Trajectory Accuracy and Error Evaluation

Now that you have all watched Stereo Visual Odometry and Visual SLAM in action, it is time to do what Kapernikov does best…
Manage Data

Task 1: Extend the C++ or Python ros node in trajectory_evaluation package that extracts trajectory and ground truth messages. You will be provided with rosbag and csv files and you can choose either to write a ROS subscriber or a simple csv parser. You can also work with a rosbag dataset or trajectory and ground truth csv files for convenience. Download them and put them in /path/to/tech-session-visual-odometry/datasets/gazebo/. You can play the rosbag dataset with:

$ rosbag play /path/to/rosbllag-dataset.bag # add -l for playing in a loop

Task 2: Create fancy plots of the trajectories
Task 3: Clean the data (align and interpolate in time)
Task 4: Calculate and plot translation (x,y,z) and rotation (wx,wy,wz,w) errors
Task 5: Create more fancy plots. You can get some ideas from: link (Optional) Task6: Create a new rosbag file including trajectory, ground truth AND the optimized map graph of VSLAM (topic: (/ackermann_vehicle)/rtabmap/mapGraph), align the three and apply tasks 2-5.

Further instructions:

1. Use gazebo_stereo_odometry_and_GT_last_messages.bag with two simple ros subscriber to receive the two messages
2. Or create a csv parser for the gazebo_trajectory_ground_truth.csv and gazebo_trajectory.csv