Three topics are proposed below, each with a proposed special syllabus and one or more specialisation project building on the syllabus.
(1) Simulation of Biomechanical Systems.
Students taking this specialisation will be part of the biomechatronics research group, and relate to our vision of a Computer Haptic Assisted Orthopaedic Surgery (CHAOS) Simulator.
Our vision is to create a surgical simulator with high fidelity, to allow surgeons to train on specific procedures. Surgical simulators have proved effective in training basic skills, but existing simulators do not have sufficient fidelity to model advanced procedures in detail nor model the pecularities of individual patients. We expect to be able, in the future, to create digital twins of individual patients, so that experienced surgeons can practice and plan advanced procedures on very specific conditions. The most immediate goal is to simulate arthroscopy, but we are also studying the modelling and simulation methods necessary for more general types of surgery.
This specialisation module and projects will contribute to this vision in different ways.
Number of students 1-4
Specialisation Module
The students will read a syllabus covering biomechanical modelling, human anatomy, haptic modelling and control. The exact contents of the syllabus will developed in collaboration with the student, in order to support the specialisation project.
Assessment Format Viva Voce
Supervision Format Group supervision
Supervisors Hans Georg Schaathun (mathematical modelling, co-ordination), Bismi Rasheed (biomechanical modelling), Terje Vagstad (orthopedic surgeon), Øystein Bjelland (haptic control)
Specialisation Project
Assessment Format Written Report and Open Source Software
Supervision Format Preferably group supervision, but this is negotiable.
Number of students 1-3
Soft Tissue Simulation This project contributes to the vision by integrating simulation models for soft tissue deformation with realistic materials models, and make them run efficiently. The project should lead to an autonomous simulation model which can easily be integrated with different visualisation tools and other models in co-simulation systems. A particular challenge is smooth simulation in real time, but this may require continuation into an MSc project.
Soft Tissue Visualisation This project contributes to the vision by creating a visualisation system for orthopedic surgery. The software needs to be built in modular fasion, so that the visualisation client can connect to different simulation engines (such as the one from the preceeding project). The main challenge here is to create an autonomous client, to visualise the behaviour of simulations such as the one developed in the project below.
A particular challenge is smooth visualisation in real time, but this may require continuation into an MSc project.
Human joint animation This project works more closely with the Aalesund Biomechanics Lab (AaBL) which is a collaboration between NTNU (IIR) and Ålesund Hospital. The research at AaBL is interdisciplinary where robotics and cybernetics intersects with orthopaedics and biomechanics. This project is will develop an animation engine for 3D skeletal models, or simply visualization of joint motion trajectories. The models should ideally just contain skeleton without soft tissue structures (muscles, tendons or ligaments). However, the motion, which the student must describe kinematically, should be as close to the natural motion of the respective joint.
Biomechanical Modelling
Develop a dynamic (functional) model of the knee, or a part of the knee, to support simulation of movement, behaviour, and eventually surgery. This project will focus more on the modelling and less on the implementation of the model. Most probably, a biomechanical modelling tool such as Abacus will be used for the task.
Machine Vision is a critical element of robotics. Autonomous robots need to be able to see their environment. It is a broad field, which includes and applies elements of optics and geometry, 3D visualisation, image processing, and machine learning. The students may attend the Level 3 module on Machine Vision, but adapt and extend the syllabus in depth to be suitable as an MSc module. Exact details are decided in collaboration with the student.
Assessment Format Viva Voce
Supervision Format Class/Group
Supervisor Hans Georg Schaathun (module convener)
Specialisation Project
Assessment Format Written Report and Open Source Software
Supervision Format Individual
Project Intergalactic Machine Vision. One of the great challenges in physics is to map the Universe. This is difficult because a lot of the matter is so-called dark matter, which cannot be observed (or photographed) directly. We can observe it indirectly, because gravitational forces defract light and distorts the images of more distant galaxies. This is known as gravitational lenses, sharing many principles with ordinary lenses as studied in machine vision. Good lense models exist, and given distributions of light and dark matter, it is possible to simulate the distortion. However, recovering the dark matter distribution from observed images is hard. Some special cases have been resolved using machine learning, and this project will aim to extend these solutions to solve more general cases.
Supervisors Hans Georg Schaathun (machine learning/machine vision) and Ben David (physics/cosmology).
Bin picking (3D vision). Machine vision is critical to robotics. Robots need to see, analyse, and understand the world around them. An important industrial task is bin picking, where a robot recognises an object among many in a bin and figures the best way to grasp it. This is applied, for instance, in industrial robots doing assembly on a conveyor belt. A common solution is by means of machine learning, which requires large amounts of data with a known ground truth. This can be accomplished by using a physics simulator to generate synthetic data, like for instance generating a set of physics objects that drop into a bin and take a picture of it.
This project will focus on generating synthetic data and comparing computer vision methods of your choice against the data. The physics simulator isn’t specified, and could be Unity, Unreal Engine or Nvidia Omniverse. A base setup for “bin picking” will be provided in the simulator.
The physics simulator and data generation systems from the specialization project will be a good foundation into a master thesis, where you could either expand the system towards AI and reinforced learning, or towards a physical integration towards a robot using ROS2.
Supervisors Adam Leon Kleppe (robotics) and Hans Georg Schaathun (machine vision)