Hi all! I'm new here. In fact, I'm new to UAVs. I had a RC plane for my 12th birthday, but it never flew, but I guess my own kids will be luckier! :-)
Some day, I'd like to build an autonomous UAV designed to follow orienteers who are running in the for filming them. Orienteering is a sport where athletes run using very detailed maps. I have an ongoing small project that outlines some background. I would really like to be able to follow them for say 30 minutes. The operator in that case should be able to just point at an athlete or group of athletes and click, the UAV will start to track them (based on their GPS position, and possibly a small radio beacon). The UAV will have to be responsible for its own safety, and that it should not crash into anything. The athletes generally run at a pace of 2-4 m/s, seldom slower, sometimes faster, the average is about 3 m/s. So, if the UAV travels at a pace of 5 m/s between the treetops, it'll need to make evasive moves (or stop, if it is an octocopter, or something like Firefly6) pretty fast if it can only detect the obstacle 2-3 meters away...
So, obstacle detection and avoidance is essential. I've been looking around, but I'm rather inexperienced, so please forgive me if what I write is already completely obvious. I found lots of inspiration in this post, and this device would completely solve my problem, but we don't have that kind of money. I suppose I wouldn't want to put more than one real camera on the UAV, and that camera is busy producing the video. Also, there'll be no screen, no walls or anything simple to just bounce a signal off of. There might be a thin branch of a tree.
After looking for range estimation techniques for a while, I figured, perhaps all I need is detection. So, that's my idea: How about mounting a class 3R laser, and a ArduEye to simply look for the red dot? If the flight path is OK, the laser will simply beam out in thin air. If the laser hits something, and the laser can be detected, then that's quite sufficient for the UAV to know it is in trouble, usually, it should just ascend. Shorter range detection should be employed in addition so that the UAV could use them to figure out what the right move would be.
I'm thinking that the laser should cover more than a single spot ahead, it should ensure that the flight corridor is safe. There are line lasers of course, but I suppose a 3R laser would then need to distribute its energy over a larger area, and thus, make the echo harder to detect. Or does a line laser in fact scan? If not, I guess a dot laser could be combined with a mirror on a motor, so that the motor could ensure that the corridor is illuminated by the laser, so the ArduEye could detect anything in that corridor. And perhaps a simple flat vibrating disk motor could do this job?
What do you think, does this idea sound reasonable?
Replies
I think I'll just respond to myself, just to document what I've found in the last few days, also after discussing with a friend.
For one thing, I'm not so sure the ArduEye is well suited, since you need quite instantaneous readout when there is a spot to be detected. Perhaps a photodiode with a lens that focuses incoming light. Moreover, perhaps visible light isn't that good either. Red laser is where the sunlight is strong. We expect to bounce off green leaves, so a green laser wouldn't have good contrast against the reflected ambient sunlight. So, I looked into infrared lasers. There are some big dips in the intensity of the solar spectrum, many due to the presence of water vapor in the atmosphere. However, I found a scientific paper (behind a paywall, I could access it at work), that argued there is very little dependency of wavelength in the 500 nm to 1500 nm in the attenuation of light due to fog (smoke was a different story) at sea level. So, that's good, what you see is what you get, also in IR. Then, I found a NASA presentation that shows that the reflective properties of leaves are really good above 900 nm or so. And, the combination of little sunlight at that wavelength with high reflectance... I think we have a winner.
Then, I found this product which has very high sensitivity at those wavelengths. It detects like 2/3 of the photons at 940 lm... There's certainly no abundance of cheap laser modules at that wavelength as there are on 650 nm, I haven't found the ideal, cheap laser module. And I haven't found the lens or the narrow-band filter, but I think I'm onto something...
I hope that this setup can detect obstacles some tens of meters away, so that the UAV has time to autonomously ascend gracefully. This could be augmented with a maxbotix range finder, so that if things get too close for comfort, the UAV can take more drastic measures, like if it can hover, stop in the air.