Location Tracking

Question

Ok so we have all seen autonomous lawnmowers. I am currently building my own which is relatively easy to do the problem is location tracking. I do not want to use a bunch of burried wire to create a path for it to follow and just setting up a perimeter makes it wander around aimlessly. GPS is not accurate enough to get propper location RF signals seem to be extremely complicated and relatively expensive so my question is is there any other way to track the location of a rover acuratley enough to be within 1 foot or less?

shabaz · Answer

Hi Robert,

Without local things to sense (like buried wires etc) I think it's not easy to do it repeatably with reliability, and that's of prime importance for your use-case I guess.

If you have the ability to mount a camera high up with visibility of the entire area (not at a shallow angle) then things could be easier, but would require something like a Pi.

With that and the camera attachment plugged in, you can use software called 'OpenCV' to recognize the lawnmower and its co-ordinates. It is quite effective, and easier still if there is some special color or large pattern on the lawnmower.

Then some maths is needed to translate that into a real position and then translate that into instructions that would direct the lawnmower correctly. It's a lot of work to do though, it wouldn't be a quick solution.

You could also check out ROS (Robotic Operating System) - that may have suitable plugins or apps for camera feeds (e.g. a camera attached to your device) and image processing, but again this is a lot of work if you've never used it before. That's not to say you shouldn't attempt it - lots of useful skills can be learnt by doing this, even if the end goal is still far away, and you may then be in a better position to decide if you wish to continue that path, or not.

If you have the time and a Linux box (e.g. Pi) and some camera, it is quite straightforward to get going with OpenCV.

To do that, go to https://www.raspberrypi.org/magpi/ and in the search box (unintuitively not at the top of the page, but below the banner) type opencv to find articles on image recognition type topics which is what the camera method approximately needs - I have not read them, but they are designed for beginners. You'll probably also want to get a book on the subject from your library or Amazon etc.

jw0752 · Answer

I have a robot vacuum cleaner that does a great job of mapping a room. It has several forms of obstacle detection but seems to have a way of counting its steps (wheel revs) and keeping track of its direction by looking at the differential between left and right wheels. This is just a guess on my part but it is the only thing that seems to explain it ability to vacuum the room just as I would mow the yard, back and forth with a slight over lap. John

Gough Lui · Answer

If not via a visual system, radio is a bit tricky to localise, but light generally is not. I wonder whether you could rig up some criss-crossed high-power infrared beams or something like that. Detect obstruction by lawnmower to determine location. Or do a little more sophisticated like some of the laser-ranging devices and actually get a distance measurement as well? A bit hard in outdoor sunlight or inclement weather though. - Gough

mcb1 · Answer

Trimble used to do this and it was used by Local authorities. This was back in the day when there was a deliberate 'wobble' on GPS signals for 'domestic' use, but as I understand they came to an agreement and it was removed. Aviation DME's use the time of signal to determine distance but the type of accuracy you're looking for is too small. Ultrasonic time of flight might work, but you're also going to encounter wind changing it. The newer 'no moving parts' Anenometers are based on this principle with three heads at a known distance to give wind speed and direction. You could certainly determine angle easily, it just the distance bit. Mark

dougw · Answer

One way to know position is simply measure 2 distances. For example put a stake at each of 2 corners of the property and measure the distance to each stake from the robot using any distance sensors. The third side of the triangle (between the stakes) is fixed and known so you know the position of the robot. The measurements can be as simple as string pots, but many other technologies can be used to measure the distance. Another simple method is to use dead reckoning using a compass and wheel counting. This method probably also needs a perimeter detection system.

rgiguere · Answer

I thpught of using ultra sonic sensors for the prototype but i just cant seem to find any that will travel across a distance greater than a few inches, great for avoidance but thats about it. I am really intrigued by the gps method i just found a gps by navspark that has RTK built in for roughly 80 dollars and works with srduino and Raspberry pi i ordered one since the price was good so i guess we will see thanks for your input much appreciated

Gough Lui · Answer

Glad Bill Clinton ordered the turning off of Selective Availability. If we had access to the military receivers, they're about 10x as accurate due to the higher chipping frequency. Carrier phase tracking GPS as used by surveyors are quite expensive - but there were some localized GPS-like units like Locata that could measure distances from a known sirvey point reference station. Still a highly expensive endeavour. If all one wants to do is mow tthe lawn, perhaps some beefy motors, rope and pulleys would do. Have the lawnmower like the head of a 2D plotter and pull it around with two ropes - X and Y. Unless the rope is cut, you know where it's at by measuring/encoding revolutions on the pulley. One thing we should be careful of is safety. We don't need motorized equipment going AWOL. - Gough

Gough Lui · Answer

As far as I know, RTK still needs a reference base station to function and those are very expensive. - Gough

dougw · Answer

Maxbotix makes lots of longer range ultrasonic sensors for example MB7363 is an outdoor sensor that reaches 10 meters.

Location Tracking

Top Replies