Accurate tool targeting under both autonomous and hands-on control was achieved.

We considered a robot-assisted neuroendoscopy, and we developed a handling interface for linking a clinically-used endoscope to a lightweight robot (tool holder) with 7 DoFs. Such a robot holds potential for soft interaction with the surgeon, yet its intrinsic compliance must be suitably tamed not to lose tool targeting accuracy. Starting from practical specifications by neurosurgeons, we designed, fabricated and preliminarily assessed a compact and ergonomic handling interface. Such an interface permitted to easily insert/retract the tool (the measured force was 2 N), and to accurately hit a predefined target (the mean targeting error was below 0.5 mm, within the accuracy level of the optical tracker used for tool localization and pose). The feedback by neurosurgeons was very positive, thus encouraging further developments.