What type of proximity sensor should I use?

Hello, my name is Robby

This is my first time using this forum, I hope it will be useful to my purpose.

I am a college student nearing my completion in Electrical Engineering degree at University of Central Oklahoma. I am looking for help for my capstone project. Let me explain the project:

We are building a dog training device that will prevent dog from counter surfing for food on kitchen counters or tables. The idea is to construct a dog collar with a proximity sensor. We have a Arduino Uno micro-controller that will be placed on the table. when the collar gets 2 or 3 feet near to the micro-controller, it will signal the Arduino. The problem i am facing is what kind of sensor should I use for for collar? I just need it to signal the Arduino when its 2 or 3 feet away from it.

Looking forward to your response, thank you!

Im not so sure on the type of proximity sensor would best fit your application. But 2 or 3 feet in reference if there is no obstruction to the line of path of the sensor, ultrasonic sensor would be my suggestion. You could do radar type but that is much complicated since it will reevaluate every now and then the positioning.

I can foresee and number of issues. You probably need line of sight(Kevin's comment), since RF systems, typically RFID, which may be a good choice in some circumstances, will work through timber and glass table tops, the dog could be underneath the surface well within your stipulated range. Optical or acoustic line of sight may present a further problem, if on a collar, it could well be obstructed by the head, ears, and fur. In both cases if the beam is broad enough, the signal would probably eventually reach the detection system. Otherwise you would need to keep the source well clear of the animals head with some sort of extension on the collar. No matter what system you use, there will always be some variation in signal triggering threshold(due to variations in transmitted and received power), which will probably result in quite a bit of variability in the measured distance. You would probably need to do a Mote Carlo analysis based on the likely variables in the chosen system. Some means of auto calibrating the system automatically would over come some of those problems.
I have done work in the past with "live" systems, in my experience you have to make some decisions around false negative and false positive outcomes, and see how the stats stack up. This means you have to do some real world experiments, trying a few line of sight systems, and see how you go.
One other option is to put a light curtain on the surface, and wait for the animals snout to break the beam. This is likely to mean that you are closer than a few feet for a positive detection, but it will give a good detection rate.
What do you intend to do with a positive detection, a shock collar?!, can't say I like them much, what options are you considering?
Just a lot of rambling thoughts, you have probably thought of most of them, too much time on my hands,
cheers,
Richard

Thank you for replying. Ultrasonic sensor is something I have looked into in the past. I could not find a suitable sensor that detects at 2 or 3 feet. I found many that ranges higher than that, which could be a probably. Is there one in particular you know that will ranges in the distance i need?

Thank you for the reply.

The dog tripping the system when under the table is something i have not thought about, I will have to consider that.

When the Arduino system detects the collar with the transmitter in range, it will signal a speaker alerting the dog in the owner's voice commanding the dog to get down.

The idea of the project is design a system that will train the dog to not counter surf without using any harmful technology. There is also treat dispenser used to reward the dog after a specific amount of time to reimburse positive behavior.

After what you have said, it seems like a motion detector would be the best option. Here is two problems that's in my mind:
1. Finding a suitable detector that will only detect within 3 feet of where it is placed.
The only ones i have found ranges more than 6 feet.
2. The reason why I thought of a collar with a transmitter in the first place is because the system can continue to operate while the humans are around the system. It will not cause trip the system like a universal motion detector would.

Is the collar with a transmitter completely out of the question now?

If you use wireless, you have to consider that an antenna does not have an isotropic radiation pattern, there are nulls and non uniform radiation patterns depending on the orientation and antenna type. Most 2.4GHz transceivers use an on board chip antenna which means a vertical orientation of the PCB. The other option is to use a small wire vertical whip.
You can't make use of the signal output itself, some digital 2.4GHz systems have a register which has Received Signal Strength (RSSI) information, which has to be read.
On lower frequency systems, where unfortunately antennas are quite a bit larger (you could use an electrically small antenna with suitable matching), you may find they have an analogue AGC output, which can be monitored for received signal strength.
In both cases, the closer you get, the stronger the RSSI or AGC signal is, and provided the system does not saturate, it will give some idea of distance, with calibration.
If a detector works at six feet, provided it does't saturate at three feet, and you have access to the system internals, then it should be possible to get a signal that will give an idea of the signal strength, and hence distance.
Haven't checked out RFID, which are quite different in they way they operate, being a transponder based method.
Cheers,
Richard