Feiyang Hong, Hongjie Wang, Xia Yu, and Liyong Ma
Ultrasound robot, trajectory optimisation, gray wolf optimisation algorithm, Q-learning, visual servo
In the context of the convergence of medical imaging and robotics, traditional ultrasound diagnosis is characterised by unstable image quality and low efficiency due to the reliance on manual operation. Consequently, this study presents an autonomous cardiac ultrasound scanning system that integrates path planning, trajectory optimisation, and visual servo control. BiTRRT is used to generate obstacle-aware initial paths, which are smoothed using fifth-order rational B-splines. A multi-objective Q-learning grey wolf optimiser (MO-QLGWO) is developed to jointly minimise execution time and jerk, yielding a 10.4% reduction in scanning time and a 6.8% decrease in joint jerk. Cardiac section detection is performed with a YOLOv11 model achieving mAP50 = 0.99, and an IBVS controller using point-cloud normals enables rapid probe alignment, converging within 1.7 s. All evaluations are conducted in a ROS2– UR5 simulation environment. Results demonstrate that the proposed system achieves smooth motion, accurate section tracking, and real- time visual convergence, providing a viable foundation for future clinical deployment of robotic cardiac ultrasound.
Important Links:
Go Back
