Autonomous tumor palpation and resection path planning using tactile array sensor and deep reinforcement learning for surgical robot

Abstract

Surgical robots have been widely used in tumor resection, but they still have shortcomings such as the lack of tactile perception, which may lead to inaccurate intraoperative tumor identification and resection. A piezoresistive tactile array sensor is proposed in this paper, which features small size (10 mm X 10 mm) as well as the high-efficiency array detection mode. The sensing principle is simply applying a constant voltage to each tactile element and measuring the current to generate a tactile image. Its effectiveness and performance are verified by finite element simulations. In addition, a deep reinforcement learning-based autonomous detection algorithm is developed to further improve the detection efficiency and facilitate the planning of the resection path, which provides an effective guarantee for accurate tumor resection in future autonomous robotic surgeries.