Visual Servoing for Beating Heart Surgery - Visual Servoing via Advanced Numerical Methods - page 101

is very large then the original ESM algorithm does not work. Nevertheless, results
on the tracking of a deformable object [29] could then be integrated.
The final results we obtained using a heart beating video are shown in Fig-
ure 6.10. The video is taken at 500 fps. The ESM algorithm runs currently at 100
fps, which is close to the estimated needed framerate for beating heart surgery.

6.5 Control Strategies

In order to achieve high accuracy tasks, robots have to be controlled using exte-
roceptive measurements. In robotized cardiac surgery, visual sensors are generally
used because they do not introduce any additional cumber to the operating field:
surgeons already use cameras to have a visual feedback, especially in totally endo-
scopic surgery. In order to compensate for the heart motion, high speed vision has
to be used [31]. This allows to avoid aliasing when acquiring the fast transients of
the heart motion and also to model correctly the robot dynamics.
For safety reasons, medical robots are lightweight and composed of thin arms.
These mechanical characteristics induce a low bandwidth of the system which is in-
compatible with the dynamics required for the heart motion tracking. In order to cope
with these limitations predictive control was introduced for heart motion compensa-
tion to enlarge the tracking bandwidth [19, 6, 44]. Predictive control laws include
prediction algorithms that anticipate future heart motion as described in Section 6.3.
Predictive control is more suitable when it is possible to directly measure the current
heart motion, i.e. , in an active compensation approach. In an active stabilization con-
text, the end-effector of the robotic system is in contact with the heart surface making
the direct measurement of the heart motion impossible. In this case, H ∞ control was

suggested [3] in order to tune the robustness of the controlled system with respect to
the model uncertainties induced by the contact with the heart muscle.
In this section, the control strategies of two heart motion compensation schemes
are described. First, a generalized predictive controller (GPC) implementation for
the active compensation approach is described. Then, an H ∞ control law used for

the active stabilization method is derived.

6.5.1 Heart-tool Synchronization

In this approach, a robotic arm is controlled to simultaneously track the motion of
the area of interest and move the tool according to the surgeon gesture thanks to a
master device.

6.5.1.1 Robotic System Description and Modeling

The testbed (Figure 6.11) consists of a SCARA like robotic arm equipped with a
wrist. This robot, holding an instrument and simulating a surgical robot, is actuated