Contemporary MRI, fluoroscopy and CTfluoroscopy
allow real-time visualization, which
enables real-time tracking of surgical instruments.
Three-dimensional ultrasound-guided interventional
systems [13,15] also include a real-time tracker in the
field of interest. In these systems, one can track
the end-effector of a surgical robot and manipulate
the device under visual servo control. It has been
known in general robotics that the operational
space formulation [16] and partitioned control [17]
can be used to alter the behavior of the system so
that it appears, kinematically and dynamically, to be
an RCM device. Unfortunately, existing kinematic
and dynamic models need to be precise, so the
joints must be fully encoded and calibrated. Extensive
research has also been devoted to visual servo
control [18], but work applied to uncalibrated and/
or unencoded robots has focused on estimating the
robot’s Jacobian rather than generating a virtual
Remote Center of Motion (Virtual RCM). Artificial
intelligence-based algorithms for robot motion have
also been investigated but not yet applied to the
needle-placement task. These algorithms have been
used in the control of uncalibrated mobile robots to
explore unknown environments and navigate familiar
environments [19]. Related research has also
examined the effect of uncertainty in robot sensors
and/or the environment [20] in generating a
collision-free map of the space.
Our present contribution combines an uncalibrated
needle-placement robot from three linear,
two rotational, and one linear insertion stages and
an AI-based motion algorithm to create a Virtual
RCM robot that requires neither encoded joints
nor complete knowledge of the robot kinematics.
Unlike classic RCM robots, the Virtual RCM
method does not require (1) the existence of a physically
fixed fulcrum point, (2) a priori knowledge of
the kinematic chain, or (3) encoding of the joints.
This relaxes many requirements previously imposed
on RCM needle-placement robots. For example,
the axes of the prismatic stages need not be orthogonal;
the axes of rotation stages need not intersect;
and kinematically unknown passive linkages are
permitted anywhere within the chain. This allows
robots using the Virtual RCM algorithm to be
simple and inexpensive to construct, eliminates
laborious calibration, and permits testing of new
robots or parts of robots to proceed rapidly without
affecting the accuracy of image guidance.
Materials