We learned last week my golden retriever has bone cancer and we will be amputating his leg and likely starting Chemo very soon. Consequently I'm moving his dog door to the main fenced in backyard of our house in order to avoid stairs and allow him access anytime. We have another old girl who would appreciate the same thing.
I'm not too worried about the code but I've so far been surprised to not see a clear tech choice for an application like this. Essentially I'd just like to toggle a motor and lock sequence based on a unique ID that can be triggered by a tag within a few feet (1M).
Here's a breakdown of my research so far (and this may be a wrong interpretation on my part):
1a. Passive RFID - not enough range, and particular about orientations.
1b. Active RFID- maybe this is viable but I haven't been able to find a low cost tags plus scanners that are also small and ruggedized.
3. BLE- these seem like a good candidate, the nordic chips seem like overkill for what I need, but when touting their battery life, size, and convenience this might well be the winner.
4. RF- 433Mhz- I've never used these before but I think I actually have the modules. Cheap and easy seems appealing.
5. WiFi - probably not going to be efficient enough to be practical.
6. Retail- I backed a kickstarter once that ended up sending me two Stick n' Find modules. Miraculously I still have them. They are nice and small and utilize bluetooth claiming up to 1yr shelf life. Again with the overkill but givent their longevity and their size (about that of a quarter) not bad. Also, if I was able to even see the device pop up (when in range) my primary goal is already done.
7. Retail 2- my brother got me a MYNT module- much the same thing and reasons for the above.
8. IR sensor- This could work but I worry would be unreliable, particularly when surrounded by a giant long furry chest. Maybe I'm wrong here. Not sure if the battery life is comparable.
9. There are assorted amazon examples here and here but I don't see any indication you could get them to do anything with Arduino at all..
I'm sure there are more that I'm missing.. but you get the gist. I'm not going nuts with security features, we're usually home anyway and if not, this dog door is proabably the hardest way to break in in any case, so just using a particuar frequency or modulation would probably be fine so that it only opens for dogs. (this is why PIR or radar aren't good options) The only bandwidth I need is to transmit some (any) type of ID that can be used uniquely and received from either side of the door.
Anyway, most of the challenge with this one will be executing a good mechanical solution, but if I could just identify a source for a tag that can work with Arduino I'd at least have the electronic element sorted out. Thanks for any suggestions!
-k
Hi,
This video may be of interest to get an idea of what you can expect from an RFID tag
Searching for "Automatic chicken coop door opener" may give you some ideas on how to build the door?
Maybe along with some kind of sensor on the dog you could also have sensors which detect human activity in the are (anything higher than a dog) and have it lock the door or warn you if anything suspicious occurring?
1 Like
There are several doggie door projects that have been done on this site. A search should reveal those projects. I remember one that even detected if the animal was passing through the door and kept the door open for the animal.
1 Like
RFID seems to be the usual solution, often with a big antenna wrapped around the dog's door. The issue with most of the other solutions you mention is that they have too long a range, so there's the issue of the door being unlocked whenever the dog is somewhere in the vicinity, or indeed, just in the house in the case of WiFi.
Take a look at commercial offerings and see what they use - I'm pretty sure though that the system my sister uses for her cat is passive RFID.
1 Like
Thanks for the feedback! I'm going to start the mechanical part today and automate it some more later as the concept evolves. I don't think there's any retail solution that would be as robust and integrated as I'd hoped and although its some work the end result I have pretty good confidence and experience with. I used RCWL-0516's (radar) sensors for two halloween projects and they were pretty good. I could use that as a stand in for now until I find a better solution that I could toggle on and off for when we leave the house. Obviously this doesn't uniquely identify a dog but better than nothing. Its true that most (particularly rf) solutions have a longer range than needed but also examples like BLE have a range indication I think I can use- maybe? Anyway, I did search the forum and I guess I was a little surprised at the ambiguity. I'd hoped this would be a silly question with an obvious answer and somebody would just say, "no, dummy- you just use this fob from amazon like everyone else!" Anyway, I guess the research continues, thanks!
Adding more poo! onto the pile, you may find that you can use a camera and some software to ID your dog over other dogs. If the algorithm ID's your dog, open door. Just as an FYI.
1 Like
haha.. you must be on my wavelength because I've been mulling that idea also for an automatic dog feeder. Our golden will gobble everything in sight and has special meds now, but our other is a grazer. So further down the line I could see that project happening sometime. If I could find a good solution to this tag problem (eg. maybe camera/AI) it might be good for both projects. Of course with the door project I'd need cameras on both sides.
2-3ft for this automatic dog door example... I wish I could just harvest the sensor from one of these. Surely it must be aviailable somewhere.
Then this link might interest you, TensorFlow-Lite-Object-Detection-on-Android-and-Raspberry-Pi/Raspberry_Pi_Guide.md at master · EdjeElectronics/TensorFlow-Lite-Object-Detection-on-Android-and-Raspberry-Pi · GitHub, is what I used to get started.
I got an IR camera doing recognition of 3 objects called 'squirrel', 'duck' and 'human'. 'squirrel' and 'duck' evolved into size of animal Tensors with "squirrel' being the small animal Tensor and 'duck' being the mid sized animal Tensor. The 'human' Tensor is used to trigger a camera to save the images. I'm currently adding another object detection Tensor to the project. My current fancy is to "watch" a detected object, if the object can no longer bee 'seen', then take images and save them. Just to see if I can and if so how.
I use ESP32's to gather the data, I use a Raspberry Pi to processes the data.
1 Like
I just got and started to look at this tutorial from Dronebot Workshop. Link: Machine Vision with HuskyLens - YouTube
The board has lot of processing capability and even a button "learn". (Learn the picture).
1 Like
Thinking about your project, I'd use Arduino_KNN - Arduino Reference, and a Arduino BLE 33 or a ESP32, to ID your dog.
Simple machine learning with Arduino KNN | Arduino Blog.
There are other KNN examples on the internet, "arduino knn library" are the words I used.
1 Like
Here's an idea using an Arduino and a magnetic loop. The article includes a link to the libraries in their github.
It's the same principle as the wireless "fences", but instead of a shock, you could use an ESP-01 and publish an MQTT message when the dog is inside the "fence". (Or whatever communication protocol you're comfortable with).
That's super cool and I want to play with it whether its a fit for this or not. I added it to my Xmas list! I'm sure I can think of lots of ways to use something like that. wow. For this particular application I wonder if I can add a 2nd camera head (maybe just a mirror?) and then read alternate image sources every second or so. That solves the door problem.
That's especially interesting as I built a circuit with some similarities that actually uses an LED as a sensor (in reverse) to relay the color of my DJI Mini 2 front LED and react accordingly. This would pretty much be the same thing I imagine but with more variance of the input data. Have you tried this yourself? A nice perk, having this work regardless of whether they're wearing their collars or not!
I have used KNN to detect objects in an image.
thanks, I actually have a pair of unused ESP32-Cam boards lying around so this was an interesting read. I will have to keep an eye on these type applications. If I was going to try the AI/cam approach I think so far I would probably try the Huskycam board mentioned above. The demo is very impressive and I'm still a novice coder so I particularly liked the simplifed interface and integration.
Hi Salukikev,
thanks for the answer. I have in mind the realization of a system to feed some stray cats, and I have to check the presence of theyr.
Andreas Spiess is a very good trainer. When I need sometthing He is the first souce.
Have a nice Job and best wishes for your Dog.
Ciao
D
The imaging solution sounds entertaining. I'd be interested to know how well it functions in practice, especially if the dog isn't looking at the camera.
