In robot-assisted orthopaedic surgery, registration is a key step, which defines the position of the patient in the robot frame so that the preoperative plan can be performed. Current registration methods have their limitations, such as the requirement of immobilisation of the limbs or the line of sight (LOS) issues. These issues cause inconvenience for the surgeons and interrupt the surgical workflow in the operating room.

Targetting these issues of current registration methods, we propose a camera-robot registration system for joint replacement. The bone geometry, which is measured directly by a depth camera, is aligned to a preoperatively obtained bone model to calculate the pose of the target. Simultaneously, in order to avoid registration failure caused by LOS interruptions, the depth camera tracks objects that may occlude the target bone, and a robot manipulator is used to move the camera away from the nearest obstacle. The optimal camera motion is calculated based on the position and velocity of the obstacle, which avoids the occlusion efficiently without changing the target position in the camera frame. Inverse kinematics of the robot is used to project the Cartesian velocity of the end-effector into the joint space, with kinematic singularities considered for stable robotic control. An admittance controller is designed as the human-robot interface so that the surgeon can directly set the robot configuration by hand according to the actual environment.

Simulations and experiments were conducted to test the performance. The results show that the proposed obstacle avoidance method can effectively increase the distance between the obstacle and the LOS, which lowers the risk of registration failure due to obstacle occlusion. This pilot study is promising in reducing distractions to the surgeon and can help achieve a fluent and surgeon-centred workflow.