Leader: Domenico Mundo (UNICAL); Other collaborator(s): UNIPD, UNIFI
A novel virtual sensing technology applied to assistance robotics will be investigated in this task to enable real-time analysis, monitoring and control of the tasks undertaken by a user. In robotic motion exercising, it will provide an estimate of the effort that a patient exerts on the device. A strong advantage of the proposed approach is the minimization of cost and complexity of the sensing instrumentation: by combining inexpensive standard sensors, such as accelerometers and strain gauges, and biomechanical models, indirect measurements of relevant physical quantities (e.g., forces, torques, motion states) will be enabled.
In the reference period, a multibody framework for biomechanical simulation of human body motion has been defined and implemented as an in-house tool that can be exploited for the subsequent activities of Virtual Sensing during robot-assisted exercising. In parallel, the integration of a basic upper-limb biomechanical model with a stochastic estimator (based on an extended Kalman Filter) is ongoing for an initial investigation of the potential for the proposed methodology to enable indirect measurements of human muscular effort. Finally, a first case study of multi-physical human-robot ecosystem modelling in the developed platform was analysed and results will be shown at the Age-It General Meeting (Venice, 20-22 May 2024).
Main policy, industrial and scientific implications:
Once completed, the ongoing work will have three main implications for the practise of robotic rehabilitation and exercise: 1. The availability of the multibody model of the robot-patient ecosystem will enable a customized design of the exercise based on the patient's conditions; 2. The indirect measures of human effort enabled by the Virtual Sensing technology will augment the set of clinical information made available to the operator by biosignal acquisition systems; 3. In-home applications, expensive and difficult-to-use biosignal acquisition systems (ECC, EMG) will be replaced by the developed virtual sensors
Please see the next reporting period.
In the reference period, by exploiting the multi-body simulation platform developed to enable biomechanical simulations of human body motion, a muscular-skeleton model of a human upper limb, with simplified force-elements representing motor muscles, has been integrated an Extended Kalman Filter to enable indirect measurement of human muscular effort. A case study has been analyzed in which muscular efforts have been estimated for a virtual exercise and the achieved results, even if preliminary, are extremely encouraging. Ongoing activities aim at including in the digital twin of the example limb a more accurate biomechanical model of the muscles (e.g., the Thelen of the Millard model), along with the implementation, in the estimation platform, of biomechanical constraints and energy functionals to solve the inverse dynamic problem robustly.
In the reference period, the digital twin of the example limb was made more accurate by including more realistic biomechanical models of the muscles, i.e. the Thelen and the Millard models. A framework to solve the inverse problem of muscular force estimation was implemented and tested by using the minimum set of generalized forces, which makes the problem’s solution unique even if the biomechanical system is over-actuated. Ongoing activities aim at including, in the estimation platform, proper optimization functions (e.g., the sum of the squared activation functions) that allow solving the inverse problem at the level of individual muscle recruitment. In parallel, the investigation of an alternative ML-based approach to muscular force estimation has started, intending to improve the computation efficiency of the platform under development.
Scientific publications
Dissemination Events