Instead of trying to caputre reflected light, why can't a camera be used to locate intense light?
A laser should be the brightest light source around and if the brightest spot moves then why couldn't a camera be able to translate a location based off brightest spot and velocity of the robot?
AIman, I assume your question was directed at my post.
>Instead of trying to capture reflected light, why can't a camera be used to locate intense light?
Several reasons;
- Simplicity is a major goal in the design from my point of view. Aside from the obvious benefits of a simpler system (cost and effort to put it together), a simpler system is more accessible by the electronics novice.
- Video processing takes serious processor horsepower, and programming knowledge to use it.
- By modulating the light source, it is MUCH easier to eliminate false signals, and of course this GREATLY simplifies the rest of the processing.
- A retroreflective target (which reflects the modulated output from laser) is much easer to create than building multiple sources that modulate the light. We need a lot of targets, because in a house the view of the targets will often be obscured, so this is another argument against a complicated target.
- We want the Roomba to identify individual targets, so the plan is to put barcodes on the reflective targets. Multiple modulated IR LEDs would tend to create interference with each other of they were very strong at all. By the way, the Roomba manufacturer, iRobot, makes available "Virtual Walls" which put out a fairly weak modulated signal to keep the Roomba from going into another room, but there is no way to distinguish one of them from another
I am certainly open to other ideas, so here is the current plan in more detail in case you have more ideas;
The Roomba will get a laser attached onto it, along with light sensors adjacent to the laser (What these light sensors will be, is still being investigated). Then around the room some retroreflective targets will be placed. When we want accurately determine the location of the Roomba, it will spin in place, and time how long the spin takes. . When the laser hits a reflective target, the sensors detect it, and that time is logged in the Stamp for processing later.
The big benefit of the laser is it will provide a distinct on-off signal when it crosses over a target. A more diffuse light beam, such as a LED, would not.
The processor, a Stamp 2, will use the "hit logs"calculate the angles where the targets were detected. If the location of each target is already programmed in, we have the necessary information to calculate the location of the Roomba vacuum.
Of course, the next challenge is in identifying individual targets from each other. While only two targets are needed for the calculations, we think we will need many targets, because some will be obscured by things in the room. Also, we want the Roomba to be able to go from one room into another and find its way back.
The barcode idea may be implemented in a few different ways. The least demanding on the electronics is to use 1/2" wide reflective tape and put stripes representing a 0 or 1 on a piece of index card. But, I think our goal is to print barcode onto transparency and place that on a wide piece of retroreflective tape.
We've used visible lasers up to this point. The detector chips, which have internal circuitry to detect the modulated frequency and an external pin to drive the external light source, tend to be infrared (we are working with one of the few visible 650 nm detectors right now). Also, their response time seems to be a bit low for our application, which needs to detect stripes at about 8 Khz.
This may sound like a strange recommendation for an electronics parts source, but...
If you just want an inexpensive laser diode, you should try Walgreens drug store. Around here they have had a pen with a blue LED light and a red IIIa laser powered by three LR44 cells (included). They are selling every day for $2.99 OK, I can't resist: "But wait there's more." They also sell them two for $5. That's hard to beat for hacking and easy availability. Heck, Rshack wants $3.99 per LR44 cell on their web site (silver oxide not alkaline though). If you need new batteries for a digital caliper or stopwatch etc. the laser pointer could be considered free. You will find it hanging next to the pens and Sharpies. The SKU is 630990822001. I've heard of people buying laser pointers for a buck, but have not come across any yet.
I've had good luck experimenting with driving a laser diode (from a different hacked pointer) with an SX28 through a PN2222 and chopping at a few kHz at least. I haven't had a reason to drive it any faster yet, so don't know what the upper limit of modulation is.
At the other end of the financial spectrum, if you have the money, the top of the line Leica laser tape has Bluetooth built in (not the Stanley product). You can output measurements directly to a spreadsheet.
In my mind the vision thing is getting way to detailed and leading to confusion.
Yes the human eye uses more things to focuse then a camera, however there are many camera's with auto focus.
A program I am working on uses a camera to navigate with a sonar as a saftey feature. The pictures allow the robot to move in any direction but if the sonar says the saftey limit is violated then the robot stops that going in that direction. In the case of using a PING its only a few inches. My point being that all you really need is a 2D camera image to use for directional findings (i.e. what direction is available to move towards) and a sonar to keep saftey limits for obsticale avoidance.
Hook the sonar up to the camera - top or bottom - and then pan the camera/sonar unit. It works much like the BOE BOT demo that used sonar to center on an object except that this uses pictures with sonar to be sure the way is clear. You· could do this with just sonar, but then if the robot was supposed to go somewhere specific you would either need a map or some other form of identifier. With this set up the robot can wander freely and not damage something or get damaged by something.
What if a laser is used to point outward at a set distance or if you want a tilt device so it can change distance. The light is bounced back to a circular set of mirrors, say something small like an inch wide by an inch tall. These mirrors rotate and are not smooth but placed around the cirumfrence of thier "home" facing straight out so that each one has a gap between the mirrors to either side. In other words, cut the middle row out of a disco ball. The mirrors rotate non-stop while the laser is in use.
Shouldn't we be able to use the dead spaces between the mirrors to capture distance? For example if we cant time a flicker of light to hit a light sensor and the amount of light changes - dimmer is farther - then would it work?
Example ································································· { outline·of dark box ······························ {==========}·············· * Light sensor ·······························{···/············· *}············· / mirror·gets light to sensor ································ <·················}············ Laser is the < ·································· /·l···l··l· \···· }··········· Mirrors in circle /ll\ ·······························{==========}··········· === top and bottom of dark
So in short the laser < fires out a pulse. The moving mirrors·/l l \ catch the reflected light and bounce it up to a stationary mirror / which sends it back to a light sensor. The box/container that everything gets mounted into is light absorbent - black painted·or some dark colored crushed velvet or foam to absorb stray light. The light sensor·won't be a need to be as fast as the laser, only as fast as the mirrors rotate. Of course a sensative light sensor would need to be used.
Just a quick reply to the eye sight thing... If our eye could emit light, and then really accurately time the return, then yeah, one eye would be all we need for depth perception. The reason we have two eyes, and binocular vision, is that we, unfortunately, aren't superheroes.
I vote we work on eye-lasers next.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
--
And, in the end, the love you take
is equal to the love you make.
--
I think you've re-created one of the earlier devices created to measure the speed of light. I suspect you need a lot more speed than you can expect from of a device with glued-on mirrors in order for it to work.
I think you are better off with one of the existing laser distance meters. Leica, and several other makers now produce them in the $200 to $300 range. Towards the top of the range you can get ones which will interface to a computer.
A microcontroller can use only one eye because it measures the time it takes for the laser to bounce off of the object your are aiming at. The Brain requires two eyes because, and I am not entirely sure about this, it focuses on an object and uses the position of the eyes to calculate a rough distance.
You do not have a laser source providing another known variable. You know the direction of the source and you know the spot seen must be on that line, once you have that you can perceive depth with one sensor. This is an easy thing for you to check with a laser pointer and a person holding a target using one eye.
Everyone,
Time of flight is used for laser range finding sometimes but they tend to look at the phase shift between the modulated output beam once received and the signal that modulated it rather than sending a pulse and waiting for it's return, this can be analogue circuitry with the output processed at your leisure. Leica handheld range finders and some of the laser trackers by Faro use these techniques.
I'm not sure if this thread is really being seen by anyone anymore, but i'll post anyways.·
I noticed everyone is talking about a laser range finder method of either time of flight or trig based... but I dont' see anyone here trying to implement a phase shift method.· This is quite simple, and can done with a lot less processing power. Actually it can be with·a plain LED that would be well under the the Watt issues of class III and higher lasers (that is, their nasty habit of blinding people)...· Its mentioned briefly above, but not explained how to do it.·
The phase shift phenom is caused by the delay in the signal strength of a returned signal that is strobed.· That is, if I strobe an LED from off to on, then to off again... do this many,· many, times, eventually that same signal will be bounced off some distant object.· The further away the object is, the greater the phase difference between an outoing sinusodial signal and incoming sinusodial signal due to the delay in time that it takes for that signal to be received back at the receiver after it is transmitted.· So if I keep sending that sinusodial signal, and receive it... I can then take my time and meausre the phase differnece between·strobed led signal·and the received·strobed signal.· The bueaty of·this method is that i can do all day long, and can continiously receive the signal rather than shoot out·a beam and then·quickly setup the receiver.· Plus if you're really clever, you can run some·filtering on the received waveform to remove any low level·junk.. say, the frequency from an AC·lightbulb if it is nearby that might skew your results.·Anyways, I'm digressing here.
So·I thought I'd put my two cents in here and put an example of how to do the calculations required for a phase shift laser range finder.··A range finder could be constructed using the phase shift method·by power cycling an LED and measureing the return signal on a photo diode..
For example if·you want to see the·viabilitiy of·this.. heres a way to determine the phase for say an object a·3rd of a meter away (basically a foot, and for arguements sake I'm using 0.3 instead of 0.33333333333.....)
That object, .3·meters away, would have a light sent to it and bounced back... with the light travelling a total of .6 meters
Light travels at 300 000 000 m/s
So the time it takes for it to travel to and fro for an object .3 meters away... would be
V=d/t
(0.6 (m))/(x ) = 300 000 000 (m/s)· ====> x = (0.6 (m))/(300 000 000·(m/s) )= 0.000000002 (s)
Okay... so 0.000000002 is really FREAKING fast...
where m=meters, s=seconds, and x is the time you are trying to find
Alright, so now onto the next part of the calculations... Next, lets determine how much time passes for one period (T), when we have a frequency of 1 000 000 hz for the strobing of our light source.·
T=1/f where T is the period, and f is the frequency... T will be in seconds
T=1/(1 000 000) = 0.00001 (s)
So what is the phase difference for a third of a meter?
Phase_difference=(time/Total_Time)*360
Where phase_difference is the phase between signals, time is the time it takes light to travel the distance specified above (where the time can be the difference between zero crossings of the recieved signal and the transmitted signal), and Total_Time is the time for one complete period... and 360 is the number of degrees (or also know was 2*Pi) that are in a cycle or period of a sinusodial function
In this case.. time=0.000000002, Total_Time=0.00001
Phase_difference=(0.000000002 (s)/0.00001 (s))*360 (deg)=0.0002*360=0.072 (deg)
Hehe, okay so hopefully I did the math correct in the example.· Alright, so wha tyou can do is modulate the frequency a bit and see what happens.. you'll find the results interesting [noparse]:)[/noparse]·
Comments
Instead of trying to caputre reflected light, why can't a camera be used to locate intense light?
A laser should be the brightest light source around and if the brightest spot moves then why couldn't a camera be able to translate a location based off brightest spot and velocity of the robot?
>Instead of trying to capture reflected light, why can't a camera be used to locate intense light?
Several reasons;
- Simplicity is a major goal in the design from my point of view. Aside from the obvious benefits of a simpler system (cost and effort to put it together), a simpler system is more accessible by the electronics novice.
- Video processing takes serious processor horsepower, and programming knowledge to use it.
- By modulating the light source, it is MUCH easier to eliminate false signals, and of course this GREATLY simplifies the rest of the processing.
- A retroreflective target (which reflects the modulated output from laser) is much easer to create than building multiple sources that modulate the light. We need a lot of targets, because in a house the view of the targets will often be obscured, so this is another argument against a complicated target.
- We want the Roomba to identify individual targets, so the plan is to put barcodes on the reflective targets. Multiple modulated IR LEDs would tend to create interference with each other of they were very strong at all. By the way, the Roomba manufacturer, iRobot, makes available "Virtual Walls" which put out a fairly weak modulated signal to keep the Roomba from going into another room, but there is no way to distinguish one of them from another
I am certainly open to other ideas, so here is the current plan in more detail in case you have more ideas;
The Roomba will get a laser attached onto it, along with light sensors adjacent to the laser (What these light sensors will be, is still being investigated). Then around the room some retroreflective targets will be placed. When we want accurately determine the location of the Roomba, it will spin in place, and time how long the spin takes. . When the laser hits a reflective target, the sensors detect it, and that time is logged in the Stamp for processing later.
The big benefit of the laser is it will provide a distinct on-off signal when it crosses over a target. A more diffuse light beam, such as a LED, would not.
The processor, a Stamp 2, will use the "hit logs"calculate the angles where the targets were detected. If the location of each target is already programmed in, we have the necessary information to calculate the location of the Roomba vacuum.
Of course, the next challenge is in identifying individual targets from each other. While only two targets are needed for the calculations, we think we will need many targets, because some will be obscured by things in the room. Also, we want the Roomba to be able to go from one room into another and find its way back.
The barcode idea may be implemented in a few different ways. The least demanding on the electronics is to use 1/2" wide reflective tape and put stripes representing a 0 or 1 on a piece of index card. But, I think our goal is to print barcode onto transparency and place that on a wide piece of retroreflective tape.
We've used visible lasers up to this point. The detector chips, which have internal circuitry to detect the modulated frequency and an external pin to drive the external light source, tend to be infrared (we are working with one of the few visible 650 nm detectors right now). Also, their response time seems to be a bit low for our application, which needs to detect stripes at about 8 Khz.
Joe Dunfee
If you just want an inexpensive laser diode, you should try Walgreens drug store. Around here they have had a pen with a blue LED light and a red IIIa laser powered by three LR44 cells (included). They are selling every day for $2.99 OK, I can't resist: "But wait there's more." They also sell them two for $5. That's hard to beat for hacking and easy availability. Heck, Rshack wants $3.99 per LR44 cell on their web site (silver oxide not alkaline though). If you need new batteries for a digital caliper or stopwatch etc. the laser pointer could be considered free. You will find it hanging next to the pens and Sharpies. The SKU is 630990822001. I've heard of people buying laser pointers for a buck, but have not come across any yet.
I've had good luck experimenting with driving a laser diode (from a different hacked pointer) with an SX28 through a PN2222 and chopping at a few kHz at least. I haven't had a reason to drive it any faster yet, so don't know what the upper limit of modulation is.
At the other end of the financial spectrum, if you have the money, the top of the line Leica laser tape has Bluetooth built in (not the Stanley product). You can output measurements directly to a spreadsheet.
Good luck with your project.
Rick
www.sales.hamamatsu.com/en/products/solid-state-division/position-sensitive-detectors/two-dimensional.php
kelvin
Well I better get to taking one apart...
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Think Inside the box first and if that doesn't work..
Think outside the BOX!
Yes the human eye uses more things to focuse then a camera, however there are many camera's with auto focus.
A program I am working on uses a camera to navigate with a sonar as a saftey feature. The pictures allow the robot to move in any direction but if the sonar says the saftey limit is violated then the robot stops that going in that direction. In the case of using a PING its only a few inches. My point being that all you really need is a 2D camera image to use for directional findings (i.e. what direction is available to move towards) and a sonar to keep saftey limits for obsticale avoidance.
Hook the sonar up to the camera - top or bottom - and then pan the camera/sonar unit. It works much like the BOE BOT demo that used sonar to center on an object except that this uses pictures with sonar to be sure the way is clear. You· could do this with just sonar, but then if the robot was supposed to go somewhere specific you would either need a map or some other form of identifier. With this set up the robot can wander freely and not damage something or get damaged by something.
What if a laser is used to point outward at a set distance or if you want a tilt device so it can change distance. The light is bounced back to a circular set of mirrors, say something small like an inch wide by an inch tall. These mirrors rotate and are not smooth but placed around the cirumfrence of thier "home" facing straight out so that each one has a gap between the mirrors to either side. In other words, cut the middle row out of a disco ball. The mirrors rotate non-stop while the laser is in use.
Shouldn't we be able to use the dead spaces between the mirrors to capture distance? For example if we cant time a flicker of light to hit a light sensor and the amount of light changes - dimmer is farther - then would it work?
Example
································································· { outline·of dark box
······························ {==========}·············· * Light sensor
·······························{···/············· *}············· / mirror·gets light to sensor
································ <················ ·}············ Laser is the <
······························· ··· /·l···l··l· \···· }··········· Mirrors in circle /ll\
·······························{==========}··········· === top and bottom of dark
So in short the laser < fires out a pulse. The moving mirrors·/l l \ catch the reflected light and bounce it up to a stationary mirror / which sends it back to a light sensor. The box/container that everything gets mounted into is light absorbent - black painted·or some dark colored crushed velvet or foam to absorb stray light. The light sensor·won't be a need to be as fast as the laser, only as fast as the mirrors rotate. Of course a sensative light sensor would need to be used.
Would it work as a range finder?
Post Edited (AIman) : 11/17/2006 10:16:10 PM GMT
I vote we work on eye-lasers next.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
--
And, in the end, the love you take
is equal to the love you make.
--
A friend of my dad's had only one eye and passed his flight exam. No clue how but he did it.
Back to topic
Would the diagram described work?
I think you are better off with one of the existing laser distance meters. Leica, and several other makers now produce them in the $200 to $300 range. Towards the top of the range you can get ones which will interface to a computer.
Joe Dunfee
You do not have a laser source providing another known variable. You know the direction of the source and you know the spot seen must be on that line, once you have that you can perceive depth with one sensor. This is an easy thing for you to check with a laser pointer and a person holding a target using one eye.
Everyone,
Time of flight is used for laser range finding sometimes but they tend to look at the phase shift between the modulated output beam once received and the signal that modulated it rather than sending a pulse and waiting for it's return, this can be analogue circuitry with the output processed at your leisure. Leica handheld range finders and some of the laser trackers by Faro use these techniques.
Graham
I noticed everyone is talking about a laser range finder method of either time of flight or trig based... but I dont' see anyone here trying to implement a phase shift method.· This is quite simple, and can done with a lot less processing power. Actually it can be with·a plain LED that would be well under the the Watt issues of class III and higher lasers (that is, their nasty habit of blinding people)...· Its mentioned briefly above, but not explained how to do it.·
The phase shift phenom is caused by the delay in the signal strength of a returned signal that is strobed.· That is, if I strobe an LED from off to on, then to off again... do this many,· many, times, eventually that same signal will be bounced off some distant object.· The further away the object is, the greater the phase difference between an outoing sinusodial signal and incoming sinusodial signal due to the delay in time that it takes for that signal to be received back at the receiver after it is transmitted.· So if I keep sending that sinusodial signal, and receive it... I can then take my time and meausre the phase differnece between·strobed led signal·and the received·strobed signal.· The bueaty of·this method is that i can do all day long, and can continiously receive the signal rather than shoot out·a beam and then·quickly setup the receiver.· Plus if you're really clever, you can run some·filtering on the received waveform to remove any low level·junk.. say, the frequency from an AC·lightbulb if it is nearby that might skew your results.·Anyways, I'm digressing here.
So·I thought I'd put my two cents in here and put an example of how to do the calculations required for a phase shift laser range finder.··A range finder could be constructed using the phase shift method·by power cycling an LED and measureing the return signal on a photo diode..
For example if·you want to see the·viabilitiy of·this.. heres a way to determine the phase for say an object a·3rd of a meter away (basically a foot, and for arguements sake I'm using 0.3 instead of 0.33333333333.....)
That object, .3·meters away, would have a light sent to it and bounced back... with the light travelling a total of .6 meters
Light travels at 300 000 000 m/s
So the time it takes for it to travel to and fro for an object .3 meters away... would be
V=d/t
(0.6 (m))/(x ) = 300 000 000 (m/s)· ====> x = (0.6 (m))/(300 000 000·(m/s) )= 0.000000002 (s)
Okay... so 0.000000002 is really FREAKING fast...
where m=meters, s=seconds, and x is the time you are trying to find
Alright, so now onto the next part of the calculations... Next, lets determine how much time passes for one period (T), when we have a frequency of 1 000 000 hz for the strobing of our light source.·
T=1/f where T is the period, and f is the frequency... T will be in seconds
T=1/(1 000 000) = 0.00001 (s)
So what is the phase difference for a third of a meter?
Phase_difference=(time/Total_Time)*360
Where phase_difference is the phase between signals, time is the time it takes light to travel the distance specified above (where the time can be the difference between zero crossings of the recieved signal and the transmitted signal), and Total_Time is the time for one complete period... and 360 is the number of degrees (or also know was 2*Pi) that are in a cycle or period of a sinusodial function
In this case.. time=0.000000002, Total_Time=0.00001
Phase_difference=(0.000000002 (s)/0.00001 (s))*360 (deg)=0.0002*360=0.072 (deg)
Hehe, okay so hopefully I did the math correct in the example.· Alright, so wha tyou can do is modulate the frequency a bit and see what happens.. you'll find the results interesting [noparse]:)[/noparse]·