex situations such as relieve pain, reduce load on painful joints and
muscles, or detect patient demotivation.
The main objective of this project is the development
of smart robotic assistance systems for efficient assessment and personalized
rehabilitation using innovative control strategy with bidirectional
multisensorial feedback for both robot and patient. The system will use
collaborative robot IIWA with hybrid position/force control that will act as
passive element during the assessment phase and as active one during the
rehabilitation. At the same time, the patient will be equipped with the
multisensorial system of two types: a) embodied easy to wear sensors that
measure his/her medical parameters during rehabilitation such as IMU+EMG
signals, O2 saturation, hearth pulse, and arms/trunk poses, and b) external
cameras that analyse the patient’s face expression, compensatory movements and
postures, to infer during execution of exercises pain, placidity, fatigue, and
later the acceptance and adherence of the therapy.
The physical patient-robot interaction will be
dynamically measured and adjusted based on robot sensors feedback and
adjustment of predefined musculoskeletal model of patients upper limb features,
according the evolution of the therapy (velocity of the upper limb motion and
applying 3D forces). It mixes in bidirectional way information provided by the robot
sensors to adapt the patients exercise and, viceversa, use the patients data to
adjust the robot motion. It also mix objective (sensor measurements) and
subjective (deduced by observation) data during the therapy. This will be
implemented by AI techniques by learning from previous
measurements/observations and the development of a descriptors set for each pathology. This
paradigm is based on a multidisciplinary approach applied learning to robot
assisted therapy that exploits automatic assessment, machine learning and
gamification technologies capabilities to dynamically adapt the rehabilitation
execution and prescription during the therapy, helping therapists to be more
precise, efficient and cost effective at resource use and increase patient adherence
to the therapies.