What is the most cost effective yet still reliable outdoor boundary system?

Question

I am working on a robot that will operate outside in people's lawns in a predefined space. The issue that I am having right now is figuring out what kind of boundary system I should implement in order to ensure that the robot stays within someone's lawn, assuming that each lawn the robot is being used in does not necessarily have a fence or the same layout but that the operating area is a rectangle or other 4 sided polygon.

I have come up with a few ideas so far. I am trying to keep the cost of the boundary system within 150 - 200 dollars USD. In addition, the idea is that setting up this boundary system should be as easy as possible for the end user. I hope that this system is easy to implement, but I am open to learning new skills to make it happen.

If one of these ideas is not feasible, please let me know. I am working on this project for school (high school, 12th standard). I have some experience with robotics but am a little bit out of my league for some of these boundary solutions.

1. Guide Wires Buried Underground (expensive, but reliable (from what I've heard), also a lot of work for the boundary system).
2. 4 Poles at each corner of the rectangular operating area. The robot will either have a camera, or some kind of ultrasonic sensor or LiDAR sensor to detect distance from the robot to each pole. Then, it can triangulate its position in the field and move accordingly. (I am guessing this will require some kind of image recognition to first detect the poles, and then figure out the distances to the pole).
3. Have the user control the robot manually once from one specific starting location around the entire boundary of the field. Measure the distances and whenever the user wants to use the robot, they can put it at the predefined starting location and it will cover the boundary path as well as any parts of the lawn inside. (This is the dumbest solution, and I am worried that the robot will slowly veer off course as it traverses the lawn up and down multiple times).
4. Some kind of AR app that a user can use, in which the app generates an environment surrounding the lawn and the user places 4 virtual markers at each corner of the lawn. This environment is uploaded to the software running on the robot, and it uses a camera (maybe spinning?) that allows it to figure out where in the environment it is. Then, it moves accordingly. (Honestly, I have no clue how this would work in the very slightest. Is it even feasible?)
5. Four poles at each corner of the operating area with infrared sensors creating some kind of invisible fence. As the robot passes through the invisible fence, and infrared receiver will pick up the signals and tell the robot to turn around. (I've heard infrared works only up to 30 feet, but this is for lawns, so I'm not sure if this would work, also how would the infrared sensors be powered (that's something I will need to check for myself)).
6. Red or blue rope about 6 inches off the ground with spikes holding it there. A sensor on the robot will detect this rope, and that forms the boundary. I like this idea a little bit but let's be honest, it would look kind of ugly in someone's front (or back) yard.
7. GPS (isn't nearly accurate or detailed enough)

As you can see, there's many potential solutions, but I'm trying to figure out which one will work 100% and is still cost effective. Thank you in advance! Also, if you have any ideas, please do let me know!

Answer 1

Having worked on robotic lawn mowers in the past. I can tell you from first hand experience that this is no easy task. Outdoor robots are hard. Much harder than indoor robots. Especially if you want them to operate over a long period of time (days). Some reasons for the difficulty:

• Everything gets dirty. So cameras and wheels will get covered in dirt, leaves, grass, bugs, etc.
• Extensive wheel slip. Your wheels will slip A LOT.
• Uneven terrain. No matter how flat the yard looks to you, it is actually very uneven. Not to mention if you go over a small rock or tree root, that can alter your heading in a drastic and unpredictable way.
• Harsh lighting. Contending with full sun (and full shade) can be quite difficult.
• Scale. Typically outdoor robots deal with much greater distances than indoor robots.
• Things change. The look of a place can change dramatically over the course of a day, let alone a season due to the movement of the sun, falling leaves, growing bushes, etc.

Here are some notes on your proposed solutions:

1. Buried Guide Wires. This is your best bet for keeping the robot contained. It won't localize the robot, but is the quickest and cheapest way to prevent the robot from running off. There is a reason every robotic lawn mower on the market uses this technology. The sensors on the robot are cheap. The expense comes from the spool of copper wire you need. I'd actually recommend that you just buy a robotic lawn mower and hack it to your needs. It will save you a lot of work.

2. LiDAR to pole markers. Note that you would probably be doing trilateration instead of triangulation. This can get tricky if the ground is very uneven and distances are large. If the robot is tilted, the LiDAR on the robot might go off the top of the pole, or hit the ground instead. And you will also have to deal with picking out the poles in the LiDAR data which might be difficult if there are other trees around. But I think I've seen a beach drawing robot that uses this technique.

3. Dead Reckoning. This won't work due to wheel slip and large distances.

4. AR tags. Besides the issues with having cameras outdoors and keeping the lens clean. This can suffer because of the distances involved. Seeing an AR tag from across the yard with enough resolution to localize to it means a VERY big tag. Then when you are close to the tag, it will be too large to use...

5. Infrared fence. I doubt this will work very well in full sun. The sun puts out a lot of infrared and will blind and washout your IR receivers.

6. Rope fence. Actually, this is probably the cheapest. You can't go wrong with a physical boundary. In general, I think the more you can engineer the environment, the easier it will be to localize the robot and keep it contained.

7. GPS. Right, GPS alone isn't all that accurate. Especially next to a house or under trees where you have limited satellites in view. There is differential GPS which uses a fixed GPS base station, and GPS on the robot. It will give you better performance. But you still have to deal with satellite visibility (and seeing the same satellites between your two receivers). But it's hard for me to say if this will be accurate enough for your application. Actually, there is one robot lawn mower that uses differential GPS.