We've been working on a laser scanner that is light enough to be carried by a small drone. Click on the image above to see a short video of the scanning action. This prototype is based on the SF30/C and an extremely lightweight motor drive. Four slip rings are carrying power, alarm and an analog image that is updating at 2000 readings per second. The unit gives saturated laser coverage to detect even narrow obstacles.
The 'scope trace below shows the distance (yellow) and alarm (blue) outputs for a 360 degree scan. Straight forwards is in the middle of the trace.
The alarm has been set to pick up the four narrow poles spaced around the center of the scan. The unit can measure out to a radius of 100m and draws less than 500mA at 5V during operation. Interfacing is easily done by either monitoring the digital alarm or collecting analog results in order to draw SLAM maps.
This scanning laser can be used for accurate positioning in complex indoor or outdoor environments or for obstacle detection whilst the drone is operating close to the ground.
We will be packaging this prototype over the next few months and plan to release it with a retail price of US$999.00
Comments
@Michal Rus Here is what LD said about power lines: http://diydrones.com/profiles/blogs/overhead-power-line-detection-a... Cool stuff!!!
Great news ! When it's available to buy, please let me know. I have a clients waiting for this. When the integration with Pixhawk is ready, you have a market here. I was hesitating about the new DJI's obstacle avoidance sensor, but I since I'm not a DJI user, I'll wait until something similar is available for Pixhawk. Laser is surely better approach than machine vision, especially in bad weather. However I'm not sure if the resolution of your sensor would be good enough in a situation like this : power line inspection from 20m close to the lines. This is my main application.. Some people are afraid to fly too close when you can hit the lines easily, especially in beyond-line-of-sight scenario where your only eyes are FPV video from thermal camera with narrow optics.. I wonder if the communication with Pixhawk would be done over the same 3.3V ADC port as your other products ? And what would happen if I want to use 1 SF30C for terrain following auto mission/landing and this new sensor for colisiona avoidance simultaneously ? I guess, 6.6V ADC port would be used then ? Or another solution ?
Thanks and keep me informed please ! Good work !
@Thorsten,
i'm start to integrate in my companion computer on Spark the support of RTK GPS with mine VR GPS 8T , could be nice to discuss about thist task ... :) The Lidar is integrated in the same project.
In the video clip we can see that the whole assembly is rotating, is it possible to have the beam steered by a spining mirror (co-axial to laser emmiter)? This would mean lower spinning mass and maybe faster spin rates depending on what angular resolution 'you' were looking for.
Also on encoders, former company used to phase lock/multiply up their optic encoders to get high resolution. Good if spin rate is fairly consistent.
@Rob - you asked about the number of steps in the encoder, you didn't ask about the pointing resolution. The encoder is only used to apply periodic and non-periodic corrections to the main pointing mechanism. The drive that we're using can easily achieve an aiming resolution of 1/40,000 per rev. This is too fine for anti-collision and SLAM applications which is the target market for this product, however it's nice to know that it's available for future mapping products.
The choice of angular resolution and firing rate that we're using is to allow us to get 100% coverage at all distances with a 0.2 degree beam divergence, in other words, 2D point cloud saturation. Accurate mapping units would have a narrower beam and a number of other enhancements that would go with higher overall precision.
Ok, so at 1000 steps on a 360 degree arc, that's 0.36 degrees per step. At 50m range, each step is 310 cm wide. So that's fairly course resolution. Probably good enough for collision avoidance, but I don't know about mapping.
@Roberto - thank you, we'd love to get your input on the system integration side. This is an area where we see the biggest hurdles.
@Rob - we've gone for a very new technology on the encoder that theoretically can give us tens of thousands of steps per revolution. However we haven't finished all the testing yet and in practice we would be happy with a thousand accurate steps with linear interpolation between them.
@Thorsten - we'll talk more.
@johnkowalsky - I can't tell you much about the motor / encoder combination except that our main objective with this unit was to keep the weight as low as possible so it's custom made.
What kind of motor is that ? You say it has an encoder ?
Roberto, great idea!
I have to contact you because of the RXM-RAWX logging.
Cheers,
Thorsten
PS: yes a separate data logger would be required I guess - if there would be some synchronized timing/trigger events - like CAM_TRIGG_DIST it should be possible to merge the data from the flight controller and the LIDAR
Cool stuff. I was going to ask if there was an encoder, but you state there is? How many steps?
-
1
-
2
-
3
of 3 Next