Revision history [back]

I'm pretty sure there isn't a map ROI package available yet.

I have a half-finished node that converts the map to an image and publishes it using image_transport. This has the advantage that one can use the compressed (jpg) or theora compression options which save huge amounts of bandwidth. Of course these compression schemes are lossy, but for visualization they're ok. For additional bandwidth saving, the node publishes a full and/or a tile based image, the latter showing only the victinty of the robot pose. This way, the map around the robot can be visualized using only a few kb/s of bandwidth. Another node running on the GUI side could reassemble the image to maps, so they can be visualized in rviz as usual. Haven't done this yet, but should only take time :) Let me know if you want to take a look at it (it's not in our public repo for a reason though :) ).

Of course, the "right way" would probably be to use RLE encoding for the transmission of occupancy maps, as they (at least at this time) only have the three discrete states "free","occupied" and "unknown", which makes it possible to get huge bandwidth saving using the lossless RLE compression scheme. We already use this to share maps between our humanoid robots in real time (video, but due to time constraints no one ported that to the ROS ecosystem yet.

I'm pretty sure there isn't a map ROI package available yet.

I have a half-finished node that converts the map to an image and publishes it using image_transport. This has the advantage that one can use the compressed (jpg) or theora compression options which save huge amounts of bandwidth. Of course these compression schemes are lossy, but for visualization they're ok. For additional bandwidth saving, the node publishes a full and/or a tile based image, the latter showing only the victinty of the robot pose. This way, the map around the robot can be visualized using only a few kb/s of bandwidth. Another node running on the GUI side could reassemble the image to maps, so they can be visualized in rviz as usual. Haven't done this yet, but should only take time :) Let me know if you want to take a look at it (it's not in our public repo for a reason though :) ).

Of course, the "right way" would probably be to use RLE encoding for the transmission of occupancy maps, as they (at least at this time) only have the three discrete states "free","occupied" and "unknown", which makes it possible to get huge bandwidth saving using the lossless RLE compression scheme. We already use this to share maps between our humanoid robots in real time (video, ), but due to time constraints no one ported that to the ROS ecosystem yet.

I'm pretty sure there isn't a map ROI package available yet.

I have a half-finished node that converts the map to an image and publishes it using image_transport. This has the advantage that one can use the compressed (jpg) or theora compression options which save huge amounts of bandwidth. Of course these compression schemes are lossy, but for visualization they're ok. For additional bandwidth saving, the node publishes a full and/or a tile based image, the latter showing only the victinty of the robot pose. This way, the map around the robot can be visualized using only a few kb/s of bandwidth. Another node running on the GUI side could reassemble the image to maps, so they can be visualized in rviz as usual. Haven't done this yet, but should only take time :) Let me know if you want to take a look at it (it's not in our public repo for a reason though :) ).

Of course, the "right way" would probably be to use RLE encoding for the transmission of occupancy maps, as they (at least at this time) only have the three discrete states "free","occupied" and "unknown", which makes it possible to get huge bandwidth saving using the lossless RLE compression scheme. We already use this to share maps between our humanoid robots in real time (video), but due to time constraints no one ported that to the ROS ecosystem yet.

I'm pretty sure there isn't a map ROI package available yet. (See edit at end of this post)

I have a half-finished node that converts the map to an image and publishes it using image_transport. This has the advantage that one can use the compressed (jpg) or theora compression options which save huge amounts of bandwidth. Of course these compression schemes are lossy, but for visualization they're ok. For additional bandwidth saving, the node publishes a full and/or a tile based image, the latter showing only the victinty of the robot pose. This way, the map around the robot can be visualized using only a few kb/s of bandwidth. Another node running on the GUI side could reassemble the image to maps, so they can be visualized in rviz as usual. Haven't done this yet, but should only take time :) Let me know if you want to take a look at it (it's not in our public repo for a reason though :) ).

Of course, the "right way" would probably be to use RLE encoding for the transmission of occupancy maps, as they (at least at this time) only have the three discrete states "free","occupied" and "unknown", which makes it possible to get huge bandwidth saving using the lossless RLE compression scheme. We already use this to share maps between our humanoid robots in real time (video), but due to time constraints no one ported that to the ROS ecosystem yet.

/edit: @Lorenz mentioned in the comments below that a node for this task indeed already exists at http://rosjava.googlecode.com/ in the compressed_visualization_transport package.