I thought about doing a project in which I would try to estimate an object's (mainly human's) position using a few sensors placed around the room. The two main options I see for this kind of project are ultrasonic sensor, or microwave radar.
My first choice would be the ultrasonic sensor, but the ones I could find have a really low measuring angle (around 15 degrees), which is a problem when it needs to observe a whole room. I could buy more and place them so that they collectively see the whole room, but that doesn't seem like the best solution (and it can get pretty expensive too).
Radars are a bit more complicated. Most of what I could find are Doppler microwave radars, which, as I understand, cannot be used for distance measurements. I could also just use a simple antenna for transmitting and another for receiving the scattered signal, but then I had to calculate the distance myself, which I'm not sure I have the sufficient math knowledge for . Also, I'm not sure if a DIY antenna could measure the signal accurately enough for me to be able to estimate the position of the objects.
Which one do you think is the best solution? Is it even feasible to do this with these kind of sensors?
You seem to misunderstand how echo location works. Ultrasonic and radar are ALL echo location systems. The controller sends a pulse of either sound or microwave energy and waits for the echo return. The exact time from transmission to reception is the time for the pulse to travel to the object and return. From that you can calculate the distance from the pulse sender to the reflecting object.
All have their limits. Your design must take this into account. You could as you mentioned use multiple ultrasonic sensors. But you would have to sequence their operation so one did not hear one of the other sensors.
as stated by Paul, You really have no idea of how what you say you want to do would work. you want sensors with as narrow a window as you can get. a 15 degree window will only give you about a 25% accuracy (likely much, much less) over a 90 degree field of view (window). You would sweep your sensor over the 90 degrees, at known steps, and note significant sensor returns. that would only give you a relative bearing of the target from the sensor's position. you would need to do the same sensor sweep from another known sensor position, to get an additional relative bearing. then a little trigonometry will give you the target's approximate position (depending on the accuracy elements mentioned above). do a lot more study, or pick a new project.
Actually, I believe there are some computer vision applications that would run on a Raspberry Pi with a webcam, so perhaps a couple of those in the corners near the ceiling.
Perhaps mount the ultrasonic unit on a servo operated mounting and "scan" one side of the room limited by opposite wall size.
Another on the 90 degree wall.
Time so servos move and sensors transmit at independent and correct times i.e. wait for servo to stop, then transmit 1 unit, wait for a return, move other servo etc. etc.
Paul_KD7HB:
You seem to misunderstand how echo location works. Ultrasonic and radar are ALL echo location systems. The controller sends a pulse of either sound or microwave energy and waits for the echo return. The exact time from transmission to reception is the time for the pulse to travel to the object and return. From that you can calculate the distance from the pulse sender to the reflecting object.
All have their limits. Your design must take this into account. You could as you mentioned use multiple ultrasonic sensors. But you would have to sequence their operation so one did not hear one of the other sensors.
Paul
They can all use echo location, but as I understand Doppler radars don't. I found a couple of articles that state that you can't extract distance data from these sensors.
Thanks for the tip regarding the interference of sensors, I'll look into that one too.
jremington:
You can buy such a system, called Pozyx. It is not cheap because technically, indoor localization is quite a difficult problem.
The point of this project isn't that I have this system, but to learn how it works and how to do it myself. I understand that it's a difficult problem, but if it's possible with these sensors, I would try it at least.
123Splat:
as stated by Paul, You really have no idea of how what you say you want to do would work. you want sensors with as narrow a window as you can get. a 15 degree window will only give you about a 25% accuracy (likely much, much less) over a 90 degree field of view (window). You would sweep your sensor over the 90 degrees, at known steps, and note significant sensor returns.
So is it not possible to measure the distance of multiple objects at once by measuring the reflection times?
123Splat:
do a lot more study, or pick a new project.
Of course I will do much more study, that's why I do this project . I just needed a general direction to go to with choosing the right sensor, as I don't know how feasible to do position estimation with them.
Paul__B:
Actually, I believe there are some computer vision applications that would run on a Raspberry Pi with a webcam, so perhaps a couple of those in the corners near the ceiling.
Yeah, that was my backup plan, I just thought that it would be more difficult to measure distance from a video than using sensors made specifically for it.
So is it not possible to measure the distance of multiple objects at once by measuring the reflection times?
How would you know what reflected the pulse? You send a pulse. It bounces off of several things. You measure the time needed for the first echo to return. You can't know how many echoes return.
geiszla:
My first choice would be the ultrasonic sensor, but the ones I could find have a really low measuring angle (around 15 degrees), which is a problem when it needs to observe a whole room. I could buy more and place them so that they collectively see the whole room, but that doesn't seem like the best solution (and it can get pretty expensive too).
One alternative to using a lot of ultrasonic sensors is to mechanically scan an narrow beam sensor. This is similar in principle to a spinning radar antenna. There exist kits to do this, but it wouldn't be difficult to prototype with cardboard and a glue gun.
The standard HC-SR04 ultrasonic sensors will detect only the first return, that is, the closest object seen on any one pulse. Thus the "image" of the room would be two dimensional in the sense that further away objects would be completely hidden behind closer objects. It is possible in principle to detect objects at multiple ranges per pulse, but the processing to do this gets very complicated.
Thanks for the anwsers, I think I'm getting it now. For now, I'll just stick with computer vision, as that seems the best and simplest way of tracking objects in real time. If that proves too difficult I'll switch to mechanical scanning.
. Also, I'm not sure if a DIY antenna could measure the signal accurately enough for me to be able to estimate the position of the objects.