Master Thesis on “Visual segmentation of elevator components with robotic platform”
Supervisor: Dr. Francesco Verdoja (francesco.verdoja@aalto.fi)
Advisors: Prof. Ville Kyrki (ville.kyrki@aalto.fi), Jari Karhu (jari.karhu@kone.com)
Keywords: object segmentation, deep learning, robotic vision, robot control, elevator maintenance
This thesis is part of an on-going collaboration with KONE Oyj
Project description
Elevator inspection currently relies on periodic on-site visits because installed sensors do not capture all maintenance needs. In this context, robotic inspection can offer significant advantages by enabling more frequent and accurate data collection, reducing the need for humans to operate inside elevator shafts. Visual segmentation of elevator components is a critical step in robotic inspection, as it allows the system to identify and analyze specific parts of the elevator for maintenance purposes.
This thesis focuses on developing and evaluating visual segmentation methods for elevator components using a mobile robotic platform. The student will work on designing and implementing algorithms that (i) allow the robot to enter the elevator shaft by walking on the roof of the elevator cabin, and (ii) segment and identify various elevator components from visual data collected by the robot. The project will involve data collection, algorithm development, and performance evaluation in real-world elevator environments.
KONE Oyj, partner in this thesis, will provide data and guidance about the components to be segmented. The thesis final evaluation is planned to be carried out on the KONE model elevator in Aalto.
Prerequisites: python/C++, ROS, machine learning, computer vision, control
Start: Available immediately
Deliverables
- Review of relevant state-of-the-art literature;
- Implementation of a navigation and control pipeline to allow the robot to enter an elevator shaft by walking of the roof of the elevator cabin;
- Implementation of an object segmentation method to identify components inside an elevator shaft;
- Evaluation of system performance in a real-world elevator.