Design of Next Generation Ultrasonic-Based Healthcare Transmission Systems exploiting MAB

Future healthcare systems will use unconventional communication systems, like Ultrasounds (US), to communicate through human body tissues safely and efficiently. US have been extensively used for medical applications in the last decades, but is well known that performance can be affected by time-varying conditions in the body, making adaptability crucial for body area networks (BAN). Coordinating the behavior of US devices via their Digital Twins (DTs) can enhance adaptation and optimize transmission parameters. Next-generation networks will have DTs of devices cooperating to improve communication reliability. To cope with the intrinsic unreliability of transmissions inside or along the body, and to achieve the required adaptivity, we assume that a learning mechanism based on Multi-Armed Bandit (MAB) can be deployed to optimize data transmission based on the current channel quality. This implies that DTs will iteratively estimate conditions and adapt US transmissions using feedback to optimize metrics like error rate and throughput. Experimental tests on tissue-mimicking phantoms confirm the feasibility of using learning techniques for robust data transmission.