I thought about this a long time. I felt that perimeter wires were great. But to make different mowing patterns would require more wires. And if the mowing area needed to change, the wires would need to be replaced. GPS didn't seem accurate enough (at the time) and could glitch too easily. The biggest drawback is the non-patterned mowing algorithm that wastes time and energy. I settled on the idea of using a scanning head that looks for reflectors placed around the yard for triangulation. This is the method used in many factories for robotic forklifts. But I don't have time to work on such projects because work keeps me busy, and family.
I have very similar conclusions, reflectors seems great compromise - Linde use them in their HDk series, and reports 1-2 cm accuracy. Don't know about behavior in full sun thou, optical methods tend to be jammed by intense sun and scanning head is difficult to use with standard remote control solution of detecting IR signals keyed in 40kHz carrier wave. Maybe 2 colors pointing same reflector? That way by having 2 peaks on diffrent frequency we'd eliminate false positive. Adding third detector that should stay low for third color would eliminate white objects or sun glare to give false positive. Alternatively, for smaller distances, ultrasound work ok for 12m - Cheap ultrasound sensors can be modifyed to talk to each other and double distance measure range, if they would be synchronised by PPS signal from chap gps reciver - could be as accurate as GPS-RTK, but order of magnitude cheaper
I actually working in development for commercial robot mowers like that. As a collision sensor you could use a magnet on the front bumper and a Hall-effect sensor to detect changes in the magnetic field. With that you have a non-mechanic detection solution which means it's less likely to fail. You can also use it to detect if someone lifts the mower.
Imponujące. Trafiłem na Pana stronę przez czysty przypadek. Jest Pan przedstawicielem coraz mniej licznej grupy osób na UA-cam, które nadal czerpią radość i satysfakcję z tworzenia i dzielenia się swoimi dokonaniami z innymi. Dziękuję! Co do tego konkretnego projektu: jedną z zasad „skutecznego koszenia” jest możliwe ścisłe przyleganie obudowy ostrza kosiarki do podłoża. Dzięki temu pod obudową powstaje minimalne podciśnienie dzięki obracającemu się ostrzu kosiarki, które przypomina śmigło. To podciśnienie sprzyja pionowemu ustawianiu się źdźbeł trawy. To z kolei ułatwia ostrzu równe ścinanie. Zastanawiam się, czy pomyślał Pan o adaptacji jednej z dostępnych w sklepach dyskontowych taniej kosiarki elektrycznej. Częściowo znalazłem odpowiedź w samym filmie: chodzi o projekt, który można będzie w pełni zrealizować samemu. Może jednak na etapie przygotowania prototypu miałoby to sens? Czekam na kolejne projekty open source / open hardware. Na przykład na pralkę. Serdecznie pozdrawiam.
Professional systems use a +/- single cycle chirp with a specific repetition or a waveform with a skewed duty cycle. This allows it to detect not only if it approaches the wire but from which direction. When the mower moves outside the boundary the signal is inverted and by using a chirp or skewed duty it can detect this. It is not possible to distinguish phase skew from inversion with a symmetrical signal. With proper amplification this also allows the system to always detect the signal anywhere inside the boundary so it can stop if power is lost to the boundary.
3:00 When you're trying to sense a signal of particular frequency, you terminate that inductor with a capacitor to create a tuned oscillator. Just use some random internet LC calculator to come up with the values. Safety feature suggestion: continuity detection for the perimeter wire. NRF24L01 is OK but keep in mind possibility of ESPnow, which is a similar thing but it communicates between ESP32/ESP8266 microcontrollers. I don't like shopping for NRF24L01 because there are a lot of fakes and their compatibility to each other including wireless protocol is very hit or miss, and the quality of the modules is as well. I haven't shopped for Nordic microcontrollers in a while but when i last checked they were on the expensive side as well, if you wanted to have radio and microcontroller on the same IC, which i find myself usually wanting that if possible. I wonder if there's some sort of base station triangulation system one can build cheaply. I would like everyone's suggestions, is sure to come in handy someday.
I've always liked the idea of a base station triangulation system. Not only would it be nice for this application but I thought it would be cool to use the same setup to figure out which room I was in in my house. Then I could make calls to turn on lights and fans relative to my current position. Plus, a triangulation system is (for me) another new thing to learn.
@@makernova8160 Have a look at UWB. It uses TOF like GPS but penetrates walls and has an accuracy of a few cm. There was an open source project on hackAday once, but the creator pulled it off and commercialized it. So you would have to start from square 1. Let me know when you succeed 😉 I want this too
Hoho Metabo, powodzi się! 😂 Dobry sprzęt, piękny garaż z OSB, widać że przemyślany. Kibicuje od dawna, i chciałem powiedzieć że bardzo szanuję za mentalność open-source.
Even if it is not for this project: For RTK you could have a look at the Adafruit ZED-F9P and RTK SSR. SSR is using correction data that can be broadcast instead of per device corrections.
Świetny projekt! Zrób wykrywacz metali na twojej wykaszarce. To samo wykryje siatki, słupki i metaliczne śmieci, który potrafią uszkodzić wykaszarkę. Z innej strony z detektorem metali wykaszarka sama potrafi omijać ogrodzenia i zmniejszy prace dla markowania miejsc, zabronionych dla niej. Jeszcze dodaj żyroskop, żeby wykaszrka jezdziła wzdłóż linii równoległych i nie omijała części pola na dużych działkach.
Seeing a video with such title appear, while I'm going to bed at 5am, after working on my thesis about RTK mowers is ... idk :D, my brain is shutting down, good night
I mean RTK GPS *is* objectively better. It's just more expensive and needs a lot more work to implement. If you have a large garden (1 acre+) then mowing randomly is super inefficient as a random path mower will be mowing some areas dozens of times while other areas might get mowed relatively infrequently (or not at all). Likewise if you have multiple smaller lawns and the mower needs to traverse areas to get to them ....with a boundary wire mower you basically need multiple mowers. Due to the relative simplicity a boundary wire mower is probably the best place to start for a DIY project though. But I'd aim to make it as modular as possible, so folks can add RTK, Lidar, vision, etc if their lawn requires it and budget allows for it.
I am really pleased to follow this project as I have been working on something similar using a commercial robot mower that I purchased on eBay. It had no charging base station, battery or working main PCB. But the BLDC motors work OK. Having no money for the expensive GPS RTK, the perimeter wire is my go-to. Can the plastic food containers withstand the UV and remain waterproof? Guess I'll find out the hard way. Have you considered using the JSN-SR20-Y1 ultrasonic detector module with car bumper-style ultrasonic weatherproof transceivers for potential collision detection? These are on my to-do list. Right now my main headache is efficient power management, as my motors are 18V and the electronics are 5V. Anyway, good luck with your progress!
Base corrections over NTRIP are used by surveyors around the world, so don't think it could only be used on your local country. It works the same on every country, you insert the address (IP) username & password, port and (optional) mount point (basically if they have multiple receivers you can choose one). That's it, you now have the corrections. This requires an internet correction tho. But it can be used on many other rovers, even drones.
There are no free services for it, are there? Like you pay for it and it only works for 50km next to a tower which are sparse in Europe. Or are there other tricks?
@@ssa7843 Depends on the country really. In W EU there are a lot of free CORS networks. If you get a paid service you have access to all base stations from that area/country. If you move you can choose the one closest to you. Most of them can automatically give you the corrections from the base closest to you based on your current location.
Well neither is a solution for the problems we have in the Netherlands. Ground and houses are so expensive that we simply can't buy them anyways :D I have a CS degree let me turn it off and on :D And then proves he does have a modern CS degree because he programs in Python BLEHHHHHH BLEHHHHH... You can use Zig even on a PiPico or C and C++. Real men don't use python on embedded devices.... ;) Oh man hilarious that when you are setting up the perimeter wire an old man in the back actually mowed his lawn in the same time :D And luckily there wasn't a mower yet attached, or we would've had an obituary on the screen: "Bartnik, taken to soon from us by the wrath of his mother". BRILLIANT VIDEO! KEEP IT UP!
Glad to see you're do this project. I also built an autonomous mower for my engineering final year project. It looks rubbish but I was suprised with how well the perimeter wire worked. My hardware is very similar to yours though. I amplified the signal by 200 using an op amp and used two inductors angled at 45 degrees to the center plane on each side of the mower. This made it very easy to program which way the robot turns when it detects the wire. Also filtering the output from the op amp with a large capcitor made it much more readable for the arduino. I have a video on my channel with the contraption in action.
What about using two sensors? One 45deg to the right of the front and another of the left. It could better detect corners and could tell which direction the wire is in. Keep up the great work!
For me personally, I'm not a fan of perimeter wire (only), because typically the robot doesn't know where it is, or what the shape of the area is that it is mowing. It can't plan a route or pattern to cover it efficiently or quickly, and it can't even be sure it got to it all. It's just statistically likely that it'll cover everything eventually (if the shape isn't too unsuitable), just by randomly bouncing off the edges. Of course this can be improved with other sensors (internal or external), but true absolute positioning would be in a different category. There are of course also ways to do absolute positioning just in a small area (multiple receivers triangulating it's position), but I do understand why that isn't and can't be the focus right now. It's still great to see the whole process you use to design, even if the product or project isn't specifically "for me". Also it being open source means everyone can add whatever positioning system, or it'll be an option V2. Either way, great project!
You could use a phase locked loop with 5 kHz or a mixer to detect the signal. It is much faster and saves computing time. It’s all analog, which can be hard. For circuit simulations I use Ltspice which saves you a lot of time. Greets from a hw dev.
Obstacle sensor - full enclosure mounted using extremely flexible tpu, somewhere place a magnet on the enclosure and hall effect sensor on the main board. If the hall effect senses "loss of magnetic field", itmwill stop. This will work when you bounce obstacle in front, but also during reversing, turning, etc.
My first thought I had when you said that it would be an issue to run the mower when you have an power outage is "just add a backup battery to the signal wire" and it's probably a good idea too have on the final version
I know the project is meant to be something accessible by most people, but i think for a robotic mower it kinda requires some more complex or expensive stuff just to make it viable. Wire loops and random movement would be fine for a small space, but one is going to need a more complex and expensive solution for anything more.
Wanted to say - don't use motor current sensors for bump detection. Worx robots use them and they tend to dig into ground and destroy their body. Use separate casing on top and hall sensor (or accelerometer) for bump detection. Second - to eliminate driving out of bounds in corners.. never ever do 90 degree angles in boundry wire. Smooth curves. All cable robots need this just to eliminate getting it stuck or going nuts in corners. Rounded corners. Third - make 2 coil circuit to detect wire. Then PID loop and try to keep wire between 2 detect coils. Better than one if signal is too high right under the sensor. 2 sensors on each side and when driving back using cable to recharge you try to keep wire between 2 sensors. So both sensor stay on limit.
Please make the wheel without the motor - in the BACK !! so the motorized wheels in the FRONT! and remember to get the main weight of the robot at the motorized wheel side!! :) its VERY important for a garden, not as perfect as yours! trust me :) Or better, chain drive - 4wd also possible :) Keep it up! Nice project and nice progress!
1:30 Husqvarna uses guide wires (at least 2 up to 4) as well, this way can mower faster return to charging station, otherwise must mower do at least 1/2 (on the long run) of the boundary wire path. On the other side, worx landroid (positec company) uses only boundary wire (for virtual fence and return path to charging station). Both those mower have lift sensor in front wheels (if you lift front part while mower is moving, then should mower stop instantly & stop the blades spining as well ). Wire, vs gps (+rtk station), vs lidar, the later is for all terrain.
Before getting my luba 5000, was working on the Rep_al mower, defenetly lots of feature creap before I got it working. :) I then decided to go for ardupilot (yuri's homebrew here on you tube has some vids about that) with rtk as wires being cut is horrible to find. the worst was when it doesnt cleanly cut the wire, and it corrodes in the ground. ardusimple is the rtk I got at $600 for a kit. Im still wanting to build one one day but think converting a mower that is exsisting would be a good start :) great project!
I do agree with the choice of the wire. It works and it's cheaper. Only a question: do you think the wire in order to be detected will have to be placed on the surface or can be put some cm deep? This in order to reduce the risk that animals (cats dogs or even mouses) can cut it.
I will have to experiment with that as well but with the right amplification on the sensor side and maybe higher voltage in the perimeter wire it should work even underground.
I spend way to much time watching my budget robot mower at home. Pointless info you probably know but Corners get churned up when it rains suddenly (thankfully unheard of in the uk 😂) due to ending up there a lot and getting in the dance mode. Mine seems to randomly choose if its turning clock wise or counter It seems to have 10 or 12 angles it will rotate to when it hits the boundary The radar object detection struggles with a washing line pole but not over hanging plants The pre cut boundary line travel is always too long, so it rarely starts in first 75% of perimeter Cutting perimeter doent happen enough and it requires manual cutting. Again sorry if this is all known and poitnless cr@p
Really interesting project, I like your journey through the problems. There was a rbpi based semi-open source lawnmover project, the guy developed everything for it including a phone app. He was selling the plans for 30eur or something. I was very inclined to build it but recently robot movers got really cheap, the one in lidl is around 300 eur with decent specs. So I gave up on building my own and have to stick to the 100 other projects I have in my head :)
AI image mapping is maybe a better option to allow a rover to move not randomly bouncing from the invisible fence… there’s already a commercially made robot that does… it can follow a learned path.
increase voltage, increase frequency. and use high frequency diode to convert ac to dc as sense and usedc output as sense. it will reduce computation of arduino!
The general purpose FFT may be a bit too much. A DFT for only a single frequency bin with precalculated values is super easy to implement. Be sure to sample at at least double the frequency you want to detect (10x is better for differentiation). The alternative is an analog bandpass filter or inverse notch.
Honestly, a wire is fine to keep a mower inside the perimeter, but as soon as you have a bigger lawn, having the mower drive in pseudo-random paths is just ... well wasting lots of time and energy. I currently have such a mower and the next one will definitely have to be more intelligent and know where exactly it is and what part of the lawn it still needs to do. These random things were nice when nothing better was available, but now there are commercial mowers that know where they are without parts (like cameras) that would need lots of maintenance (I guess via GPS), so I'm certain my next one will be like that, even if it costs a bit more.
another suggestion: you could leave the sensitivity as it is rn and bury the cable some cm deep in the soil, in order to cover potential hazard for person/animals!
I will have a microcontroller in the base anyway and that gives me more flexibility to adjust the frequency or even use more complex signal. Thank you!
RTK is expensive but you can actually make money off it. I also suspect that any commercial robo-mower that comes with RTK probably does sell your RTK data already. But you won't see a dime of that. So setting up your own base station is actually beneficial.
Add a second perimeter to make your robot halt till reset. And LoRa “keep running” signal on the same board. It’ll not escape then. I’m pretty sure you eventually loose the robot without such system.
I've had my neighbors cut my perimeter wire so many times, or the sand pushes the wire up, or the mower itself cuts the wire and commits xuicide every session. It's a massive nightmare. Plus the ping-pong method of attack is super frustrating. Some parts of the yard won't get mowed for months
I personally think neither solution is great. GPS RTK is too expensive and not always that reliable depending on your location (tall buildings are a huge problem). And the perimeter wire is messy and a pain to install. I think you should consider solutions such as (bluetooth) beacons and/or lidar as well. Perhaps a combination of several approaches by also using a compas and accelerometer to interpolate your location. Perimeter wire solves your border issue, but not your location.
Hello, great project. What if you install magnets in your tyres? And then you can measure the rotational speed of your wheels with a Hall sensor on the housing.
if you want to make it follow the perimeter wire, what about a second perimeter wire outside the first at a different frequency and an additional sensor on the mower so it can orient itself?
Perimeter wire + decent logic is the only hassle free way. That is if you just want your lawn to look neat. If the objective is to tinker and mess with it all the time, then go wireless.
Have you thought to add a sort of ultrasound emitter in order to keep dogs/cats/animals away from this unattended blade spinner? Could be helpful and save lives!
RTK is the future. Why bother developing 1st gen tech when the entire industry is going RTK? I got my Landroid for 800CAD, not much more than a regular mower. People whos watching this isn't in the market of building a mower since they're already low cost. Better content would be to convert a perimeter wire mower to RTK or to DIY RTK mower in my opinion. Also perimeter wires are a real PITA.
O ile dobrze zrozumiałem to twoje wykrywanie perymetru działa na zasadzie wykrywania przewodu? Natomiast w komercyjnych kosiarkach nie chodzi tyle o wykrycie przewodu co cały czas prze asymetrię sygnału kosiarka dostaje informację czy jest wewnątrz pętli czy na zewnątrz, jeżeli czujnik dostanie odwrócony sygnał oznacza to że jest poza obszarem.
Takie klasyczne podejście jak w kosiarkach komercyjnych rozwiązuje też twój problem z wykryciem zaniku zasilania i zniknięciu obszaru, kosiarka cały czas na całej powierzchni dostaje informację że znajduje się wewnątrz obszaru, jeżeli zaniknie zasilanie przewodu sygnał znika kosiarka staje.
Encoders on the motors won't help you detect the wheels slipping over the grass. The wheel will still turn, just not move the mower forward on that side. Really, knowing exactly your position and (from subsequent position updates) heading is the way to go. Anything else is by now old tech.
You’re trying to reinvent too many wheels. Study the decisions that were made before you by projects like autonomous vacuum cleaners (and lawnmowers obv.). For instance there’s a reason why they didn’t simply rely on current sensors or motor-side encoders. Use the caster wheel to monitor actual motion. Also increase its diameter, it’s way too small. Base-station-less RTK is the way to go.
I like his learn-as-you-go method better. It's better from an intellectual standpoint, and it makes for more interesting videos. Of course in a business setting, building on "prior art" (i.e., the work of *other* people) is the only approach the boss will permit, but Niko is his own boss. He has no profit expectations, deadlines, or marketing considerations. He's living (and sharing) in the open-source universe, not the commercial one.
Hi, Im also working on a Mower, but i settelt on a sistem wich registers the distance betwen the Robot and The Antenna, the "Torches" Are Placed on the Eges of the field, im currently not sure, how these torches will work, but im thinking about it, may someone can tell me, how this could work.
Ok. Aren’t you just attempting to recreate a roomba? Just swap the vacuum and sweeper for a whipper snipper? Or can’t you just set boundary markers at the corners to create a “wall” or a single point of reference (like a charging base station) instead of wire or gps. It seems like you’re making the area factor a bigger problem than it should be. If it has a point of reference and can detect range or has two points of reference so I can triangulated position. It should be simple. It doesn’t require freaking GPS or physical wires. Edit: maybe I misunderstood what you were trying to accomplish
@@user-jp1qt8ut3s I think his parents help him in that regard, also he seems to be a talented young engineer, who I'm sure figured out how to monetize his skills
The first 500 people to use my link skl.sh/nikodembartnik07241 will get a 1 month free trial of Skillshare premium!
"When your new project kills the flowers, your mom kills you" Thats gold man
I thought about this a long time. I felt that perimeter wires were great.
But to make different mowing patterns would require more wires.
And if the mowing area needed to change, the wires would need to be replaced.
GPS didn't seem accurate enough (at the time) and could glitch too easily.
The biggest drawback is the non-patterned mowing algorithm that wastes time and energy.
I settled on the idea of using a scanning head that looks for reflectors placed around the yard for triangulation. This is the method used in many factories for robotic forklifts.
But I don't have time to work on such projects because work keeps me busy, and family.
I have very similar conclusions, reflectors seems great compromise - Linde use them in their HDk series, and reports 1-2 cm accuracy. Don't know about behavior in full sun thou, optical methods tend to be jammed by intense sun and scanning head is difficult to use with standard remote control solution of detecting IR signals keyed in 40kHz carrier wave. Maybe 2 colors pointing same reflector? That way by having 2 peaks on diffrent frequency we'd eliminate false positive. Adding third detector that should stay low for third color would eliminate white objects or sun glare to give false positive.
Alternatively, for smaller distances, ultrasound work ok for 12m - Cheap ultrasound sensors can be modifyed to talk to each other and double distance measure range, if they would be synchronised by PPS signal from chap gps reciver - could be as accurate as GPS-RTK, but order of magnitude cheaper
I actually working in development for commercial robot mowers like that.
As a collision sensor you could use a magnet on the front bumper and a Hall-effect sensor to detect changes in the magnetic field.
With that you have a non-mechanic detection solution which means it's less likely to fail.
You can also use it to detect if someone lifts the mower.
I can see you are still a beginner engineer your desk is still way to clean.
Haha, that hit hard! Sitting here at my desk at work that's piled high in electronics, 3D prints, and injection mold test shots
Well
Obvi he looks
Likes he’s 16 can’t be to old !
@@evanbarnes9984the way it should be. I had youngins harass me about the "clutter". This coming from kids without a drop of imagination.
I need a bigger desk. Help.
This is truth 😂
Imponujące. Trafiłem na Pana stronę przez czysty przypadek. Jest Pan przedstawicielem coraz mniej licznej grupy osób na UA-cam, które nadal czerpią radość i satysfakcję z tworzenia i dzielenia się swoimi dokonaniami z innymi. Dziękuję!
Co do tego konkretnego projektu: jedną z zasad „skutecznego koszenia” jest możliwe ścisłe przyleganie obudowy ostrza kosiarki do podłoża. Dzięki temu pod obudową powstaje minimalne podciśnienie dzięki obracającemu się ostrzu kosiarki, które przypomina śmigło. To podciśnienie sprzyja pionowemu ustawianiu się źdźbeł trawy. To z kolei ułatwia ostrzu równe ścinanie.
Zastanawiam się, czy pomyślał Pan o adaptacji jednej z dostępnych w sklepach dyskontowych taniej kosiarki elektrycznej. Częściowo znalazłem odpowiedź w samym filmie: chodzi o projekt, który można będzie w pełni zrealizować samemu. Może jednak na etapie przygotowania prototypu miałoby to sens?
Czekam na kolejne projekty open source / open hardware. Na przykład na pralkę.
Serdecznie pozdrawiam.
Professional systems use a +/- single cycle chirp with a specific repetition or a waveform with a skewed duty cycle. This allows it to detect not only if it approaches the wire but from which direction. When the mower moves outside the boundary the signal is inverted and by using a chirp or skewed duty it can detect this. It is not possible to distinguish phase skew from inversion with a symmetrical signal. With proper amplification this also allows the system to always detect the signal anywhere inside the boundary so it can stop if power is lost to the boundary.
That's interesting. Is there a specific name to that to learn more about it?
@@5eurosenelsuelo There was a patent I read ages ago. I will try and find it. No luck so far though.
@@5eurosenelsuelo I was unable to post the patent it seems. Probably blocks links.
The patent number is US8532822B2. You find it on google patents.
@@5eurosenelsuelo My previous reply was deleted by youtube.
The patent number is US8532822B2. You find it on google patents.
The patent number is US8532822B2. You find it on google patents
All my posts about this seem to disappear? Been trying for 5 days!
3:00 When you're trying to sense a signal of particular frequency, you terminate that inductor with a capacitor to create a tuned oscillator. Just use some random internet LC calculator to come up with the values.
Safety feature suggestion: continuity detection for the perimeter wire.
NRF24L01 is OK but keep in mind possibility of ESPnow, which is a similar thing but it communicates between ESP32/ESP8266 microcontrollers. I don't like shopping for NRF24L01 because there are a lot of fakes and their compatibility to each other including wireless protocol is very hit or miss, and the quality of the modules is as well. I haven't shopped for Nordic microcontrollers in a while but when i last checked they were on the expensive side as well, if you wanted to have radio and microcontroller on the same IC, which i find myself usually wanting that if possible.
I wonder if there's some sort of base station triangulation system one can build cheaply. I would like everyone's suggestions, is sure to come in handy someday.
I've always liked the idea of a base station triangulation system. Not only would it be nice for this application but I thought it would be cool to use the same setup to figure out which room I was in in my house. Then I could make calls to turn on lights and fans relative to my current position. Plus, a triangulation system is (for me) another new thing to learn.
@@makernova8160 Have a look at UWB. It uses TOF like GPS but penetrates walls and has an accuracy of a few cm. There was an open source project on hackAday once, but the creator pulled it off and commercialized it. So you would have to start from square 1. Let me know when you succeed 😉 I want this too
A triangulation system could also be used when the robot detects its battery is running low and calculates the fastest route to the charger.
Hoho Metabo, powodzi się! 😂 Dobry sprzęt, piękny garaż z OSB, widać że przemyślany. Kibicuje od dawna, i chciałem powiedzieć że bardzo szanuję za mentalność open-source.
Dzięki!
Even if it is not for this project: For RTK you could have a look at the Adafruit ZED-F9P and RTK SSR. SSR is using correction data that can be broadcast instead of per device corrections.
Świetny projekt! Zrób wykrywacz metali na twojej wykaszarce. To samo wykryje siatki, słupki i metaliczne śmieci, który potrafią uszkodzić wykaszarkę. Z innej strony z detektorem metali wykaszarka sama potrafi omijać ogrodzenia i zmniejszy prace dla markowania miejsc, zabronionych dla niej.
Jeszcze dodaj żyroskop, żeby wykaszrka jezdziła wzdłóż linii równoległych i nie omijała części pola na dużych działkach.
Seeing a video with such title appear, while I'm going to bed at 5am, after working on my thesis about RTK mowers is ... idk :D, my brain is shutting down, good night
I mean RTK GPS *is* objectively better. It's just more expensive and needs a lot more work to implement.
If you have a large garden (1 acre+) then mowing randomly is super inefficient as a random path mower will be mowing some areas dozens of times while other areas might get mowed relatively infrequently (or not at all). Likewise if you have multiple smaller lawns and the mower needs to traverse areas to get to them ....with a boundary wire mower you basically need multiple mowers.
Due to the relative simplicity a boundary wire mower is probably the best place to start for a DIY project though. But I'd aim to make it as modular as possible, so folks can add RTK, Lidar, vision, etc if their lawn requires it and budget allows for it.
I am really pleased to follow this project as I have been working on something similar using a commercial robot mower that I purchased on eBay. It had no charging base station, battery or working main PCB. But the BLDC motors work OK. Having no money for the expensive GPS RTK, the perimeter wire is my go-to. Can the plastic food containers withstand the UV and remain waterproof? Guess I'll find out the hard way. Have you considered using the JSN-SR20-Y1 ultrasonic detector module with car bumper-style ultrasonic weatherproof transceivers for potential collision detection? These are on my to-do list. Right now my main headache is efficient power management, as my motors are 18V and the electronics are 5V. Anyway, good luck with your progress!
Base corrections over NTRIP are used by surveyors around the world, so don't think it could only be used on your local country. It works the same on every country, you insert the address (IP) username & password, port and (optional) mount point (basically if they have multiple receivers you can choose one). That's it, you now have the corrections. This requires an internet correction tho. But it can be used on many other rovers, even drones.
There are no free services for it, are there? Like you pay for it and it only works for 50km next to a tower which are sparse in Europe. Or are there other tricks?
@@ssa7843 Depends on the country really. In W EU there are a lot of free CORS networks. If you get a paid service you have access to all base stations from that area/country. If you move you can choose the one closest to you. Most of them can automatically give you the corrections from the base closest to you based on your current location.
Well neither is a solution for the problems we have in the Netherlands. Ground and houses are so expensive that we simply can't buy them anyways :D
I have a CS degree let me turn it off and on :D And then proves he does have a modern CS degree because he programs in Python BLEHHHHHH BLEHHHHH... You can use Zig even on a PiPico or C and C++. Real men don't use python on embedded devices.... ;) Oh man hilarious that when you are setting up the perimeter wire an old man in the back actually mowed his lawn in the same time :D
And luckily there wasn't a mower yet attached, or we would've had an obituary on the screen: "Bartnik, taken to soon from us by the wrath of his mother".
BRILLIANT VIDEO! KEEP IT UP!
Glad to see you're do this project. I also built an autonomous mower for my engineering final year project. It looks rubbish but I was suprised with how well the perimeter wire worked. My hardware is very similar to yours though. I amplified the signal by 200 using an op amp and used two inductors angled at 45 degrees to the center plane on each side of the mower. This made it very easy to program which way the robot turns when it detects the wire. Also filtering the output from the op amp with a large capcitor made it much more readable for the arduino. I have a video on my channel with the contraption in action.
What about using two sensors? One 45deg to the right of the front and another of the left. It could better detect corners and could tell which direction the wire is in. Keep up the great work!
For me personally, I'm not a fan of perimeter wire (only), because typically the robot doesn't know where it is, or what the shape of the area is that it is mowing. It can't plan a route or pattern to cover it efficiently or quickly, and it can't even be sure it got to it all. It's just statistically likely that it'll cover everything eventually (if the shape isn't too unsuitable), just by randomly bouncing off the edges. Of course this can be improved with other sensors (internal or external), but true absolute positioning would be in a different category. There are of course also ways to do absolute positioning just in a small area (multiple receivers triangulating it's position), but I do understand why that isn't and can't be the focus right now.
It's still great to see the whole process you use to design, even if the product or project isn't specifically "for me". Also it being open source means everyone can add whatever positioning system, or it'll be an option V2. Either way, great project!
You could use a phase locked loop with 5 kHz or a mixer to detect the signal. It is much faster and saves computing time. It’s all analog, which can be hard. For circuit simulations I use Ltspice which saves you a lot of time.
Greets from a hw dev.
Obstacle sensor - full enclosure mounted using extremely flexible tpu, somewhere place a magnet on the enclosure and hall effect sensor on the main board. If the hall effect senses "loss of magnetic field", itmwill stop. This will work when you bounce obstacle in front, but also during reversing, turning, etc.
My first thought I had when you said that it would be an issue to run the mower when you have an power outage is "just add a backup battery to the signal wire" and it's probably a good idea too have on the final version
i picked up two italian mowing robots for $100. they have encoded motors. hopefully i can run gps rtk open source project on them some day.
I know the project is meant to be something accessible by most people, but i think for a robotic mower it kinda requires some more complex or expensive stuff just to make it viable. Wire loops and random movement would be fine for a small space, but one is going to need a more complex and expensive solution for anything more.
What a nice project and progress! Thanks Nikodem!!
RTK is the way. Yes it's expensive but accurate. Buying a robot vacuum cleaner and modifying it to cut grass may work though
Wanted to say - don't use motor current sensors for bump detection. Worx robots use them and they tend to dig into ground and destroy their body. Use separate casing on top and hall sensor (or accelerometer) for bump detection.
Second - to eliminate driving out of bounds in corners.. never ever do 90 degree angles in boundry wire. Smooth curves. All cable robots need this just to eliminate getting it stuck or going nuts in corners. Rounded corners.
Third - make 2 coil circuit to detect wire. Then PID loop and try to keep wire between 2 detect coils. Better than one if signal is too high right under the sensor. 2 sensors on each side and when driving back using cable to recharge you try to keep wire between 2 sensors. So both sensor stay on limit.
Please make the wheel without the motor - in the BACK !! so the motorized wheels in the FRONT! and remember to get the main weight of the robot at the motorized wheel side!! :) its VERY important for a garden, not as perfect as yours! trust me :) Or better, chain drive - 4wd also possible :)
Keep it up! Nice project and nice progress!
1:30 Husqvarna uses guide wires (at least 2 up to 4) as well, this way can mower faster return to charging station, otherwise must mower do at least 1/2 (on the long run) of the boundary wire path.
On the other side, worx landroid (positec company) uses only boundary wire (for virtual fence and return path to charging station).
Both those mower have lift sensor in front wheels (if you lift front part while mower is moving, then should mower stop instantly & stop the blades spining as well ).
Wire, vs gps (+rtk station), vs lidar, the later is for all terrain.
I just found this channel, very nice video!
16:45 of the wheels slip on grass, it will still not go in a straight line. I don't know if it really needs to. Great progress!
All robots mowers I've seen have an IMU to account for the slippage (there's going to be a lot of it!) and drive straight
Before getting my luba 5000, was working on the Rep_al mower, defenetly lots of feature creap before I got it working. :) I then decided to go for ardupilot (yuri's homebrew here on you tube has some vids about that) with rtk as wires being cut is horrible to find. the worst was when it doesnt cleanly cut the wire, and it corrodes in the ground. ardusimple is the rtk I got at $600 for a kit. Im still wanting to build one one day but think converting a mower that is exsisting would be a good start :) great project!
I do agree with the choice of the wire. It works and it's cheaper.
Only a question: do you think the wire in order to be detected will have to be placed on the surface or can be put some cm deep? This in order to reduce the risk that animals (cats dogs or even mouses) can cut it.
I will have to experiment with that as well but with the right amplification on the sensor side and maybe higher voltage in the perimeter wire it should work even underground.
I spend way to much time watching my budget robot mower at home.
Pointless info you probably know but
Corners get churned up when it rains suddenly (thankfully unheard of in the uk 😂) due to ending up there a lot and getting in the dance mode.
Mine seems to randomly choose if its turning clock wise or counter
It seems to have 10 or 12 angles it will rotate to when it hits the boundary
The radar object detection struggles with a washing line pole but not over hanging plants
The pre cut boundary line travel is always too long, so it rarely starts in first 75% of perimeter
Cutting perimeter doent happen enough and it requires manual cutting.
Again sorry if this is all known and poitnless cr@p
Really interesting project, I like your journey through the problems. There was a rbpi based semi-open source lawnmover project, the guy developed everything for it including a phone app. He was selling the plans for 30eur or something. I was very inclined to build it but recently robot movers got really cheap, the one in lidl is around 300 eur with decent specs. So I gave up on building my own and have to stick to the 100 other projects I have in my head :)
If LIDL in Europe has such a machine, maybe it will make its way to the USA in the future. I will be on the lookout for it.
Can't wait for the next episode already
Thanks!
AI image mapping is maybe a better option to allow a rover to move not randomly bouncing from the invisible fence… there’s already a commercially made robot that does… it can follow a learned path.
increase voltage, increase frequency. and use high frequency diode to convert ac to dc as sense and usedc output as sense. it will reduce computation of arduino!
An Automower knows if it is Inside or outside of the loop. I believe it is done by putting asymmetric pulses on the loop wire...
i was so confused i read this as "Paramotor wire is better than GPS RTK."
The general purpose FFT may be a bit too much. A DFT for only a single frequency bin with precalculated values is super easy to implement. Be sure to sample at at least double the frequency you want to detect (10x is better for differentiation).
The alternative is an analog bandpass filter or inverse notch.
Honestly, a wire is fine to keep a mower inside the perimeter, but as soon as you have a bigger lawn, having the mower drive in pseudo-random paths is just ... well wasting lots of time and energy. I currently have such a mower and the next one will definitely have to be more intelligent and know where exactly it is and what part of the lawn it still needs to do. These random things were nice when nothing better was available, but now there are commercial mowers that know where they are without parts (like cameras) that would need lots of maintenance (I guess via GPS), so I'm certain my next one will be like that, even if it costs a bit more.
another suggestion: you could leave the sensitivity as it is rn and bury the cable some cm deep in the soil, in order to cover potential hazard for person/animals!
Why not use a simple 5kHz generator with an NE555? Way cheaper/simpler? Love the project though..
I will have a microcontroller in the base anyway and that gives me more flexibility to adjust the frequency or even use more complex signal. Thank you!
Glad you're protecting your lif.... your mom's plants😁
RTK is expensive but you can actually make money off it.
I also suspect that any commercial robo-mower that comes with RTK probably does sell your RTK data already.
But you won't see a dime of that.
So setting up your own base station is actually beneficial.
Check Segway navi mow.i just got mine couple of weeks ago
Great project 🎉 I do cheer for you and your career!
I wonder how such wire "antenna" propagates and if it's causing radio interference... do you have some spectrum analyzer (TinySA maybe) to see ?
Add a second perimeter to make your robot halt till reset. And LoRa “keep running” signal on the same board. It’ll not escape then.
I’m pretty sure you eventually loose the robot without such system.
I've had my neighbors cut my perimeter wire so many times, or the sand pushes the wire up, or the mower itself cuts the wire and commits xuicide every session. It's a massive nightmare. Plus the ping-pong method of attack is super frustrating. Some parts of the yard won't get mowed for months
Sad to see that GPS RTK is dropped.
Could we make a simple GPS system using ultrasound?
Just to make the stakes higher and more fun?
I personally think neither solution is great. GPS RTK is too expensive and not always that reliable depending on your location (tall buildings are a huge problem). And the perimeter wire is messy and a pain to install. I think you should consider solutions such as (bluetooth) beacons and/or lidar as well. Perhaps a combination of several approaches by also using a compas and accelerometer to interpolate your location.
Perimeter wire solves your border issue, but not your location.
Hello, great project.
What if you install magnets in your tyres?
And then you can measure the rotational speed of your wheels with a Hall sensor on the housing.
if you want to make it follow the perimeter wire, what about a second perimeter wire outside the first at a different frequency and an additional sensor on the mower so it can orient itself?
Great project! please, use ESP32 or ESP8266 and OTA. this way you can update code live or even use micro python
skoszenie kwiatów mamy to poważne przestępstwo, racja , lepiej nie ryzykować
Amazing video, once again.
Would it still work if the wire is underground? It'd great to avoid tripping with it.
you should do a dockstation
Young Stark. How would you build a exoskeleton with giro to lift heavy things and carry it on uneven surface?
Perimeter wire + decent logic is the only hassle free way. That is if you just want your lawn to look neat. If the objective is to tinker and mess with it all the time, then go wireless.
respect bratku
Awesome 😎
i wonder if placing a strobe light at the 4 corners would work. then use a 360 deg light sensor to detect the strobe lights.
How many gardens are a simple rectangle?
An optical encoder can be used to control the speed, which is inexpensive
What about using BLE beacons for tracking?
something like iBeacon coupled with ESPresence?
Good work. But maybe not MVP. It *is* POC: Proof of Concept.
Awesome!
UWB modules are pretty cheap, and they have alright accuracy ( sub 10cm).
Then, is a light fence even better than boundary wire?
Why don't you use overboard motors and map the place with your lidar sensor?
Lidar won't work outside.
Have you thought to add a sort of ultrasound emitter in order to keep dogs/cats/animals away from this unattended blade spinner? Could be helpful and save lives!
There is $40 PX1125R RTK capable receivers.
I don’t want to lay over half a kilometer of boundary wire… for just the outer perimeter, not counting the internal boundaries.
Brother give me solution when I jogging my machine there are showing grbl ($ for help) and stop it.. Please I have no idea
RTK is the future. Why bother developing 1st gen tech when the entire industry is going RTK? I got my Landroid for 800CAD, not much more than a regular mower. People whos watching this isn't in the market of building a mower since they're already low cost. Better content would be to convert a perimeter wire mower to RTK or to DIY RTK mower in my opinion. Also perimeter wires are a real PITA.
O ile dobrze zrozumiałem to twoje wykrywanie perymetru działa na zasadzie wykrywania przewodu? Natomiast w komercyjnych kosiarkach nie chodzi tyle o wykrycie przewodu co cały czas prze asymetrię sygnału kosiarka dostaje informację czy jest wewnątrz pętli czy na zewnątrz, jeżeli czujnik dostanie odwrócony sygnał oznacza to że jest poza obszarem.
Takie klasyczne podejście jak w kosiarkach komercyjnych rozwiązuje też twój problem z wykryciem zaniku zasilania i zniknięciu obszaru, kosiarka cały czas na całej powierzchni dostaje informację że znajduje się wewnątrz obszaru, jeżeli zaniknie zasilanie przewodu sygnał znika kosiarka staje.
Subbed!
Encoders on the motors won't help you detect the wheels slipping over the grass. The wheel will still turn, just not move the mower forward on that side. Really, knowing exactly your position and (from subsequent position updates) heading is the way to go. Anything else is by now old tech.
Uwb stations to triangulate position and map perimeter?
You’re trying to reinvent too many wheels. Study the decisions that were made before you by projects like autonomous vacuum cleaners (and lawnmowers obv.). For instance there’s a reason why they didn’t simply rely on current sensors or motor-side encoders. Use the caster wheel to monitor actual motion. Also increase its diameter, it’s way too small. Base-station-less RTK is the way to go.
I like his learn-as-you-go method better. It's better from an intellectual standpoint, and it makes for more interesting videos. Of course in a business setting, building on "prior art" (i.e., the work of *other* people) is the only approach the boss will permit, but Niko is his own boss. He has no profit expectations, deadlines, or marketing considerations. He's living (and sharing) in the open-source universe, not the commercial one.
Hi, Im also working on a Mower, but i settelt on a sistem wich registers the distance betwen the Robot and The Antenna, the "Torches" Are Placed on the Eges of the field, im currently not sure, how these torches will work, but im thinking about it, may someone can tell me, how this could work.
you had me at schizophrenic hallucinations.
Thing with these is they only need to drive onto the road or footpath once...
Thank you Uncle Sam for free GPS
Instead of gps you may want to try UWB tracking
"I have a degree in computer science so let me turn it off and on."
🙃
Why not try something more innovative like lidar or something, seems like u have experience with it. Now it feels like ur just reinventing the wheel
Ok. Aren’t you just attempting to recreate a roomba? Just swap the vacuum and sweeper for a whipper snipper?
Or can’t you just set boundary markers at the corners to create a “wall” or a single point of reference (like a charging base station) instead of wire or gps.
It seems like you’re making the area factor a bigger problem than it should be. If it has a point of reference and can detect range or has two points of reference so I can triangulated position. It should be simple. It doesn’t require freaking GPS or physical wires.
Edit: maybe I misunderstood what you were trying to accomplish
learn to weld and use metal and bigger wheels ^^
Why do you need to design a land mower? You have already swiped all lawnmowers from western Europe, haven't you?
whats ur first language
At 5:26 he says "In Poland we have". So I guess there's your answer.
@@kxtof thx
How do you earn enough to have such a huge workshop?
yeah good question
@@bananabuilder2248 wanna be inspired too?
this is not "a huge" workshop from my point of view
@@muratkabilov whatever. How did he get those millions to pay for it?
@@user-jp1qt8ut3s I think his parents help him in that regard, also he seems to be a talented young engineer, who I'm sure figured out how to monetize his skills
perimeter
[pə'rɪmɪtə]
an awesome project. as always!
Thank you!