International Challenge on Surgical Autonomy in Real-world Settings
International Challenge on Surgical Autonomy in Real-world Settings
23 June, London
Vision
As part of the MRC-Hamlyn Workshop 2026, we invite global innovators to push the boundaries of surgical autonomy. While robotic-assisted surgery has transformed minimally invasive procedures, the next frontier lies in task automation, empowering robots to perform repetitive or delicate sub-tasks with superhuman consistency and safety.
This challenge seeks to benchmark the latest advancements in data-driven control, computer vision, and soft-tissue manipulation, fostering collaboration between engineers and clinicians to accelerate the translation of autonomous capabilities into clinical reality.
Challenge Tasks:
Goal: grasp the needle and move it to a target position while avoiding collision with obstacles.
Variation Parameters:
Needle Size: 2 different sizes (diameter > 1.0 mm and diameter < 0.5 mm)
Needle Pose: 2 different poses (upright on a flat surface and recessed, partially embedded up to 2 mm in tissue)
Obstacle Configuration: 2 different configurations (simple: 2 obstacles with >10 mm clearance, and complex: 6 obstacles with <5 mm clearance)
Light Intensity: 2 different levels (bright: normal endoscopic illumination, and low: endoscopic light source turned off)
Success Criteria:
An attempt is successful if:
The needle is grasped and transported to the target zone without dropping;
The needle centre is within 5 mm of the target zone centre;
The needle remains stable in the target zone for at least 2 continuous seconds;
Zero contact with any obstacle occurs throughout the attempt.
Any obstacle contact or human intervention constitutes a failed attempt.
Goal: control bi-manual arms to grasp suture and tie a knot to close a tissue wound .
Variation Parameters:
Tissue Geometry: 2 different geometries (flat: planar surface and curved: convex surface)
Suture Thickness: 2 different thicknesses (thick (e.g. 2-0 gauge ~0.3 mm) and fine (e.g. 4-0 gauge ~0.15 mm))
Wound Orientation: 2 different orientations (vertical: aligned with the camera's vertical axis and inclined: rotated at angle of 30 or 45 degrees)
Success Criteria:
An attempt is successful if:
Both arms grasp the suture ends;
A minimum of two throws are completed;
The tissue model is not torn, with wound edges approximated to within a 2 mm gap.
Tissue tearing, suture breakage due to excessive robot force, or human intervention constitutes a failed attempt.
Goal: drive the clip applier to approach the blood vessel and apply the clip on the vessel at millimeter level.
Variation Parameters:
Vessel Type: 2 different types (cystic duct and cystic artery, implicitly distinguishing vessel thickness)
Clipping Position: 3 different positions (first, second, and third clip on the same vessel)
Angular Orientation: 2 different angular orientations of the vessel
Light Intensity: 2 different levels (bright: normal endoscopic illumination and low: endoscopic light source turned off)
Success Criteria:
An attempt is successful if:
The clip is applied with the large hook side facing downward;
The structure is fully occluded with no visible lumen remaining;
The clip remains in place without dislodging when the structure is gently manipulated after application.
Incomplete occlusion, incorrect clip orientation, clip dislodgement, vessel perforation, or human intervention constitutes a failed attempt.
* The tasks shown in the demo videos are all performed through teleoperation rather than automation for reference
You are free to use any surgical robot platform.
We also provide the da Vinci Research Kit and Sentire® Surgical System (Cornerstone Robotics) for testing and challenge at MRC
Note: Participants may also use general-purpose dual-arm manipulators equipped with standard laparoscopic tools and external sensors are also allowed. But all of these factors will be taken into consideration of the experts' evaluation.
Registration 31 Mar. 2026
30 April 2026
Video submission deadline 10 May 2026
22 May 2026
Real-time Broadcast at MRC-Hamlyn Workshop 23 Jun. 2026
Award Announcement TBA 2026
Detailed Participation & Submission Process:
Phase 1: Registration (Deadline: March 31st April 30th)
Fill out the registration form, confirming your intent to participate and selecting your desired track(s).
Phase 2: Development & Submission (Deadline: May 10th 22 May 2026)
Participating teams will develop and test their algorithms on their local platforms and submit their final results online.
Video Demo: Submit one video per task to demonstrate the robot executing the task autonomously (MP4 preferred, maximum 2 minutes, multi-angle footage is recommended).
Technical Report: A brief document explaining the hardware platform used, the vision/control algorithm framework, system setup, and your approach to achieving autonomy.
For submissions, please email: autosurgchallenge2026@gmail.com
Phase 3: Preliminary Evaluation & Finalist Selection (Late May)
The organizing committee's expert panel will comprehensively evaluate all submitted videos and reports to select the top finalist teams who will advance to the live demonstration stage.
Phase 4: Global Real-time Broadcast (June 23rd)
The selected finalist teams will be invited to conduct a Real-time Broadcast of their algorithms running live to global peers at the MRC-Hamlyn Workshop in London, sharing their technological breakthroughs.
Phase 5: Final Evaluation & Champion Announcement
Based on the comprehensive performance during the live broadcast and the submitted materials. The expert panel will conduct the final evaluation and announce ONE CHAMPION for EACH TASK.
Expert panel will score submissions and live performances based on the rigorous demands of real-world surgical scenarios, focusing on the following dimensions:
Task Success Rate & Completeness: The ability to stably and fully achieve the task objectives (e.g., successful grasp, tight knot, accurate clipping position).
Operational Safety: Obstacle avoidance performance and whether any accidental collisions or damage occur to surrounding simulated tissues/environment (especially critical for Task 1).
Accuracy: Accuracy during action execution (e.g., the millimeter-level precision required in Task 3).
Efficiency: The time taken to autonomously complete the task while ensuring safety and accuracy.
System Robustness: The algorithm's adaptability to real-world environmental diversities and noise.
Clinical Applicability: The potential of the proposed method to be translated into real-world clinical settings, considering factors like hardware setup, workflow integration, and realistic assumptions.
Autonomy Level: Evaluated based on the degree of task automation achieved, specifically measuring how much (or how little) human intervention is required during the execution of the task.
Qi Dou
Associate Professor
Department of Computer Science and Engineering
The Chinese University of Hong Kong
Xiangyu Chu
Research Assistant Professor
Department of Mechanical and Automation Engineering
The Chinese University of Hong Kong
Jerry Wang
CTO, COO & Co-Founder
Cornerstone Robotics Limited
Hon Chi Yip
Assistant Professor
Department of Surgery
The Chinese University of Hong Kong
Yunxi Tang
PostDoctoral Fellow
Department of Mechanical and Automation Engineering
The Chinese University of Hong Kong
Kejian Shi
PhD Candidate
Department of Computer Science and Engineering
The Chinese University of Hong Kong
Samuel Au
Co-Director of Multi-Scale Medical Robotics Center
Professor at Department of Mechanical and Automation Engineering, The Chinese University of Hong Kong
Philip Chiu
Co-Director of Multi-Scale Medical Robotics Center
Professor at Department of Surgery, The Chinese University of Hong Kong