Visual awareness has a significant impact on the effectiveness of control for operators in teleoperation tasks. The awareness can degrade when time delays are introduced into the system. In this MSc presentation, a new concept is presented that introduces a visual model-mediated approach that can be used in telemanipulation. A dense visual SLAM generated virtual model is created which is asynchronously updated with new information from a moveable RGB-D sensor. The operator is able to look around freely in the generated virtual reality environment.
The performance of the aforementioned concept was evaluated in this thesis. The performance requirements of the system were determined, and shortcomings were identified by using a root cause analysis. Functional tests were performed to improve upon these shortcomings where possible. It is shown that dynamic object tracking is still a severe problem when using a commonly available visual SLAM-based approach (RTAB-map).
Segmentation of moveable objects is required in future implementations to generate a virtual model that can cope with dynamic environments.
Meeting URL
https://bluejeans.com/168105200?src=join_info
Meeting ID
168 105 200
Want to dial in from a phone?
Dial one of the following numbers:
+31.20.808.2256 (Netherlands (Amsterdam))
(see all numbers - https://www.bluejeans.com/numbers)
Enter the meeting ID and passcode followed by #
