Last Updated: Feb 19th, 2018 - I will be updating my original post as people correct me and I learn new things.

Well, I am next off to monkey around with LIDAR technology and interestingly there is not one single document available that can explain different LIDAR technologies available for PixHawk "or" Pixhawk compatible devices, what each device does, limitations, programmability etc. 


I will attempt to take a shot at what I have learned so far but I would like the community to chime in so we can have a consolidated document for future reference. Just like a wiki page.

LIDAR = A detection system that works on the principle of radar, but uses light from a laser.

Applications:

1D - Range Finders. Typically work in combination with FC such as PixHawk.
2D - Short Range Obstacle detection.
3D - Used for surveying, mapping and point clouds. A method that measures the distance to a target by illuminating the target with pulsed laser light and measuring the reflected pulses with a sensor. Differences in laser return times and wavelengths can then be used to make digital 3-D representation of the target. 

LIDAR Types: So it appears there are two types of LIDAR sensors as far as the technology goes behind it.

1. traditional LIDAR sensors
2. Time of Flight (ToF) - More to come later...

SETUP:

So what are the components of an overall LIDAR system/solution?

Well, first you need to decide what the LIDAR will do in the real life world. That will determine the type of LIDAR sensor you will need for your fixed wing aircraft or the copter you are planning to use.

Do you really need a flight controller like PixHawk? the answer is NO BUT it depends on the application. If the application is rangefinder or does object avoidance, Pixhawk along with Mission planner can support such application on a drone.


If you are going to use the LIDAR for Mapping or analysis, that's whole new setup but do you really need a flight controller for such LiDAR to work? My understanding is NO, as long as we can provide LIDAR with the position of the aircraft using a separate GPS and IMU for processing data.

LIDAR SENSORS - Bufete :)

I am planning to put a power point presentation at the end which will consolidate all the tips, discussions here for future reference. Thanks for joining the discussion.

Views: 1920

Replies to This Discussion

@Igor Nastashchuk - I think Laser Developer has provided some good reads. check them out. I believe the technology is still evolving so let's see what others have to share with us.

I have to apply pollen to the tops of trees at 8-12 feet. I would like to use LIDAR to get this close but adjust for orchard canopy variation. Which unit is best for this? Will it engage in an AUTO mission in TERRAIN mode? Will it work on PIXHAWK 1 or only on PIXHAWK 2? I was thinking of testing with the Benewake Mini.

Chris Anderson said:

You've got a wide range of sensors there, which all do different things. They break down into this:

1D: These are just range finders. (1,2,4)

2D: These are useful for short-range obstacle detection (3,5,6, although you missed the best of them, the RP-Lidar A2)

2.5D: These are meant for creating point clouds. (6,7,8)

I have updated the table with range support of each LiDAR.

The PixHawk compatible LiDAR works with both PixHawk models. I have not personally tested it for what you are trying to do, but I think LiDAR Lite V3 should be sufficient for what you are trying to accomplish and it's fairly cost-effective. You will have to set your waypoints using absolute height inside Mission planner and using LiDAR to avoid drone crashing into the canopy of the trees.

Jimmy Oliver said:

I have to apply pollen to the tops of trees at 8-12 feet. I would like to use LIDAR to get this close but adjust for orchard canopy variation. Which unit is best for this? Will it engage in an AUTO mission in TERRAIN mode? Will it work on PIXHAWK 1 or only on PIXHAWK 2? I was thinking of testing with the Benewake Mini.

Chris Anderson said:

You've got a wide range of sensors there, which all do different things. They break down into this:

1D: These are just range finders. (1,2,4)

2D: These are useful for short-range obstacle detection (3,5,6, although you missed the best of them, the RP-Lidar A2)

2.5D: These are meant for creating point clouds. (6,7,8)

So will the LIDAR push past the absolute height say if a larger tree is in the mix?

I recently bought 3 Lidar and mount one on a plane with Arduplane software.

From my tests the Lidar on Arduplane 3.8.4 is only used in Land mode since in Auto mode the EK2_RNG_USE_SPD will prevent from my understanding the use of the Lidar since almost no plane can fly at less than 6 m/s speed.

EK2_RNG_USE_SPD: Range finder max ground speed

max range 2.0 - 6.0

The range finder will not be used as the primary height source when the horizontal ground speed is greater than this value.

Makes sense. Above those speeds reaction time cannot be trusted. I am running missions at 5.38 (12 mph) so I could be in luck. I have purchased the Benewake TF01 so I am hoping for good things. Good fortune if I can run the hills. I will test PH1 & PH2 and post any success for others to learn from.

You and Luc have raised some good questions. So what is the math behind Lidar accuracy vs. Speed vs. Height?

It appears so far LiDar would fail to perform any tasks at high speed and furthermore if we are beyond LiDAR maximum read capacity it would not provide any data.


Any takers?

@Laser Developer - Do you guys have any white papers that describe and explain the math and science behind how LiDAR works?. More in-depth discussion than jus thigh-level explanations I read in your users manuals so far.

Jimmy Oliver said:

Makes sense. Above those speeds reaction time cannot be trusted. I am running missions at 5.38 (12 mph) so I could be in luck. I have purchased the Benewake TF01 so I am hoping for good things. Good fortune if I can run the hills. I will test PH1 & PH2 and post any success for others to learn from.

I found this white paper. It's a starter.

http://web.pdx.edu/~jduh/courses/geog493f12/Week04.pdf 

or read enclosed doc..very interesting for sure.

Attachments:

Well LIDAR does move at the speed of light and with reflection 1/2 the speed of light. Is the copter's computer fast enough to react? I have seen a few LIDAR videos and the PH1 seems to be slow to react although the unit was nadir going over cars, etc. If the angle was fixed 30 to 45° downward and the tree canopy could send back enough reflection I could have good luck. The problem is that the job calls for pollination on some orchards that are merely sticks in the air (Cherries) whereas Almonds have significant bloom to catch the beam(s). I am currently in California central valley so most orchards are topped and the ground is level, but I have calls for Avocados on hillsides so this is a problem that needs solving.  

There is very little detailed information available on how a LiDAR works at the electronic and signal processing levels. Most companies that make LiDARs keep their designs private. High level overviews on how a specific LiDAR performs are available but these are mostly for marketing purposes and don't give away any interesting technical information.

A number of universities have post graduate programs focussed on LiDAR technology and their thesis projects are often available to the public. These tend to emphasize how difficult it is to make a LiDAR work properly rather than offering practical insight into the design of a LiDAR.

The current state of technology is that, for practical purposes, anything that you can see with your eyes can be measured using laser light. So while a LiDAR won't see through walls, it can measure the tops of trees and the ground (simultaneously) at very high speeds. As far as a laser pulse is concerned, the world is a static environment and even the speed of an aircraft is negligible.

However, most LiDAR systems take more than one laser pulse to make a measurement, sometimes thousands of pulses. This gives better accuracy and improves the signal-to-noise ratio. The time between laser pulses becomes the limiting factor and it can take a significant time to get a result. Movement of the target during this time can lead to errors.

Unfortunately, what a LiDAR can do compared with what you can actually buy are very different things. With so few companies developing LiDAR systems, there is a very limited range of products available. This product range is dictated entirely by commercial needs - if a company doesn't make a profit then it just goes away, forever.

What is peculiar about the LiDAR industry is that there is a lack of basic knowledge available to engineers, enthusiasts and want-to-be entrepreneurs. This is completely different to other technology fields where gigabytes of information can be gathered in seconds and many companies provide raw components, data sheets and reference designs. The situation is so bad that I have not found even one design document that correctly describes in detail how to make a direct time-of-flight LiDAR using components that you can actually buy somewhere.

With interest in LiDAR at an all time high, the industry is ripe for a disruptive attack on this problem. But I guess that is another discussion ;).

This is reflected in the lack of tested units and available media with any concrete use situations. Randy McKay is the only person on YouTube that has a complete test in a use case scenario with a LiDAR and a Pixhawk in an terrain following auto mission. This is over flat ground (what seems to be a ski area sans snow) At this point terrain following is all I am interested in achieving. The scanning looks very cool, but I have no experience in how to bring this to a client with data he/she can use. The recent scan of Purépecha Empire city in Mexico is really groundbreaking in my opinion. The ability to find this kind of data before exploration must be a great benchmark for funding.

McKay LiDAR video

Purépecha Empire city in Mexico

@Jimmy Oliver - As long as you promise to send all of us a bag full of California Almonds, Cherry jams we are all in it :)

I think I understand what you meant by LiDAR move at the speed of light. I think you meant the laser source is travelling at speed of light, But the LiDAR sensor scanning is a derivative of two moving components. The aircraft speed and the housing that controls the movement of the laser source (if the laser source is rotating for scanning)

I have seen Randy video before. He is great! is he located in Japan?

That said if I make a bold guess here the key LiDAR sensor parameters one should look for are;

1. Distance limit.
2. Refresh rate, which directly defines the maximum speed an aircraft can fly before senor starts to lose its accuracy.
3. The processing speed of the IMU to process incoming data and can store it for real-time or future processing.
4. Other soft parameters will be;

     a) Lens type used.
     b) Stepping accuracy of the laser rotating source to catch the reflective beam on time.

For what you are trying to accomplish, I doubt any off the shelf solution will work in a plug and play mode without any customization.

You are trying to solve two challenges:

1. Maintain absolute height - which is easy to accomplish.
2. Push drone to predefined push limit in advance if one tree height is higher than others.

How about this approach? First, you scan the playing field in 3D. Develop a Geo Map of the area which I am assuming will help you to predetermine the different heights of the trees, and then you construct a flight plan accordingly?

Has anyone ever thought of using a blimp? Slow and steady probably more viable for pollination application.

RSS

Groups

Season Two of the Trust Time Trial (T3) Contest 
A list of all T3 contests is here. The current round, the Vertical Horizontal one, is here

© 2018   Created by Chris Anderson.   Powered by

Badges  |  Report an Issue  |  Terms of Service