Abstract
This work is concerned with robotic assisted beating heart surgery. First, a new image-based visual servoing scheme based on Nonlinear Model Predictive Control is proposed. With this controler, contraints of the system can naturally be taken into account by simultaneously handle the trajectory planification and the control. Experiments on a robotic platform with a classical camera validate the proposed approach. This scheme has also been adapted for echographic vision. It allows to control tool tip motion while ensuring intersection constraint with the echographic plane. Experimental validations highlight the efficientcy of the control scheme. This second part of the work takes place in a project of robotic assisted endovascular surgery, particularly for valve surgery. In this context, the real time localisation of the mitral valve is realized through a new method based on the tracking of the junction of the leaflet and the heart wall and a 2 dof model of the valve. The porposed algorithm has been valided with in vivo sequence. In the last part, in a first evaluation for heart motion compensation, we focus on 3D heart motion estimation using in vivo endoscopic images.