For convenience, here are instructions to change the map for the Localization project:
It is easiest to take photos of the new map with a cell phone or other camera. Take them at a height of 25cm and use an image stitching software to generate the map. We recommend auto-stitch. Replace map.jpg with your new map and change the following four parameters in "student_run_localization.py", "student_localization_helper.py": MAP_PIXEL_WIDTH, MAP_PIXEL_HEIGHT, MAP_REAL_WIDTH, MAP_REAL_HEIGHT. You may need to resize the image to be smaller if it is too large.
Version 2.0 of the PiDrone software comes with 5 programs for vision outlined in the table below. The two main algorithms are Localization and SLAM (Simultaneous Localization and Mapping). They both use the drone’s camera to produce a pose estimate, but localization requires a map of the environment before hand (map.jpg) whereas SLAM builds a map of the environment as it runs. SLAM is much slower than localization. The table shows which of the algorithms are supported under the following conditions: onboard (running on the raspberry pi), offboard (running on a separate Linux computer), online (running in real time) and offline (collecting data first and then computing in retrospect).
| Online | Offline | |
|---|---|---|
| Onboard | Localization & SLAM | SLAM |
| Offboard | Localization & SLAM | Not Supported |
As you can see, nothing is supported onboard and offline because the offboard computer is assumed to be quick enough to run the programs in real time. Also, localization is not supported onboard and offline because localization cannot run in retrospect, only in real time. Finally, note that SLAM onboard and online is supported, but runs too slow to work accurately. Do not expect this program to produce accurate poses.
The following instructions describe how to use each of the five supported programs:
Onboard: run "vision_localization_onboard.py" on the pi. You must fly over the area captured in map.jpg. Press ‘r’ to toggle localization.
Offboard: run "vision_localization_offboard.py" on the pi and "offboard_localization.py" on the remote computer. You must fly over the area captured in map.jpg. Press ‘r’ to toggle localization.
Change the map: It is easiest to take photos of the new map with a cell phone or other camera. Take them at a height of 25cm and use an image stitching software to generate the map. We recommend auto-stitch. Replace map.jpg with your new map and change the following four parameters in "offboard_localization.py", "onboard_localization.py", and "localization_helper.py": MAP_PIXEL_WIDTH, MAP_PIXEL_HEIGHT, MAP_REAL_WIDTH, MAP_REAL_HEIGHT. You may need to resize the image to be smaller if it is too large.
Offboard: run "vision_localization_offboard.py" on the pi. On the offboard computer run "offboard_slam.py." Press ‘r’ to toggle SLAM.
Onboard: run "vision_localization_onboard.py --SLAM" on the pi. Press 'r' to toggle localization.
Onboard: run "vision_localization_onboard.py --SLAM --offline" on the pi. Press “m” to toggle mapping mode on, during which time you can fly to collect data for the map. Pressing “m” again will stop the mapping and begin running SLAM offline. It is highly recommended that you land the drone for this part. Once it tells you that it is finished making the map, press “r” to toggle localization over the map you have just made.
Happy Flying!