What's available off the shelf?

[Simon Ritchie]

In this posting I look at what equipment is available off the shelf for Lidar mapping without having to build stuff yourself.

 “Traditional” Lidar sensors use lasers and moving mirrors to make them scan.  They are very accurate and very expensive and designed to be carried in a light passenger aircraft.  Now there is a new technology emerging, solid state lidar using integrated circuits.  In mass production these should be cheaper and more robust than the earlier sensors, but it’s early days.  

   

This is an article from January 2018 about Lidar sensors that can be attached to a drone:  https://www.dronezon.com/learn-about-drones-quadcopters/best-lidar-sensors-for-drones-great-uses-for-lidar-sensors/.  It doesn’t give prices.  A quick bit of research shows that the devices it describes range from $650 to $85,000!

The cheapest is the Leddartech’s VU8 system.  It costs 852 Canadian Dollars, plus cables and accessories.  It’s just the bare electronics so you will also need to buy a case.  It weighs 128 grams, but you need to add the weight of the other bits and bobs, and a small computer such as a Raspberry Pi to control it.  That all looks very promising, but the Leddartech website says that it’s intended for “collision avoidance, navigation and altimetry”.  Nothing about mapping.  Apparently they only envisage you using it to tell how high up your drone is, and to avoid it running into things.  It takes a 12V power supply, which is a nuisance because most drones use a 5V supply.  You will need some sort of power converter, which means extra weight and some power wastage.

The article also lists the Velodyne HDL-32E. The Velodyne website says that it is suitable for mapping.  It weighs 1Kg, which means you will need a fairly hefty drone to lift it.  The website does discuss mapping and shows some impressive pictures.  However, it doesn’t give a price, it just invites you to fill in a form and get a quote.  That’s always a bad sign.  According to The Register website “you can expect to pay $85,000 for one of [Velodyne’s] top-of-the-range LIDAR boxes”: https://www.theregister.co.uk/2017/05/18/lidar_backlog/

This is because driverless cars don’t exist yet and Velodyne are selling their devices in tiny quantities to car designers for research and development.  The same article continues:  “Velodyne told us its goal is to eventually whittle the price down to $250 per unit ‘for solid-state LIDAR sensor[s] when produced in mass volumes.’ Self-driving systems – whether they are part of a car or an add-on feature – are not going to enter the mainstream if they're more expensive than a house deposit, of course. A spokesperson for the Velodyne LiDAR team said there was "no specific date" to reach that sub-$300 price point. "But we’ll be able to lock down pricing once mass-production volumes are identified," she concluded.”

Which is to say, if they can persuade a car manufacturer to use their scanner into its cars, volume production will bring the price down.

Another start-up company Quanergy are working on a solid-state scanner for cars.  If you’re interested in the gory details of lidar scanners, this Youtube video is worth watching:  https://www.youtube.com/watch?v=uPqzo29Q9Wc.  It’s a one-hour seminar given at Stanford University by the CEO Louay Eldada: It’s fairly discursive.  He describes how Quanergy use electronics to make the laser beam scan, rather than mounting it a physical moving platform.  He criticises rival offerings, some of which simply miniaturise the analogue laser beam and moving platform and put it on a microchip, and others which use a very bright (and expensive) laser to light up the scene. 

However, turning a good technical concept into a consumer product is hard.  This article from Bloomberg suggests that Quanergy is struggling:   https://www.bloomberg.com/news/features/2018-08-13/how-a-billion-dollar-autonomous-vehicle-startup-lost-its-way.

Meanwhile, back in reality, what is available off the shelf today?

I came across a German company called Sick (sick.com) recently at a trade show.  Their LD-MRS model costs about $8,000 and weighs 1Kg.  Target applications include “High measurement data processing, e.g. for 3D mapping”.  https://www.sick.com/ag/en/detection-and-ranging-solutions/3d-lidar-sensors/ld-mrs/c/g91913.  Sick make a range of Lidar sensors, but it looks as if this the is their cheapest mapping product.

That looks a lot more promising.  As I described in another posting, the price of your drone depends on the weight of the payload it will carry.  On paper the sick scanner allows you to build a flying Lidar sensor for about $15,000.  Compared to $100,000 or more, that’s a game-changer.  Unfortunately, my personal budget doesn’t quite stretch, so I’m not going to buy one to try it out.

[Simon Ritchie]

I mentioned Quanergy in this posting:  https://groups.google.com/forum/#!topic/lidar-mapping/Kqzxu-0u1g0

This article from Lidar News gives another take on that company:   https://lidarnews.com/articles/uav-lidar-400-ft-faa-ceiling-limitation/.  The author Jeff Fagerman describes how his company is using a Quanergy scanner for land surveying.  He clearly loves it.  He also gives a price - $10,000, which puts it in the same ballpark as the scanner from sick.com that I found. 

That's still a lot of money, of course.  If Quanergy can make their scanner a mass market product, for example by persuading a car manufacturer to use it, it will get a lot cheaper..  That would be nice.  

,

[David Fairbairn]

The Quanergy M8 does make for a good alternative to the lower-tier Velodyne sensors, though I'd be curious to learn more about whether Quanergy and Velodyne are going to be making LiDAR sensors akin to those manufactured by Riegl (e.g. the LiDAR unit in the VQ 1560i, which has max range somewhere around 1000-3000 m). I'm not sure if I would call these LiDAR sensors 'off the shelf' though (unless I misunderstand what you mean). The LiDAR USA/Phoenix LiDAR/Yellowscan mapping units are more what I would say don't involve having to build the system yourself (and they're pricy).

Using the Quanergy M8 or a similar spinning LiDAR sensor certainly involves some building, although the first step is the most tangible - recording the stream of LiDAR data on something like a Raspberry Pi or STM32, and incorporating IMU and GPS measurements into the payload. These IMU/GPS measurements end up needing to be pretty accurate and frequent though - I've played around a bit with having a simple 10 Hz IMU/GPS onboard a drone, and while that might be good enough for something like recording electromagnetic readings, it leads to a lot of smearing and distortion in a LiDAR point cloud. After recording the data, getting the trajectory reconstruction down seems to be a pretty tricky process of using position/orientation data that's been through an Extended Kalman Filter, smoothed forwards and backwards, and interpolated heavily. If you're trying to do this stuff in real-time, then the SLAM/LOAM/loop closure stuff can get pretty involved as well it seems.

[Simon Ritchie]

>  I'm not sure if I would call these LiDAR sensors 'off the shelf' 

A good point.  I guess I meant something that you could mount on a drone, fly it over a piece of ground and get a point cloud out of it.  None of the cheaper Lidar scanners will do that because they are designed for other purposes.  They all need extra bits and pieces and you would have to do work to integrate it all together.

You're obviously ahead of me.  So far all I've done is a bit of work with the RPLidar and some research into the available products.

>  If you're trying to do this stuff in real-time

I guessed that doing that wouldn't be possible.  I envisaged putting a solid-state disk on board, writing the raw data to that and processing it later.  You could start while the drone was still in the air and finish it off after it had landed.  You would need a high-capacity disk but they are small, light and cheap.

> These IMU/GPS measurements end up needing to be pretty accurate and frequent though

What about the Emlid Reach system that I mentioned in another posting?  It's very cheap and it looks good on paper:  https://emlid.com/reach/

Concerning scanners, I'm now looking at the Benewake CE 30 D:  http://www.benewake.com/en/canshu/show-216.html.  Problem 1 is that it needs a 12V power supply.  Your typical drone has a 5V supply, so you would need a voltage converter.  More expense, more weight and more drain on the flight battery.  Or you could have a separate 12V battery, but that would increase the weight and the cost.

For the control computer I would use Pine's Rock64, which runs Linux and has a USB3 channel to control the disk:  https://www.pine64.org/?page_id=7147.  (The Raspberry Pi only has USB2, which is slower.)

Looking on the bright side, that's $1,200 for the scanner, $265 for the Reach, $100 for a computer and a disk, plus cables and something to hold it all together physically.  Maybe $1,500 in total.  The scanner weighs 334g, so maybe 0.5Kg for the whole thing.   You wouldn't need an exotic drone to carry that kind of weight.

Whether that configuration would actually produce useful results is another matter, of course.

[David Fairbairn]

I'm curious about what could be done with the Emlid Reach as well. While it may not compare to the fancy GNSS/INS systems out there like the $15k+ Novatel and Applanix devices, it still looks novel from a price point vs. cost perspective. This Youtube channel seems to have some useful information re: working with it: https://www.youtube.com/watch?v=xxUinOtjG2U 

[Simon Ritchie]

Quanergy have released a new scanner, this time at $5,000.  Jeff Fagerman has given it a glowing revue; https://lidarnews.com/articles/quanergy-m8-ultra-digital-lidar/