7 ms·

Perception is the hardest part [1] for autonomous aircrafts, and from my experience, perception research for aircrafts is still in its infancy, albeit a hot res

Perception is the hardest part [1] for autonomous aircrafts, and from my experience, perception research for aircrafts is still in its infancy, albeit a hot research topic. In some ways, it is easier than self driving cars due to the added DOFs, but in other ways can be quite complicated due to weather conditions. [1] http://www.aviationweek.com/Article.aspx?id=/article-xml/AW_08_05_2013_p53-601281.xml&p=2# http://www.aviationweek.com/Article.aspx?id=/article-xml/AW_...

krasin 13y ago
Part of that issue is the lack of high-performance and lightweight board for real-time machine vision (you won't have an NVIDIA GPU + x86 board as a part of your multicopter). All hopes that Parallela board (http://parallella.org http://parallella.org) will make it feasible.
- neurotech1 13y ago
  Depends on how small, how powerful, how reliable and costs. Its amazing to think the core processors on an original (1990s design) F-22 Raptor could be duplicated on an embedded PC today. Current Multi-core PowerPC processors and FPGAs would be capable enough. I backed Parallela but haven't received the board yet.
neurotech1 13y ago
A Navy jet like a F/A-18F Super Hornet can fly a "coupled approach" in gusty winds, onto a pitching deck and still put the tailhook onto the #3 arresting cable accurately. Navy F/A-18s can fly with terrain following radar to avoid hazards. Auto-pilot with "Terrain Aviodance" mode is related to the Terrain Awareness and Warning System (TAWS) [1] Auto-Throttle on a commercial airliner like a 777 can correct for deviations caused by gusts and other weather conditions. Careful parsing of METAR[2] information can guide jets away from weather fairly easily. Airline pilots manually set the autopilot to weather hazards but this could be done automatically in the near future. [1] http://en.wikipedia.org/wiki/Terrain_awareness_and_warning_system http://en.wikipedia.org/wiki/Terrain_awareness_and_warning_s... [2] http://en.wikipedia.org/wiki/METAR http://en.wikipedia.org/wiki/METAR

krasin 13y ago

Part of that issue is the lack of high-performance and lightweight board for real-time machine vision (you won't have an NVIDIA GPU + x86 board as a part of your multicopter). All hopes that Parallela board (http://parallella.org http://parallella.org) will make it feasible.

neurotech1 13y ago
Depends on how small, how powerful, how reliable and costs. Its amazing to think the core processors on an original (1990s design) F-22 Raptor could be duplicated on an embedded PC today. Current Multi-core PowerPC processors and FPGAs would be capable enough. I backed Parallela but haven't received the board yet.

neurotech1 13y ago

A Navy jet like a F/A-18F Super Hornet can fly a "coupled approach" in gusty winds, onto a pitching deck and still put the tailhook onto the #3 arresting cable accurately. Navy F/A-18s can fly with terrain following radar to avoid hazards. Auto-pilot with "Terrain Aviodance" mode is related to the Terrain Awareness and Warning System (TAWS) [1] Auto-Throttle on a commercial airliner like a 777 can correct for deviations caused by gusts and other weather conditions. Careful parsing of METAR[2] information can guide jets away from weather fairly easily. Airline pilots manually set the autopilot to weather hazards but this could be done automatically in the near future. [1] http://en.wikipedia.org/wiki/Terrain_awareness_and_warning_system http://en.wikipedia.org/wiki/Terrain_awareness_and_warning_s... [2] http://en.wikipedia.org/wiki/METAR http://en.wikipedia.org/wiki/METAR