The Lily Camera drone offers autopilot supposedly so reliable, it doesn't even come with a controller.
There has been a huge amount of hype in recent days about the Lily Camera Drone, a product that flies itself while taking pictures or videos of whatever the user may be doing. This type of autopilot offers extreme promise, but it is a technology that UAV companies have been struggling with for some time, and no one has quite perfected yet. Unfortunately, while many advancements have been made in developing a good autopilot, products like the Lily are likely to disappoint real world users.
A good autopilot system is one of the most useful features of drones. From inspecting power lines, to surveying crops, to monitoring construction sites, autopilot means that you don't need an experienced operator to use the drone. However, anyone that has used these autopilots, knows that there are a myriad of problems keeping them from going mainstream.
Some of the problems are small and good progress has been made in addressing them. Sometimes the tracking system just has a bad connection and the drone doesn't follow it exactly. Or there is interference in the GPS or compass based guidance system and the drone may fly away. These can be serious issues, but are ones that current technology, if well executed, is capable of addressing.
The most serious problem for autopilots, and especially the Lily Camera, is obstacle avoidance. No mass market drone yet has any reliable form of collision avoidance. This means that any drone on autopilot, especially those at lower altitudes, can crash into an obstacle with no way of automatically staying safe. If you are skiing, it is extremely easy for the drone to hit a tree, or potentially another skier, posing a serious safety issue. It might hit a bridge, power lines, trees, or banners, completely wrecking the drone. Anyone that has used "return home" features on other drones knows that they suffer the same issue, and the autopilot may confidently fly straight into a tree on the way home. In the real world, this may do $1,000 damage or more to the drone itself, making it a very expensive problem.
The issue with collision avoidance is that it is technologically extremely hard. A drone has to be aware of obstacles in front of it, behind it, to its sides, as well as above and below. This means a lot of cameras or sensors. Added to that, it has to detect things as small as power lines, which are invisible to many available sensors or hard to detect. It would also need to know about any people that may get too close to the drone, such as another person. Finally, has to recognize any of these problems this with enough room to stop or change course, which is difficult if the drone is going 30 mph or higher.
Here, a DJI Phantom under its "return to home" autopilot is unable to avoid a very large and obvious rock.
None of this technology is impossible. In fact, there are dozens of companies working on extremely promising technologies in this space. However, they are all still just experimental or prototype products, and none has been demonstrated to be truly reliable. It will likely be a few years until a reliable obstacle avoidance system can be developed and integrated into off the shelf consumer and business drones.
The Lily Camera is an exceptional product, and hopefully it can deliver on its promise. However, the technology still is not the autopilot panacea that many people think it to be. It will still crash into trees, bother the public, and be more complicated to use than it seems. Despite this, there will be a few people for whom the Lily is the perfect drone. For the others, you may just need to wait until autopilot becomes more reliable.