Adnan Munawar, an assistant research scientist at the Laboratory for Computational Sensing and Robotics (LCSR), shared his progress on the open-source Asynchronous Multi-Body Framework (AMBF) simulator used for several applications in surgical robotics on Sept. 11. The talk shed light on the use of reactive digital twins for surgical environments. His paper on AMBF was published in the Intelligent Robots and Systems (IROS) program.
While traditional surgical procedures provided open access to the body’s anatomy, they often resulted in large wounds, significant bleeding, and increased risk of damage to non-target structures. Over time, advancements in surgical techniques led to minimally invasive surgery (MIS) which uses smaller openings and offers better outcomes for patients, including reduced recovery times and less pain.
Still, the physical and ergonomic strain on surgeons persisted, setting the stage for the next evolution — robot-assisted minimally invasive surgery (RAMIS). Robotic systems like da Vinci, Hugo, and Asensus integrated the benefits of MIS with the added comfort and precision robots provide.
As surgical robotics progressed, so did the need for advanced simulation platforms, which allowed rapid prototyping, kinematic and dynamic evaluations, and skill training.
“Simulation platforms assist research by allowing rapid prototyping of complex robotic actions in a safe, virtual environment, facilitating the training of surgeons and the testing of new robotic configurations without the risks associated with real-world procedures,” Munawar said. “Due to the pandemic, there is an ever-increasing need for robust simulators for surgical robotics.”
Before AMBF, the most popular simulators were GAZEBO and v-rep (now Coppelia Robotics). However, the complexity of setting up the environment makes them unsuitable for medical purposes due to the need to switch between operating systems, software, and description files.
Specifically, eight steps were needed to load a computer-aided design medical robot into these simulators, including file type conventions, tweaking dynamics, and kinematics, ensuring smooth structures. and failing any of the intermediate steps required re-starting the process from the beginning.
“We have to do things in a very weird fashion to make it work in the simulator,” Munwar explained.
Another motivation for developing AMBF was the limitations of existing simulation platforms, which were not designed for real-time or high-speed control interfaces. Specifically, traditional simulators faced challenges in handling physics loops which, when combined with collision detection, became computationally expensive and slowed down the simulation process. This hindered their ability to incorporate haptic devices which require fast, real-time updates for precise feedback. Additionally, integrating multiple input devices with different update rates was impractical with sequential simulations in the old platform.
AMBF addressed these limitations by offering more flexible, high-speed simulations capable of interfacing with diverse haptic devices while enabling real-time feedback loops.
“This asynchronous architecture is enabled by ADF, or the AMBF description format, which allows a steady 1000 HZ latency rate, in comparison to 200 HZ offered by previous simulators,” he said.
Munawar highlighted one key research application of the AMBF in his presentation, which involved a collaborative project with associate professor in the Department of Computer Science Mathias Unberath on simulating robotic suturing and skull surgeries.
The project primarily focused on skull surface surgery for conditions like craniosynostosis — a birth defect that affects skull development. In this simulation, control of robotic tools was integrated with real-time updates, allowing for precise robot control within the simulation loop. This proved particularly useful for simulating brain surgeries, where cadaver practices and systems like Phacon were used in conjunction with AMBF to develop advanced surgical simulations. Additionally, these simulations laid the groundwork for a new digital twin paradigm called TWIN-S, in which a virtual replica of the surgical environment was used to optimize real-world outcomes.
ABMF’s third version is currently in development, which would provide Robot Operating System 2 support and integrate with Blender’s rendering engines. The AMBF’s versatility also extends beyond medical applications such as space robotics, further showcasing its broad applicability in different robotic control scenarios.
