Visual SLAM Approach with a Low-Power Remote Robot Agent

Abstract

This master’s thesis examines the viability of using a low-performance, low- power-consumption remote robot agent to collect sensor data for use in a vSLAM system. Selections for the components of the robot agent were made based on results from a small initial study, the robot was assembled, and software to run it was created. A remote server application was created to communicate with the robot agent and receive sensor data from it, subsequently processing and using the data with ORB-SLAM3.

The robot agent successfully collects and transmits sensor data to the server ap- plication, which successfully processes and uses the data in a vSLAM approach using ORB-SLAM3. Results indicate that the approach is viable, but due to hardware limitations overall performance is lower than intended. An analysis was performed and proposals for improvements to mitigate the limitations were suggested. Suggestions for further work was given.