I live in the EU, where I have five separate waste streams: glass, paper, plastic/metal, RVM plastic, and general trash. They all fill at different rates, while the collection schedule is essentially nondeterministic. The only reliable signal is that they arrive at 6 a.m. sharp in trucks specifically engineered to maximize acoustic output.
My solution is similarly overengineered: a vision system combined with an olfactory system feeds into a biological neural network, which monitors the bins and determines when intervention is required.
The neural network is annoyingly power-hungry, occasionally unreliable, and has yet to support OTA firmware updates. I also have a Home Assistant as a backup for when the neural network fails. It provides a gentle nudge that perhaps the bins need taking out.
Shocking, I know.
show comments
qurren
> Home Assistant already knows which bins (trash cans, for the international readers) are due for collection. [...] What it can’t tell me is whether I’ve actually done anything about it.
My first gut reaction to this was: Have you tried using a vision-language model?
I used to have an elaborate Home Assistant setup with tons of zigbee sensors here and there telling me if front door is not closed properly, if the rice cooker is left on (via current monitoring), if the garage door is left open (with a magnetic contact sensor), if window is left open, if the soil is dry, etc. I had to keep changing batteries of all these things and manage flaky Zigbee connections in the past.
More recently, I've found the absolute most reliable, ~zero-maintainence solution to be an LLM. Home Assistant already has an LLM plugin, you can wire it up to a local LLM if you want, and you can just feed it an image and ask "Is this garage door open or closed" and I found that to be much better for my sanity than changing batteries every 6 months and dealing with flaky sensors sending me false alerts.
I even found it useful for "Did the cat vomit on the floor" (if so notify me and pause the cleaning robot schedule) and "Do any plants look like they need attention"
For TFA's use case, if you already have a security camera facing front, you can just feed it that image and ask "Are the garbage bins out"
Also, if you're worried about cameras sending home images to the cloud, there are non-cloud options. (I use Unifi cameras and Home Assistant gets a RTSP stream over LAN.)
show comments
walrus01
If the author is reading this, I did a ctrl-f in your writeup for "battery" and "life" and didn't find anything specific about how long the batteries on these last. Can you elaborate a bit more on the battery types in use and lifespan? Charging? Since use lithium? 18650? CR2032 or CR2477 size coin cells? Any other details would be very interesting to read.
My biggest barrier to doing something like thins in my home automation setups is that I already have an extensive schedule of things that need periodic disposable battery changes (like CR2477 cells in zigbee humidity/temperature sensors), and I'm really hesitant to add even more into my house. In addition to things like battery-backed smoke detectors in every bedroom, etc, which need periodic 9V replacements on a schedule.
In fact I've started looking at going the opposite direction, where I can run a cat5e cable from my utility room, down and across the ceiling of my basement, and come up in a location where I can install an 802.3af to arbitrary DC voltage DC-DC converter (3.3, 5VDC, 12VDC out, or similar) as a PoE splitter, to make some of these things not dependent on batteries. Same idea as how you can find 9V or AA battery to direct DC wiring "dummy battery" battery eliminators now. Obviously such an approach wouldn't work for garbage bins but that's how much I am trying to eliminate having dozens of batteries to replace, as they're already all over the place with window and door sensors.
This is one of the problems where I think a simple camera or two would suffice.
Years ago I worked as a consultant on a projects which is quite similar to this, but larger scale and involving moving machinery. The task was basically to come up with sensors that would register when and which machines would be at what place in a structure. In the end we came up with a solution which only involved cameras and object detection...this was before AlexNet etc., so fun times with classical machine learning.
rapjr9
If position were measured instead of range this could be used to document activities-of-daily-living, for the elderly or to measure health information for anyone (humans or pets). Precision location could be used to detect cooking, fridge visits, bathroom visits, exercise, time in bed, possibly time sleeping, overall activity (or lack thereof, possibly indicating a health decline if extended over a long time), and more. There has been research on trying to do this with accelerometers and RFID tags, possibly recent research has tried UWB tags.
aenis
In my case, one of the external cameras points at where the bins normally reside. Each has a distinct colour. I wrote a simple algo that bins colours on the image coming from the camera, and, accounting for things like direct sunlight skewing the distribution, tells me which bins are currently on the street. That was pre-LLM age, these days I would have told claude to write that thing for me. Works as a charm, no batteries nor maintenance of any type. Runs on pyscript.
bschwindHN
A related over-engineered project for residential trash management:
I bought a bunch of UWB boards 2 years ago for a POC with a transit company, that didn't materialise.
I like the tech, and would like to do something with the boards. We only have 1 bin, and as others point out, I could just ask HA if the bin is out on the street on Friday morning (already have cameras around the house).
I've thought of a garage door opener that senses when the car is home. I would still like to do something public though, just not sure what.
I have a transit info app at bus stops, so maybe a bus aware arrival board, but GPS solves that problem better, and the UWB devices would just get stolen anyways.
knotimpressed
How much of this was written with/edited by an LLM? Some parts of it feel somewhat Claude-ish, and I’m very curious if that’s a false positive.
Super cool work! I had no idea UWB tags had gotten so cheap, or that CR2077 was a size.
show comments
mikestew
That’s all well and good until the truck driver operating the mechanical arm drops your bin into the truck. Spoiler alert: the bin will not be fished out, it is now garbage.
So those tags are going to have to be a little cheaper before I’ll use them. :-)
show comments
scun
Transmitters on the bins seems cumbersome and expensive. I would use metal tape on the bin to complete a circuit on the fence.
Then I can overengineer the fensor circuit! Ultralow power, bin recognition, alarms, fox nerfs, auto-lids and so on.
show comments
fennecbutt
Hmmmm. Idt UWB needed for this. Just good old Bluetooth beacons/rssi. Oh and I think BTLE 5.6 or ish actually has UWB like positioning built in down to centimetres AFAIK. And they're SUPER low power.
fwip
A cheaper solution is some vibration sensors. Ignore short vibrations (putting garbage in). Several seconds of vibration on trash night mean you've put it out. Big vibration on trash day means they picked it up. And then one more long one when you roll it back in later that day.
georgemcbay
> Or the bin’s precise location - how many fixed reference points would I need to triangulate it on a 2D plane?
Back in 2019 I worked for a company (in San Diego) that added UWB-based positioning to the 7 train line in NYC.
The code I wrote ran on what was basically a combined Raspberry Pi 3 compute module connected to a UWB device installed into the train's bonnet which would ping UWB devices set along the track side to determine the position of the train.
In most situations I could get a unique position based on only two beacons, but I had the benefit of the trains being on one of a limited number of fixed tracks which severely limited where they could practically be in space... well, assuming there hadn't been a train derailment.
The system didn't work in a 2D plane, though, our surveyed track data and track side UWB positions were all 3D encoded as ECEF coordinates.
In some situations depending upon track layout you needed more than 2 beacons to determine the correct position if you were just sampling single points in isolation, but since you're working with a continuous stream of changed positions as the train ran along the track at predictable speeds, you could pretty easily figure out which of multiple possibilities it had to be based on what came before even if you had potential multi-positional results for the current sample.
Of course, we would use as many of the track side beacons as we could see to strengthen confidence in the positioning, and the beacons were placed so we could generally see a good 4-6 of them most of the time.
Snow was the biggest challenge for the system, in a heavy thick snow the UWB signals bounce all over the place and you get severe multipath effects that can make it pretty difficult to determine which is the true line of sight signal for your time of flight calculations.
Fun job, I had to get certified to be able to walk the NYC train tracks because we installed a lot of the track side beacons ourselves, and I spent a lot of late nights doing live testing runs on trains that were empty other than a couple of people from our company and some MTA workers assigned to our test trains.
jcomp
Enjoyed this, thank you for writing it up and posting
show comments
PunchyHamster
I'd just do it with camera and some rudimentary image recognition, having bin present/absent (in process of being taken care of) signal should've been enough
hk1337
Seems like a fun project but ultimately not worth it. Just setup a calendar alert to remind you to put the bins out and pull them back in, they're usually picked up on a schedule so it couldn't be too difficult to figure out when to go pull them back to the house.
I live in the EU, where I have five separate waste streams: glass, paper, plastic/metal, RVM plastic, and general trash. They all fill at different rates, while the collection schedule is essentially nondeterministic. The only reliable signal is that they arrive at 6 a.m. sharp in trucks specifically engineered to maximize acoustic output.
My solution is similarly overengineered: a vision system combined with an olfactory system feeds into a biological neural network, which monitors the bins and determines when intervention is required.
The neural network is annoyingly power-hungry, occasionally unreliable, and has yet to support OTA firmware updates. I also have a Home Assistant as a backup for when the neural network fails. It provides a gentle nudge that perhaps the bins need taking out.
Shocking, I know.
> Home Assistant already knows which bins (trash cans, for the international readers) are due for collection. [...] What it can’t tell me is whether I’ve actually done anything about it.
My first gut reaction to this was: Have you tried using a vision-language model?
I used to have an elaborate Home Assistant setup with tons of zigbee sensors here and there telling me if front door is not closed properly, if the rice cooker is left on (via current monitoring), if the garage door is left open (with a magnetic contact sensor), if window is left open, if the soil is dry, etc. I had to keep changing batteries of all these things and manage flaky Zigbee connections in the past.
More recently, I've found the absolute most reliable, ~zero-maintainence solution to be an LLM. Home Assistant already has an LLM plugin, you can wire it up to a local LLM if you want, and you can just feed it an image and ask "Is this garage door open or closed" and I found that to be much better for my sanity than changing batteries every 6 months and dealing with flaky sensors sending me false alerts.
I even found it useful for "Did the cat vomit on the floor" (if so notify me and pause the cleaning robot schedule) and "Do any plants look like they need attention"
For TFA's use case, if you already have a security camera facing front, you can just feed it that image and ask "Are the garbage bins out"
Also, if you're worried about cameras sending home images to the cloud, there are non-cloud options. (I use Unifi cameras and Home Assistant gets a RTSP stream over LAN.)
If the author is reading this, I did a ctrl-f in your writeup for "battery" and "life" and didn't find anything specific about how long the batteries on these last. Can you elaborate a bit more on the battery types in use and lifespan? Charging? Since use lithium? 18650? CR2032 or CR2477 size coin cells? Any other details would be very interesting to read.
My biggest barrier to doing something like thins in my home automation setups is that I already have an extensive schedule of things that need periodic disposable battery changes (like CR2477 cells in zigbee humidity/temperature sensors), and I'm really hesitant to add even more into my house. In addition to things like battery-backed smoke detectors in every bedroom, etc, which need periodic 9V replacements on a schedule.
In fact I've started looking at going the opposite direction, where I can run a cat5e cable from my utility room, down and across the ceiling of my basement, and come up in a location where I can install an 802.3af to arbitrary DC voltage DC-DC converter (3.3, 5VDC, 12VDC out, or similar) as a PoE splitter, to make some of these things not dependent on batteries. Same idea as how you can find 9V or AA battery to direct DC wiring "dummy battery" battery eliminators now. Obviously such an approach wouldn't work for garbage bins but that's how much I am trying to eliminate having dozens of batteries to replace, as they're already all over the place with window and door sensors.
This is great fun. For where I live I decided to "under engineer" a solution that required 0 Turing Machines: https://blog.jgc.org/2024/09/i-made-rubbish-clock.html
This is one of the problems where I think a simple camera or two would suffice.
Years ago I worked as a consultant on a projects which is quite similar to this, but larger scale and involving moving machinery. The task was basically to come up with sensors that would register when and which machines would be at what place in a structure. In the end we came up with a solution which only involved cameras and object detection...this was before AlexNet etc., so fun times with classical machine learning.
If position were measured instead of range this could be used to document activities-of-daily-living, for the elderly or to measure health information for anyone (humans or pets). Precision location could be used to detect cooking, fridge visits, bathroom visits, exercise, time in bed, possibly time sleeping, overall activity (or lack thereof, possibly indicating a health decline if extended over a long time), and more. There has been research on trying to do this with accelerometers and RFID tags, possibly recent research has tried UWB tags.
In my case, one of the external cameras points at where the bins normally reside. Each has a distinct colour. I wrote a simple algo that bins colours on the image coming from the camera, and, accounting for things like direct sunlight skewing the distribution, tells me which bins are currently on the street. That was pre-LLM age, these days I would have told claude to write that thing for me. Works as a charm, no batteries nor maintenance of any type. Runs on pyscript.
A related over-engineered project for residential trash management:
https://news.ycombinator.com/item?id=34216415
Yes, and don't call me Shirley!
I bought a bunch of UWB boards 2 years ago for a POC with a transit company, that didn't materialise.
I like the tech, and would like to do something with the boards. We only have 1 bin, and as others point out, I could just ask HA if the bin is out on the street on Friday morning (already have cameras around the house).
I've thought of a garage door opener that senses when the car is home. I would still like to do something public though, just not sure what.
I have a transit info app at bus stops, so maybe a bus aware arrival board, but GPS solves that problem better, and the UWB devices would just get stolen anyways.
How much of this was written with/edited by an LLM? Some parts of it feel somewhat Claude-ish, and I’m very curious if that’s a false positive.
Super cool work! I had no idea UWB tags had gotten so cheap, or that CR2077 was a size.
That’s all well and good until the truck driver operating the mechanical arm drops your bin into the truck. Spoiler alert: the bin will not be fished out, it is now garbage.
So those tags are going to have to be a little cheaper before I’ll use them. :-)
Transmitters on the bins seems cumbersome and expensive. I would use metal tape on the bin to complete a circuit on the fence.
Then I can overengineer the fensor circuit! Ultralow power, bin recognition, alarms, fox nerfs, auto-lids and so on.
Hmmmm. Idt UWB needed for this. Just good old Bluetooth beacons/rssi. Oh and I think BTLE 5.6 or ish actually has UWB like positioning built in down to centimetres AFAIK. And they're SUPER low power.
A cheaper solution is some vibration sensors. Ignore short vibrations (putting garbage in). Several seconds of vibration on trash night mean you've put it out. Big vibration on trash day means they picked it up. And then one more long one when you roll it back in later that day.
> Or the bin’s precise location - how many fixed reference points would I need to triangulate it on a 2D plane?
Back in 2019 I worked for a company (in San Diego) that added UWB-based positioning to the 7 train line in NYC.
The code I wrote ran on what was basically a combined Raspberry Pi 3 compute module connected to a UWB device installed into the train's bonnet which would ping UWB devices set along the track side to determine the position of the train.
In most situations I could get a unique position based on only two beacons, but I had the benefit of the trains being on one of a limited number of fixed tracks which severely limited where they could practically be in space... well, assuming there hadn't been a train derailment.
The system didn't work in a 2D plane, though, our surveyed track data and track side UWB positions were all 3D encoded as ECEF coordinates.
In some situations depending upon track layout you needed more than 2 beacons to determine the correct position if you were just sampling single points in isolation, but since you're working with a continuous stream of changed positions as the train ran along the track at predictable speeds, you could pretty easily figure out which of multiple possibilities it had to be based on what came before even if you had potential multi-positional results for the current sample.
Of course, we would use as many of the track side beacons as we could see to strengthen confidence in the positioning, and the beacons were placed so we could generally see a good 4-6 of them most of the time.
Snow was the biggest challenge for the system, in a heavy thick snow the UWB signals bounce all over the place and you get severe multipath effects that can make it pretty difficult to determine which is the true line of sight signal for your time of flight calculations.
Fun job, I had to get certified to be able to walk the NYC train tracks because we installed a lot of the track side beacons ourselves, and I spent a lot of late nights doing live testing runs on trains that were empty other than a couple of people from our company and some MTA workers assigned to our test trains.
Enjoyed this, thank you for writing it up and posting
I'd just do it with camera and some rudimentary image recognition, having bin present/absent (in process of being taken care of) signal should've been enough
Seems like a fun project but ultimately not worth it. Just setup a calendar alert to remind you to put the bins out and pull them back in, they're usually picked up on a schedule so it couldn't be too difficult to figure out when to go pull them back to the house.