• this repo implements the mono version of OrcVIO, and is developed for analyzing its covariance.
• the original msckf is in the msckf branch

• platform ubuntu 18.04
• using conda conda env create --file environment.yml
• install pip packages
  • pip install sophuspy

• setup python path

export PYTHONPATH="${PYTHONPATH}:/home/erl/orcvio/orcvio-covariance-python"

• perform tests

python ./tests/test_msckf.py
python ./tests/test_quaternions.py
python ./tests/test_triangulation.py
python ./tests/test_twopoint_ransac.py

• run demo on euroc MH02 easy

python ./examples/run_on_euroc.py --euroc_folder /mnt/disk2/euroc/MH_02_easy/mav0 --use_viewer --start_timestamp 1403636896901666560

• result

Green is groundtruth trajectory, red is the estimated trajectory from OrcVIO-Mono

• covariance analysis

From figure below we can see MSCKF is making overconfident estimations whereas OrcVIO does not, due to the closed-form covariance propagation. The results are similar to similar to Fig 6, 8 in Camera-IMU-based localization: Observability analysis and consistency improvement

• https://github.com/Edwinem/msckf_tutorial
• https://github.com/uoip/stereo_msckf