Odometry Validation - AutoModelCar/AutoModelCarWiki GitHub Wiki

Odometry Validation Package

This package helps to validate the odometry of the car using external camera.

Initial Setup.

• Mount camera on ceiling facing downwards.
• Attach red square on the car in center of rear axle such that it is visible all the time from ceiling camera.
• Measure lenth of camera coverage and calculate the pixels per meter. It is currently 279
• Currently car is placed on width axis of camera facing the (0,0) pixel. Please adjust code in lines (97-100) of odom_camera.py if the initial orientation of car is different.

Run

Run code: Use usb_cam package to obtain images from the ceiling camers.

Note: Do not run any other camera(realsense) while using for odometry validation, you may have images from different sources to process

To record odometry values calculated from the ceiling camera:

rosrun odom_validation odom_camera.py

To record odometry from car:

rosrun odom_validation odom_car.py

The above two programs will save coordinates to odom_camera.txt and odom_car.txt files. Use python script 'graph_plotter.py' in src folder to plot them in 3D. Note: Update text files names and path in the graph_plotter.py script. Reference text files are provided in the the src folder.

   

Trouble shooting:

Adjust the focus and other parameters of the camera for proper image

v4l2-ctl -d 1 --set-ctrl focus_auto=0 v4l2-ctl -d 1 --set-ctrl focus_absolute=0 v4l2-ctl -d 1 --set-ctrl contrast=0 v4l2-ctl -d 1 --set-ctrl brightness=0

If the timestamps in odometry of camera and car does not match, synchronize both the clocks.