Abstract
Europe is experiencing a progressive demographic shift characterized by an aging population and a growing prevalence of metabolic disorders, including obesity, type 2 diabetes, and cardiovascular diseases. Clinical evidence shows that these conditions benefit from regular, supervised physical activity focused on balance, posture, and stretching. At the same time, the COVID-19 pandemic highlighted the need for more sustainable healthcare systems capable of reducing the workload of healthcare professionals while maintaining high-quality patient care. Within this context, the I-TROPHYTS project investigated the integration of the Internet of Things (IoT), Artificial Intelligence (AI), and robotics to design next-generation motor rehabilitation systems supporting personalized therapies, continuous patient monitoring, and semi-autonomous rehabilitation coaching. The envisioned I-TROPHYTS scenario consists of smart rehabilitation environments, such as homes, gyms, or hospital rooms, where patients perform physiotherapy routines while wearing lightweight IoT sensors measuring body movements and physiological parameters. The collected data are processed locally through edge-computing techniques and exploited by a humanoid robot that assists patients by demonstrating exercises, suggesting alternative routines, scheduling recovery periods, and notifying clinicians whenever anomalies are detected. Such an approach enables a single physiotherapist to supervise multiple patients simultaneously while maintaining continuous monitoring and personalized rehabilitation. The developed platform follows a three-layer architecture. The Sensing Layer focuses on wearable, low-cost, privacy-preserving IoT technologies based primarily on inertial measurement units (IMUs), avoiding camera-based solutions. Research activities included the design of self-configuring wireless mesh networks for reliable acquisition of wearable data, together with Machine Learning and Deep Learning techniques executed at the edge for movement classification, anomaly detection, and evaluation of exercise quality. The Knowledge Layer provides a semantic representation of rehabilitation sessions by integrating real-time sensor observations, patient information, prescribed therapies, and clinical history into a unified ontology extending DOLCE. This semantic layer enables interoperability among heterogeneous sensing technologies while supporting context-aware reasoning and personalized rehabilitation planning. The Robotic Layer introduces cognitive robotic capabilities through a humanoid NAO robot able to interact with patients and clinicians, receive information from wearable sensors, reason on the rehabilitation context, and autonomously adapt therapeutic suggestions according to both patient conditions and exercise execution. The contribution from life sciences was equally fundamental. Medical partners identified representative rehabilitation scenarios, defined reference motor routines for selected patient categories, established sensing procedures and sensor placement, and provided the clinical knowledge required to model rehabilitation protocols and reasoning mechanisms.
Results achieved
Overall, I-TROPHYTS demonstrated the feasibility of integrating IoT sensing, edge intelligence, semantic knowledge representation, and cognitive robotics into a unified ecosystem for personalized motor rehabilitation. Beyond the technological achievements, the project established a multidisciplinary collaboration between engineering and medical sciences, producing reusable datasets, software platforms, mobile and web applications, and experimental testbeds. The results were disseminated through numerous scientific publications, workshop organization, collaborations with healthcare institutions, and the exploration of future research opportunities within European and international funding initiatives. The project successfully achieved its objectives in three directions: 1) Objective O1 – IoT-enabled smart environment. Significant advances were achieved in the design and validation of wearable IoT infrastructures for monitoring physiological and motor conditions during physiotherapy sessions. The project developed reliable wireless sensing platforms capable of acquiring movement data through wearable IMU devices while preserving user privacy and reducing deployment costs. To improve communication reliability, digital-twin models of BLE Mesh networks were developed, enabling self-configuration and automatic optimization of network parameters according to application requirements. This represents one of the first investigations of digital twins for adaptive wireless body-area communication in rehabilitation environments. The project also designed edge-computing architectures integrating Machine Learning and Deep Learning algorithms capable of classifying rehabilitation exercises, assessing movement quality, and supporting real-time monitoring without relying on cloud infrastructures. These solutions demonstrate that intelligent processing can be executed directly near the sensing devices, reducing latency and improving privacy. 2) Objective O2 – Ontology-based adaptive robotic architecture. The project produced a comprehensive ontology for representing rehabilitation sessions, integrating patients, exercises, wearable sensors, physiological observations, motor performance, and quality indicators into a coherent semantic model. The ontology extends existing foundational ontologies while introducing concepts specifically tailored to motor rehabilitation. Building upon this semantic representation, a cognitive robotic architecture was developed to enable autonomous reasoning during rehabilitation sessions. The robotic system integrates heterogeneous sensor information, interprets patient status, and generates context-aware therapeutic suggestions aligned with prescribed rehabilitation plans. The architecture allows adaptive behaviour by combining semantic reasoning with real-time sensor observations, enabling personalized assistance throughout therapy. 3) Objective O3 – Integrated proof of concept. Medical and technological partners jointly defined representative datasets of rehabilitation exercises performed by target patient groups. These datasets supported the identification of optimal sensing configurations, movement quality indicators, and reference thresholds for exercise evaluation. The project implemented and experimentally validated an integrated rehabilitation platform combining wearable IoT sensing, edge intelligence, ontology-based reasoning, and humanoid robotics. The proof of concept demonstrated the capability of measuring joint movements, evaluating exercise quality, and dynamically adapting rehabilitation protocols according to both physiological parameters and real-time patient performance. The outcomes of the project were disseminated through multiple channels, including: the publication of 18 articles in PE-related editorial venues and of 17 articles in LS-related editorial venues, the submission of 6 additional contributions, the organization of a workshop (IEEE W3R-HEALTH) on topics closely related to the I-TROPHYTS project, the participation to public engagement events. Additionally, selected components of the I-TROPHYTS framework were validated through small-scale testbeds, and software platforms (a mobile app and a Web dashboard) were developed to support motor routine IoT data collection, visualization, and sharing with medical professionals (e.g., physiotherapists).Project details
Unibo Team Leader: Marco Di Felice
Unibo involved Department/s:
Dipartimento di Informatica - Scienza e Ingegneria
Coordinator:
ALMA MATER STUDIORUM - Università di Bologna(Italy)
Total Eu Contribution: Euro (EUR) 228.113,00
Total Unibo Contribution: Euro (EUR) 66.771,00
Project Duration in months: 29
Start Date:
28/09/2023
End Date:
28/02/2026
