Modified Dynamic Movement Primitive-Based Closed Ankle Reduction Technique Learning and Variable Impedance Control for a Redundant Parallel Bone-Setting Robot

Traditional fracture reduction relies heavily on the surgeon’s experience, which hinders the transmission of skills. This specialization bottleneck, coupled with the high demands on physical strength, significantly limits the efficiency of daily treatments in trauma orthopedics. Currently, most fracture surgery robots focus on open or minimally invasive reduction techniques, which inherently carry the risk of iatrogenic damage due to surgical incisions or bone pin insertions. However, research in closed reduction-oriented robotic systems is remarkably limited. Addressing this gap, our study introduces a novel bone-setting robot for the closed reduction of ankle fractures designed with a redundant parallel platform. The parallel robot’s design incorporates three sliding redundancy actuators that enhance its tilt flexibility while maintaining load performance. Moreover, a singularity-free redundant kinematic solver has been developed, optimizing the robot’s operational efficacy. Building upon the demonstrations from professional closed reduction techniques, we propose the use of a multivariate Student-t process as a multi-output regression model within dynamic movement primitive for accurately learning stable reduction maneuvers. Additionally, we develop an anthropomorphic variable impedance controller based on inverse dynamics. The simulation results demonstrate convincingly that the developed ankle bone-setting robot is proficient in effectively replicating and learning the nuanced closed reduction techniques.