Need advice about sensors for collision avoidance outdoors.

Hi!
I'm working on a robot for outdoor usage. Now I'm reaching the stage were I am supposed to choose some range sensor(s) for collision avoidance and would like some input from a third part (you guys). The robot should be able to work in all kinds of weathers , except snow or perhaps extreme rain, be able to see >3m and have resolution around 1-2cm.

The sensors I am considering so far is:

1. Laser scanner using triangulation, similar to the one used by the neato robotics vacuum-cleaner https://www.neatorobotics.com/

2. Ultrasound with Time Of Filght measurement and signal processing to filter unwanted noise.

3. Millimeter wave radar. Seems incredibly robust, but I can't find any price. Will not try to build one myself.

4. Structured light. Seems interesting.

Vision sensors of any kind will not work I think, since the robot have to able to work nighttime aswell as daytime.

Laser scanners with Time Of Flight measurement is to expensive.

The budget I have is approx. 200 dollars.

Any of you guys have experience of using any of these sensors outdoors? Which of them should I choose?

My experience:

1. Triangulation is OK but difficult to get accurate at longer distances (even 3m) depending on the separation of the transmitter/receiver. Some nice Hamamatsu PSD diodes around but they aren't cheap.

2. Ultrasound sounds possible -plenty of cheap parts, used on cars for parking. Not sure about the accuracy and the speed of sound is not fixed (although that may not matter to you). Not very directional (compared to light)

3. Radar - not used

4. Structured light - not used.

5. Lasers - I have done time of flight and heterodyne. Pure time of flight is tricky at short distances. Heterodyne is easier and can be done with an LED. You modulate you light source with say 10MHz then mix the returned signal with something very close, maybe 10kHz lower. The resultant 10kHz has the phase preserved so you now only need to measure the phase shift of 10kHz rather than 10MHz. You will have an absolute offset though and it will vary with temperature.

I think I would be looking at ultrasonics.

Keith.

Ultrasonics is probably the safest way to go, but laser is still more accurate and if I can make it work I think i will have a better solution. And it's more fun with lasers.

What kind of circuit do you use to mix the signals for the phaseshifting?

Any double balanced mixer. I have used the NE602/SA602 in the past, but I am not sure if they are still available. The MC1496 is another possiblity. There are probably some very fast RF mixers around from people like Analog Devices, although if may be difficult to get them without a lot of superfluous circuitry.

A useful Thesis for you may Michael James Brownlow - Oxford University 1993 - A Time-of-Flight Optical Range Sensor for Mobile Robot Navigation. Unfortunately I only have a hard copy - not pdf.

Keith.

Did you consider the effort for designing a laser distance sensor from the scratch? Profound experience with medium fast, high
dynamic analog electronics is obviously required for it. At a somewhat lower signal frequency range, it's the same with ultrasonic
TOF measurement.

It's O.K., if the distance sensing point is your main project, otherwise, I would fall back on existing sensor solutions.

Re: Need advice about sensors for collision avoidance outdoo

"A useful Thesis for you may Michael James Brownlow - Oxford University 1993 - A Time-of-Flight Optical Range Sensor for Mobile Robot Navigation. Unfortunately I only have a hard copy - not pdf. "

Click to expand...

I can probably get hold of it through my university article database, so thx for the advice. Will check it out.

"Did you consider the effort for designing a laser distance sensor from the scratch? Profound experience with medium fast, high
dynamic analog electronics is obviously required for it. At a somewhat lower signal frequency range, it's the same with ultrasonic
TOF measurement.

It's O.K., if the distance sensing point is your main project, otherwise, I would fall back on existing sensor solutions."

Click to expand...

This is the reason I dismissed TOF measurement of laser from the beginning.
Hopefully triangulation of some sort won't need the same kinds of advanced electronics as TOF flight would, since the biggest problem in that case is signal proccessing in the software.

My main project is actually to evaluate potential solutions to the problem and present it as a prototype, so it doesn't have to be perfect, aslong as I can prove it would work better if i used better signal filters or detection algoritms etc.

The heterodyne approach is still really TOF but it has the advantage of working mainly with lower frequencies. 10MHz would probably be sufficient for you. The main problems will be the temperature drift and crosstalk. The temperature drift can be solved if you have a "reference". In a scanning system this would just be a close target within the system. The crosstalk issue means that you need to make sure very little of the transmitted signal leaks through into the receiver. Whereas this can be a problem with a conventional TOF it is easy to distinguish crosstalk from your drive circuit and reflected signal from the target because of the time difference. With a heterodyne system the crosstalk will add to the received signal and shift the phase by an amount dependant on the relative phase and amplitude of the received signal and the crosstalk.

If you can get the thesis I am sure it will be useful for you.

Keith.

It seems that I can't find that thesis. Is it the Oxford university that has published it?

Yes. "Department of Engineering Science" "University of Oxford" "Trinity Term 1993"

this is the only link I could find to it on the internet:

https://www.amazon.co.uk/flight-optical-sensor-mobile-navigation/dp/B001OPNF9S

Keith.

I have decided that I will probably not try to build any LASER sensor myself. But I would need something of the shelf for testing purpose. Any tips on systems that could work?
Both heterodyne systems and TOF is interesting.

These **broken link removed** might be worth a look. I have used the PSD diodes but not the modules.

**broken link removed** might be of interest.

or

or: **broken link removed**

Search for rangefinder module on Google. Maybe use "-laser" to get rid of those in the search.

Cypress have a complete rangefinder on a chip project. http://www.cypress.com/?rID=21

I am sure there is another Cypress one but I cannot find it on their site, but I have a zip of the complete project of you want it. It may be on http://www.psocdeveloper.com/

Keith.

Added after 42 minutes:

I have uploaded the Cypress PSoC project here. It is in the "temp" area at the moment:



Keith.

Thx for the links and for your help.

But I think you misunderstood me litlle bit, I am still interested in LASER-module, I just don't want bulid it myself.

I have found some systems myself, but it would be nice to hear something about a system from someone who isn't bias.

Though I must say that the Cypress-chip look interesting.

Sorry, I misunderstood. Well, the best laser modules are probably from Noptel in Finland **broken link removed** probably helped by the fact they seem to have continuous projects going on at the Oulu University. They are not cheap though.

One problem with the laser rangefinders will be minimum distance as they are generally be designed for long range (>1km).

Leica probably still sell OEM modules but I doubt they can handle the repetition rate for a scanning system, although I must admit I haven't looked at their product range for a few years. Again, not cheap (no laser systems will be).

Analog Modules make some laser rangefinder building blocks which may be of interest, but I have never used them: https://www.analogmodules.com/

Keith.

Added after 2 minutes:

Also, https://vigo.com.pl/ may be worth a look, and **broken link removed** make some time to digital converters (TDC), I think.

Keith.

Re: Need advice about sensors for collision avoidance outdoo

Thx again for the links and the help.

I will have to consider this a little longer. I was hoping to find something like the one used by Neatorobotics. Their system is apperently very cheap to build by yourself, approx 30 dollars. But I don't know how it will work outside. THey have only stated that it will work in some outdoor environments, without saying more about it.

Re: Need advice about sensors for collision avoidance outdoo

Hewhowalk said:

I was hoping to find something like the one used by Neatorobotics.

Click to expand...

Definitely not a TOF laser system then! While I think you could get a TOF laser system down to <$100 it would take a lot of development and custom ICs (been there, done that). Assuming you don't have a huge budget you will have to be more "creative". Some systems aren't necessarily as clever as you might think. For example, you could just measure returned light intensity. It won't give you distance measurement because of different object colours, but until you get close enough to hit something it probably doesn't matter, and you can use a form of proximity switch to avoid hitting things.

Also, you can do useful things by the optics design. For example, you can choose where to focus the optics and the angle that the transmit & receive optics point to, and their separation (and hence the point of optimum coincidence). So, you can create a peak signal at a certain distance which isn't necessarily infinity.

Anyway, I suspect they are using simple triangulation. My reason for this assumption is that the laser appears to scan in a narrow range. If you use a single photodiode then you can easily measure distance with focussed optics, just not with great accuracy. Such a system would be very cheap - a simple laser pointer, cheap photodiode. It would probably need to be modulated to cater for ambient light.

Keith.

Keith.

Re: Need advice about sensors for collision avoidance outdoo

The reason I considered a TOF-system is because there is systems that i know work outdoors that uses the TOF-principle.

i know that neato use triangulation in their system but I don't know how robust the system is or how bad i will perform in direct sunlight.
They use a laser pointer as transmitter and a cheap cmos image array as reciever. Using a DSP and some signal proceesing, they get a relativly accurate system.

Since I want the robot to perform somekind of collision avoidance algoritm, like potential field, it wont work with just a proximity sensing system.


I have considererd optics like swep focus, but they all demand an active lightsource as soon as the ambient light gets too weak. The solution could be to switch between two different cameras, one normal camera and one nighvision camera, depending on the ambient light.
But thats probably an expensive solution.

Re: Need advice about sensors for collision avoidance outdoo

As an example of simple laser rangefinders state-of-the-art, see Bosch PLR25, which has an about 100 EUR retail price in europe.
It's said to give 2 mm accuray over a 50 mm to 25 m range. There are possbly cheaper devices from far east companies.

I don't had hands on it yet, but I guess it's rather phase difference than pulse TOF.

There is, by the way, a principle difference between single frequency phase shift and pulse TOF measurement. Phase shift can
only give a scalar measurement, pulse TOF can (in principle) detect multiple reflections at different distances, e.g. when
measuring through a window.

Thx for the advice, but the problem with those devices seems to be the interface. I can't seem to find any way to use it with a computer or microprocessor.

I have read about similar divices from Leica and they seems to be using face-shift in their devices in the same price range.

Since I will be measureing outdoors i hope that I won't have to consider windows as a problem.

Re: Need advice about sensors for collision avoidance outdoo

FvM said:

As an example of simple laser rangefinders state-of-the-art, see Bosch PLR25, which has an about 100 EUR retail price in europe.
It's said to give 2 mm accuracy over a 50 mm to 25 m range. There are possibly cheaper devices from far east companies.

I don't had hands on it yet, but I guess it's rather phase difference than pulse TOF.

There is, by the way, a principle difference between single frequency phase shift and pulse TOF measurement. Phase shift can
only give a scalar measurement, pulse TOF can (in principle) detect multiple reflections at different distances, e.g. when
measuring through a window.

Click to expand...

The problem with devices such as the PLR25 is they are slooowwww so no use in a scanning system. Interesting to see the Bosch unit though - I worked on a similar thing 20 years ago but it never made it to market for various reasons.

With regard to multiple targets, as you say, pulse TOF can discriminate multiple targets. Phase shift is tricky if it hits two targets because you will get a misjudged reading from the combined effect of two sine waves. Not usually a big problem, but worth bearing in mind.

The other problem with phase shift is that it is ambiguous after a certain distance depending on the modulation frequency. If the frequency has 360 degrees phase shift at 30m then 40m will produce the same phase shift as 10m. The usual way round that is two use two or more frequencies to resolve any ambiguity. The system I designed used 3 frequencies and while the first ambiguity was at 7.5m I could unambiguously measure over 400m.

Keith.

Re: Need advice about sensors for collision avoidance outdoo

The other problem with phase shift is that it is ambiguous after a certain distance depending on the modulation frequency. If the frequency has 360 degrees phase shift at 30m then 40m will produce the same phase shift as 10m. The usual way round that is two use two or more frequencies to resolve any ambiguity. The system I designed used 3 frequencies and while the first ambiguity was at 7.5m I could unambiguously measure over 400m.

Click to expand...

I've read about that and I actually thought that you could only use 180 degrees phase shift or perhaps even less, before you got ambiguity unless you do several measurements