LapSurgie: Humanoid Laparoscopic Teleoperation

• LapSurgie is a humanoid-robot-based teleoperation framework that uses inverse-mapping kinematics to control passive, manual wristed laparoscopic instruments under fixed remote center-of-motion constraints.
• It enables deployment in standard human-oriented operating rooms by integrating a master console with stereo vision and calibrated ArUco markers without requiring specialized infrastructure.
• Preliminary user studies indicate improved surgical accuracy and reduced manual errors compared to traditional laparoscopy, despite current limitations in speed and haptic feedback.

LapSurgie is a humanoid-robot-based laparoscopic teleoperation framework in which a general-purpose humanoid robot performs laparoscopic tasks with unmodified, off-the-shelf, manual wristed laparoscopic instruments under remote center-of-motion constraints. Its stated objective is deployability in environments designed for humans without extensive infrastructure modifications, including operating rooms, while preserving precise hand-to-tool control through an inverse-mapping strategy and a stereo vision control console (Liang et al., 3 Oct 2025).

1. Clinical rationale and conceptual scope

LapSurgie is positioned against a specific deployment problem in robotic laparoscopy: adoption of surgical robotic platforms remains largely confined to high-resource medical centers, while practical deployment in underserved communities remains an unsolved challenge. The framework therefore treats the humanoid body not as a novelty in itself, but as a way to operate directly in standard human-oriented operating rooms, with standard laparoscopic ports, endoscopes, and instruments, rather than with a dedicated, highly specialized bedside robotic platform (Liang et al., 3 Oct 2025).

The system is teleoperated rather than autonomous. A common misconception is to read the humanoid embodiment as implying autonomous surgery; the framework instead uses a surgeon-operated master console based on da Vinci Research Kit master manipulators, and the central technical contribution is the kinematic inversion that maps desired tool-tip motion to the wrist pose required to drive a passive manual wristed laparoscopic instrument. This places LapSurgie within telesurgery and robot-assisted minimally invasive surgery rather than within autonomous surgical manipulation (Liang et al., 3 Oct 2025).

Another defining feature is instrument choice. Instead of specialized robot-native tools, LapSurgie uses manual wristed laparoscopic instruments, specifically ArtiSential bipolar fenestrated forceps, whose distal wrist is mechanically coupled to handle motion. This design decision is integral to the claim of no additional setup requirements for the surgical tool chain, but it also creates the framework’s main modeling challenge: the robot must infer and command the handle configuration that will realize the desired distal tip pose through a passive mechanism (Liang et al., 3 Oct 2025).

2. Physical architecture and console design

The hardware is divided into a master-side control console and a humanoid slave. The slave is a G1 humanoid robot whose two arms hold standard manual wristed laparoscopic instruments through custom coupling mounts. The master side uses two da Vinci Research Kit Master Tool Manipulators mounted on a small mobile cart, a dual-1920p stereo endoscopic camera, a GOOVIS G3 Max head-mounted stereo display, foot pedals, and a control workstation running ROS2. Console and humanoid communicate over a ROS2 network (Liang et al., 3 Oct 2025).

The instrument interface is external to the commercial tool. A custom coupling mount rigidly attaches the laparoscopic handle to the humanoid hand in a repeatable pose, and a small servo actuator drives fingers inserted into the instrument’s finger rings to actuate the jaws. The MTM gripper signal is mapped to a jaw opening range of $0^\circ$ to $60^\circ$, with $60^\circ$ treated as fully open. This preserves the native instrument while giving the humanoid explicit control of grasping (Liang et al., 3 Oct 2025).

The visual subsystem is deliberately compact. A stereo endoscope is rigidly mounted near the humanoid and pointed at the surgical workspace, and its feed is displayed immersively to the operator. For trocar geometry, the framework calibrates the remote center of motion using ArUco markers attached to the trainer board, allowing the 3D trocar point $x_{rcm}$ to be expressed in the same coordinate system as the tool model. The paper presents this arrangement as a portable cart-based teleoperation unit that can be brought into an ordinary room rather than a dedicated robotic suite (Liang et al., 3 Oct 2025).

3. Instrument model and inverse-mapping kinematics

The central mathematical object in LapSurgie is the inverse mapping from a desired tool-tip pose to the handle pose required by a passive wristed instrument. Let the humanoid end-effector pose be

$$P_{ee} = f_{FK}(\boldsymbol{q}) \in \mathbb{R}^{4 \times 4},$$

where $\boldsymbol{q}$ are the humanoid arm joints. A fixed transform ${}^{ee}T_{h1}$ defines the handle-1 frame relative to the wrist, so

$P_{h1} = P_{ee}\,{}^{ee}T_{h1}.$

The tool is modeled with handle 1, handle 2, and the distal shaft/wrist/tip. If $x_{h1}$, $x_{h2}$, and $60^\circ$0 denote the positions of handle 1, handle 2, and tool tip, then the geometry must satisfy the remote center-of-motion constraints

$60^\circ$1

together with the perpendicularity condition

$60^\circ$2

Defining

$60^\circ$3

the paper gives

$60^\circ$4

and then places the distal tip on the shaft line through the trocar as

$60^\circ$5

This construction enforces the trocar fulcrum geometrically rather than through a dedicated mechanical RCM linkage (Liang et al., 3 Oct 2025).

The manual wristed tool has internal passive angles $60^\circ$6 on the handle side and $60^\circ$7 at the distal wrist. The framework models the coupling with a measured gear ratio $60^\circ$8:

$60^\circ$9

The handle-side angles are computed from geometry as

$60^\circ$0

and, with

$60^\circ$1

$60^\circ$2

Using signed variants $60^\circ$3 and $60^\circ$4, the distal orientation is assembled as

$60^\circ$5

which yields the complete tip pose $60^\circ$6 (Liang et al., 3 Oct 2025).

Teleoperation requires the inverse problem. Given a desired tool-tip pose $60^\circ$7, LapSurgie solves for $60^\circ$8 with a trust-region reflective nonlinear least-squares solver. The residual is

$60^\circ$9

with $x_{rcm}$0. The optimized handle pose $x_{rcm}$1 is then converted back to a wrist command by a fixed offset. The result is a teleoperation chain in which the surgeon specifies a desired distal pose, while the controller computes the proximal pose necessary to realize it through a passive instrument at a trocar-constrained fulcrum (Liang et al., 3 Oct 2025).

4. Teleoperation workflow and operator interaction

From the operator’s perspective, LapSurgie behaves like a stereo telesurgical system. MTM motions are mapped to desired tool-tip motions in the laparoscopic task frame, and the inverse-mapping block converts those distal commands into a target wrist pose for the humanoid arm. Jaw motion is controlled independently by mapping the MTM gripper to the servo-driven finger mechanism on the laparoscopic handle. The control is unilateral and visual; the paper does not report force or haptic feedback (Liang et al., 3 Oct 2025).

The workflow depends on explicit RCM calibration. ArUco markers on the training board provide the trocar location, and the inverse mapping uses that calibrated point throughout teleoperation. This choice is consequential: it allows the framework to use a humanoid arm with standard industrial-style kinematics while still respecting the trocar fulcrum, but it also means that calibration accuracy becomes part of the surgical control chain (Liang et al., 3 Oct 2025).

The paper emphasizes that the system’s logic is centered on tool-tip intent rather than direct imitation of hand pose. This is especially important for manual wristed instruments because their distal wrist is a mechanically coupled function of handle motion. A direct master-to-handle mapping would not preserve intuitive tip behavior, whereas the inverse-mapping formulation explicitly solves for the proximal configuration that best realizes the surgeon’s intended distal pose under RCM and angle-limit constraints (Liang et al., 3 Oct 2025).

5. Comparative evaluation and user study

Evaluation was performed as a comparative user study with 14 participants: 2 professional surgeons and 12 novices. Each participant performed a bi-manual peg-transfer task on three platforms: manual laparoscopy with ArtiSential instruments, a dVRK setup, and the LapSurgie humanoid system. Platform order was randomized. Each participant had approximately 5 minutes of training per platform plus two practice trials, followed by 8 measured trials. The task used a board with four pegs arranged in a square 40 mm apart, two rubber O-rings, and drinking straws placed below the rings for grasp clearance (Liang et al., 3 Oct 2025).

The study used two principal metrics. The first was completion time. The second was a weighted error score composed of failed pick-up (2), stretch ring on pegs (2), stretch ring during hand-off (4), drop ring outside of pegs (5), collision with pegs, ground, or tools (3), and straw displacement (3). This weighting scheme differentiates between benign imprecision and more consequential failures such as dropping the object during transfer (Liang et al., 3 Oct 2025).

Platform	Novices	Surgeons
Manual laparoscopy	Error $x_{rcm}$2; Time $x_{rcm}$3 s	Error $x_{rcm}$4; Time $x_{rcm}$5 s
Humanoid LapSurgie	Error $x_{rcm}$6; Time $x_{rcm}$7 s	Error $x_{rcm}$8; Time $x_{rcm}$9 s
dVRK	Error $$P_{ee} = f_{FK}(\boldsymbol{q}) \in \mathbb{R}^{4 \times 4},$$0; Time $$P_{ee} = f_{FK}(\boldsymbol{q}) \in \mathbb{R}^{4 \times 4},$$1 s	Error $$P_{ee} = f_{FK}(\boldsymbol{q}) \in \mathbb{R}^{4 \times 4},$$2; Time $$P_{ee} = f_{FK}(\boldsymbol{q}) \in \mathbb{R}^{4 \times 4},$$3 s

For novices, LapSurgie’s error score was on par with dVRK and lower than manual laparoscopy, although the differences versus manual were not statistically significant at the reported sample size; LapSurgie was substantially slower than dVRK, with $$P_{ee} = f_{FK}(\boldsymbol{q}) \in \mathbb{R}^{4 \times 4},$$4 for time. For surgeons, the lowest mean error was observed on LapSurgie, while the shortest completion times remained with dVRK. The paper therefore characterizes LapSurgie as comparable to dVRK in accuracy and significantly better than manual operation in accuracy, but slower in execution, primarily because of hardware and control limitations rather than because of the teleoperation concept itself (Liang et al., 3 Oct 2025).

The error breakdown is also informative. Manual operation produced more ring stretching and hand-off errors than either teleoperated platform. LapSurgie showed fewer collisions than both dVRK and manual operation, while dVRK showed fewer failed pick-ups. Post-study questionnaires further reported significantly lower mental and physical demand for LapSurgie than for manual laparoscopy, and similar motion accuracy, feedback quality, and overall performance to dVRK, albeit with slightly higher mental and physical demand than dVRK (Liang et al., 3 Oct 2025).

6. Relation to laparoscopic robotics, limitations, and likely directions

LapSurgie occupies a distinct place within laparoscopic robotics. Earlier compact robotic assistants such as the patient-mounted Light Endoscope Robot and the ViKY robotic endoscope holder focused on camera holding, voice control, and reduction of assistant dependence rather than on teleoperated manipulation with wristed instruments (0711.4944, Long et al., 2012). More recent robot-agnostic laparoscopic research platforms have emphasized deterministic remote center-of-motion control on industrial manipulators such as UR5e and Franka, with teleoperation, demonstration recording, and autonomous policy deployment (Rodriguez et al., 9 Mar 2026). LapSurgie differs from both lines by making the humanoid body itself the deployable bedside manipulator while preserving off-the-shelf handheld laparoscopic instruments (Liang et al., 3 Oct 2025).

The framework also sits alongside parallel work on laparoscopic perception and visualization. Multi-view 3D visualization systems such as EasyVis2 estimate tool pose from a trocar-mounted camera array and YOLOv8-Pose (Sun et al., 2024), while laparoscopic scene-analysis pipelines have been developed for intraoperative visualization of gamma-probe sensing areas through tool tracking, depth estimation, segmentation, and 3D reconstruction (Huang, 3 Jan 2025). This suggests that LapSurgie’s teleoperation stack could, in principle, be paired with richer scene-understanding modules, although that integration is not part of the reported system.

The current prototype has explicit limitations. It provides no force or haptic feedback; participants and surgeons perceived higher latency and slightly lower precision than dVRK; the instrument model uses a simplified geometric coupling with fixed $$P_{ee} = f_{FK}(\boldsymbol{q}) \in \mathbb{R}^{4 \times 4},$$5; RCM calibration currently depends on ArUco markers and trainer-board setup; and evaluation is limited to a peg-transfer task rather than suturing, knot tying, dissection, or tissue retraction. The paper also states that it is a feasibility study rather than a clinically approved platform, with no animal or human in-vivo trials and without a detailed treatment of operating-room safety, collision management, or regulatory pathway (Liang et al., 3 Oct 2025).

The proposed future direction is correspondingly broad: lower-latency electronics and communications, better humanoid joint controllers, more accurate geometric models, haptic or shared-control extensions, expansion to suturing and tissue tasks, integration of autonomy modules, and progression to cadaver or animal tissue. A plausible implication is that LapSurgie is best understood not as a replacement for dedicated surgical robots in its present form, but as an argument that humanoid teleoperation can be made technically compatible with trocar-constrained laparoscopy using commercially available manual wristed tools.