MFfCUAVSLAM
Map Fusion for Collaborative UAV SLAM
This project was a semester project I did at the Vision for Robotics Lab (www.v4rl.ethz.ch) at the Swiss Federal Institute of Technology (ETH) durgin my master studies in Information Technology and Electrical Engineering.
Report
The report that I had to write for this semester project is written in LaTeX and located in the folder report.
Presentation
The presentation I gave is located in the folder presentation. There is also an extended version, which does describe certain things in more detail, e.g. the different optimization algorithms.
Contained scripts/ROS nodes
Scripts
evaluation/scripts/evaluation.py
Python script to calculate the root mean squared error (RMSE) from a recorded experiment.
evaluation/scripts/evaluation_with_offset_estimate.py
Python script to calculate the root mean squared error (RMSE) from a recorded experiment with an estimated time offset.
ROS nodes
record_vicon
Receives ground truth positions and writes it into a text file.
retime_messages
Republishes topics with retimed timestamps.
v4rl_mcpslam-mapfusion
Modifid multi client SLAM system, in which the proposed approaches of this semester project are implemented.
Evaluation
The text files with the recorded timestamps and coordinates of each experiment/dataset are in the folder evaluation.
The data sets are "close", "far", "frontal", and "uav". "vicon" represents the ground truth. mX_skY describes how many KeyFrameMatches were required to fuse the maps (X) and how many KeyFrames were skipped (Y) after a KeyFrameMatch was detected.
Papers
The relevant research papers and some slides are in the folder papers.