A Guided Sampling Enhanced Rapidly-Exploring Random Tree Path Planning Algorithm for Robot-Assisted Flexible Needle Insertion.

Percutaneous puncture techniques have been widely adopted across various domains of modern clinical interventional therapy due to their high diagnostic specificity, minimal invasiveness, and rapid postoperative recovery. However, the nonholonomic kinematics arising from the interaction between flexible needles and soft tissues, combined with the complexity of human anatomical structures, pose significant challenges to robot-assisted flexible needle insertion. To address these challenges, this paper proposes an improved rapidly-exploring random tree (RRT) path planning algorithm incorporating soft actor critic (SAC)-guided sampling. By integrating SAC-guided sampling strategies, the algorithm offers effective sampling guidance for path planning, significantly reducing the randomness of the search process, minimizing the generation of invalid nodes, accelerating convergence, and improving both path quality and planning efficiency. A hybrid sampling strategy is employed to balance global exploration and local exploitation capabilities, thereby enhancing adaptability and planning performance in complex anatomical environments. Furthermore, a navigation and positioning robot is integrated to autonomously guide the needle toward the target, thereby improving the autonomy of the insertion procedure. Target insertion experiments demonstrate an error of 0.97 ± 0.41 mm in synthetic biomimetic tissue, demonstrating strong potential for clinical translation.