Hey,

I'd like to use your block_xmit implementation for a project of mine, but the example 'a simple test suite' seems little slow. I was able to speed it up a little by setting packet_count and max_queue of the tick method.
Increasing the bandwidth and chunk_size helped too, but could you provide an example optimized for performance?
If I set bandwidth to 5MB/s a 600MB file should be sent in about 2 Minutes, but it takes much longer.
With a bandwidth of 100MB/s it takes around 80 Seconds.

bandwidth = 100000000
num_blocks = int(total_size/65535)
chunk_size=65535
packet_count=1000
max_queue=50

Thanks!

The documented libjpeg-turbo8 dependency is not available on Raspbian (Raspberry Pi Debian). It looks like the package might be called libturbojpeg1-dev, on Raspbian, but not sure.

When I tried to follow the install instructions for Ubuntu (on WSL2) here I ran into the following errors

• sudo apt-get install python-pip libusb-1.0.0-dev libdc1394-22-dev
  • E: Unable to locate package libdc1394-22-dev
  • Maybe this package should be updated to "libdc1394-dev" according to some random googling
• sudo apt-get install libjpeg-turbo8-dev python-opencv python-wxgtk3.0
  • E: Unable to locate package python-opencv
  • E: Unable to locate package python-wxgtk3.0
  • E: Couldn't find any package by glob 'python-wxgtk3.0'
  • Maybe it is not "python-opencv" but rather "opencv-python" and maybe it should be installed with pip3? e.g. "pip3 install --upgrade opencv-python"
• pip install numpy future gooey
  • ERROR: Failed building wheel for wxpython
  • maybe this command needs to be added to the list, "sudo apt-get install libgtk-3-dev"

When attempting to run, "python geosearch.py " the application started but this error appeared in the console, "could not find module piexif" so maybe we also need, "pip3 install --upgrade piexif"

Install docs say:

On Raspian, libjpeg-turbo8-dev is instead named as libturbojpeg1-dev.

On my Raspian-lite, it's actually libturbojpeg0-dev. Suggest changing the doc to say libturbojpeg-dev and let apt-get select the appropriate integer (not sure how that works, but it seems to :)

Some things that need doing:

• fix RGB <-> BGR on map
• add lidar based height to mav_position.py
• only send ADSB_VEHICLE msgs if they are close enough to a vehicle
• send ADSB_VEHICLE msgs to both relay and retrieval vehicle
• delay takeoff if DNFZ blocking path
• can't measure distances in images in camera_air
• brightness control broken in mosaic
• fix rcout set_failsafe_pwm on ChibiOS
• place marker not working in camera_air
• image fetched marker not showing on map
• show image from map not working in camera_air
• disable SBUS failsafe handling parameter
• scale obstacle margin for obstacle speed and turn rate
• prevent pkts from 3G on porter from coming back down the rfd900x link
• send ADSB_VEHICLE pkts over all 3 links
• work up a list of OS configuration changes required on laptops which are to be "hands-off" (screensavers, automatic updates, crash-catching tools like whoopsie, ...)
• in search area WP_MAX_RADIUS should be zero. In main area should be 50
• add LZ exclusion areas for trees
• add method to reset battery state
• wp list gets stuck on companion
• look at bunny hop landing from 30th august
• allow ALT_OFFSET in terrain following mode
• cope with dual-aircraft NMEA
• adjust cuav_check parameter checking
• use USER WPs for AFS targets
• check GEOID for asterix data
• check GEOID for NMEA

From all-up test 2018-09-10:

• Kraken needs to get obstacle data. tridge's laptop will need to relay asterix data to Stephen's laptop as well as send it to vehicle; Stephen's needs to display but not process further
• fence auto-enable activated too late on Kraken; activated at WP1, should have been active earlier
• altitude is being read out too often (fix in mavinit scripts on both groundstations)
• Kraken struggled to reach altitude - review dataflash log
• motors were spinning when vehicle was still in NAV_WAIT_AIRSPACE
• Porter had no obstacles but did not pass through NAV_WAIT_AIRSPACE
• vehicles may be sending ADSB packets to ground - stop that!
• Optus Pi dead (doesn't ping)
• check for compass failure on Kraken
• need to fit df13 to USB externder to Porter (get from Ranger?)
• display collision data on ground while disarmed
• need another all-up test using quadstar and the Ranger
• need to make sure airspeed calibration is done!
• GREG: pull-ring on e-stop
• GREG: file down connector
• GREG: safety pin for GPS
• check scheduling on Porter
• log counts of objects currently being held for avoidance
• add into list of things to consider when plotting mission: choose first waypoint to avoid trees!
• add minimum flight time for Porter; full-fuel-loadout makes for hard second-takeoffs!
• DF-13 extender for Pixhawk in Porter (steal from Ranger?)
• add 5-second skies-are-clear-so-I'm-going message
• check data rates on GCS
• check why secondary links from Porter weren't helping links be non-claggy
• turn off IMU batch sample logging
• PETER: work out why mavlinklogdump won't write ISBD packets
• fix no-obstacles-on-Kraken issue
• add to checklist: turn camera on on Kraken
• ensure alt readout rates are correct for feet in mavinit.scr
• set numcells for Kraken in mavinit.scr
• consider motor stopping on descent at far site
• sudden drop after levelling off in takeoff
• ensure thermal throttling is off in ESCs.
• get to bottom of erratic avoidance
• look at Leonard's PR to aid in climb-out at remote site (prioritize climb vs yaw)
• consider getting HIL working
• improve filtering on GCS for obstacles
• use SITL to reproduce avoidance issues (erratic aircraft)
• AVDM doesn't do fence-avoidance logging

To recreate the problem do this:

• run cuav on a set of images which have pixel width and height larger than the default
• when a region of interest appears in the Mosaic window, click "Show Image"
• from the "View" window right-click and "Place Marker" somewhere within the blue box
• on the "Map" view the marker will appear quite far from where the region-of-interest appeared on the map