Amazon reveals drone delivery safety details

http://www.washingtonpost.com/news/innovations/wp/2015/07/28/amazon-details-its-plan-for-how-drones-can-fly-safely-over-u-s-skies/?tid=HP_national?tid=HP_national

Some highlights

  • Altitude will determine low speed, high speed, and no-fly
  • Urban areas will have the best collision avoidance drones
  • Network will provide collision avoidance updates i.e. other drones, updates from customer, helicopter vicinity
  • Four class levels of types of drones to deploy
2 Likes

I actually just heard the FAA administrator talk about this topic, and he shared his concern that the traditional aircraft mentality is “safe by design” while the silicon valley design mentality is “move fast and break things”.

I think that is a legitimate concern, and it will be a while before Amazon is allowed to do what they’re asking for.

The present FAA rules - LOS, operated by someone with a pilot’s license - will never allow for viable commercial operations for Amazon or most others demanding autonomy. Even with a relaxation of the LOS and pilot’s license requirements, having someone actively fly (or even just monitor) an autonomous drone (or two, or four) is less effective than having someone operate a truck that can also make many times more deliveries per unit of time, without the mass/volume restrictions a drone will face.

The framework sketched out in the article seems reasonable at first glance. Autonomous, high-speed drones with some considerable heft will need excellent sensors and good P2P communications. But making that work reliably and with sufficient resilience to handle unforeseen events - loss of uplink to FAA/ground control, malicious actors, catastrophic weather, countless platforms - will take some work. I can see some synergy with the work being done on self-driving cars - especially the various ideas being kicked around with P2P communications about road conditions.

The good thing about drones is that the models being proposed by Amazon aren’t going to be too large or heavy, thus the risk presented to the public will likely be small enough that the FAA allows for some experimentation. suspect that the first live pilots will do the bulk of their flying over utility right-of-way, greenbelts, public roads, and then private property so as to minimize risk. Even without the P2P networking that the article talks about, I think that these drones can be made reasonably safe in a few years. Bet they even test them out at that University of Michigan test track (along with the cars).