- Title: Robotic Sonographer: Learning Robotic Ultrasound With Sparse Expert’s Feedback

• Ref: https://ieeexplore-ieee-org-ssl.libproxy.snu.ac.kr/document/10684826
• Summary: Robotic ultrasound (RUS) offers a viable solution to address its limitations; nonetheless, achieving human-level proficiency remains challenging. Imitation learning or learning-from-demonstrations (LfD) methods have been explored in RUS, which learns a policy prior from offline demonstrations to encode the mental model of expert sonographers. This paper suggests a coaching framework (i.e., active feedback) for RUS to amplify its performance. This novel framework combines DRL (self-supervised practice employing an off-policy Soft Actor-Critic (SAC) network) with sparse expert feedback through coaching. The coaching by experts is modeled as a Partially Observable Markov Decision Process (POMDP).