Yes, a magnetometer, defined as “an instrument used for measuring magnetic forces, especially the earth’s magnetism” is what sends readings of the magnetic field in 3 dimensions. Inside Ardupilot, these are stored in variables named mX, mY and mZ (for compass #1) and the Ardupilot calculates its heading from them.
And as you probably have seen, getting a stable accurate reading is difficult on a large metal body with electromechanical devices and moving metal parts, aka a mower!
My experience:
Thanks for the information. That clears things up a bit. Supposedly the RM3100 magnetometer is a sensitive military grade magnetic sensor that will help this issue. We all want our mowers to be robust and reliable.
Although I’ve never actually used them, I think if you get two F9 GPSs it’s possible to use GPS-For-Yaw. There’s a video from Yuri at the bottom of the linked wiki page who seemed to have good success.
By the way, I don’t think the RM3100 will work any better than the compass built into most of the GPS/Compass units out there. Getting the compass away from metal seems to be the big issue, the particular model of compass is not so important.
In Rover-4.1 we think it may be possible that “GSF” will allow the Rover to get a heading estimate using a single GPS without needing a compass although this is not confirmed yet.
@SJohnson, I am eventually headed to the GPs for Yaw (Moving Base) that Randy is talking about. But I’m not in too big of a hurry to spend the $200+ for as another F9P because my mower is very functional as is. Turns are not perfect but repeatable, so pretty or not, the grass gets cut.
Hi,
Been considering such a project for a long time. But the one big question stopping me is:
“What is the mower actually stopping from running over an injured animal or small child playing on the lawn ??”
I certainly don’t want to put a damper on such great project but for me safety has to come first.
Any push bar activated safety switch? Light barrier ahead of mower? …any safety device at all?
If it needs to be supervised all the time than the whole automation thing is little more than fun.
…just saying.
Yes, this worries me as well sometimes. We don’t need to be running over any baby partridges or other living things.
From a technical point of view, it’s tricky to recognise obstacles using a 360 lidar (or even the 3D depth cameras like the Intel RealSense 435) so I think the solution is most likely to be something like this blog discusses - using AI to recognise out-of-place obstacles and then the mower can avoid them.
ArduPilot already has path planning so there’s not much to do except add a new mavlink message to pass in the obstacles direction and distance in. The harder part is the AI part which would need to run on a companion computer (perhaps an NVidia or RPI4). AI is popular these days though so I suspect someone will show up at some point to work on this.
I was thinking of something much more basic, yet effective - at least until some other technology is easily available and reliable:
A bumper switch for example. - Having a flexible PVC tube / pipe across the front of lawnmower with tension spring keeping it away from front of lawnmower. Connected to the mounting arm is a micro- or lever switch which will kill the ignition of motor once a obstacle is being hit.
I like your idea of the switch. I am concerned though especially on the big zero turn mowers with 25hp engines, if you have already made contact is too late.
I read about the moving base option somewhere and for me, if it made my mower work like it is is on rails, I would do it. I have big open areas but I also have several things to mow around. If I mow down my wife’s garden or other plants, I would be in trouble. I don’t think she will let me blame it on the mower.
Do you still need the fixed base station also in moving base option being discussed? The other question I have is how close to the rover does the fixed station need to be? I guess for now I am assuming that I will use the radio antenna option from rover to base. Will I have to move the base station around my property?
The rule of thumb you will see (which comes from surveying guidelines) is that the base should be less than 10km from the rover. I saw mention of a 3 cm inaccuracy at 10km but I don’t really know. I have never noticed a problem with my base located 3000 feet from my rover.
Thanks for the information. I only have 8 acres so if I mount it outside my workshop, the radio on the Ardusimple F9P will reach anywhere on the property with sufficient RTK accuracy.
For talking to the rover from the computer to pass Ardupilot communication I am leaning towards WIFI instead of telemetry radios.
Re. safety switch:
Yes, I’m aware of the issue with having momentum on a large machine/vehicle. It would need such a safety bumper to be arranged well in front of the mower to be effective.
But rather than killing the engine it might also be possible to briefly change mower into reverse. - About 35 years ago I connected a resistor parallel to a servo (internal potentiometer). That resistor was switched via a water sensor inside a RC submarine and changed dive setting to rise sub to the surface no matter what the control setting called for in the event that water had gotten inside.
So same could be done here: Bumper switch activates and overrides gear servo to go into reverse and then perhaps stop.
…Much better would be to have an “obstacle avoidance function” within ArduRover. If the bumper switch is activated the mower moves 90 degree to one side, covers another 3m or so an then moves another 90 degree back and eventually returns to original track after avoiding obstacle.
I am trying to decide which model Pixhawk flight controller to be the heart of the mower build. It seems like the newest Holybro Pixhawk 4 with all the nice built-in I/O ports would be a nice choice. It is bothersome that it is on the closed hardware list for Ardupilot. Is the shortage of memory for the firmware going to be a limitation for the rover configuration now or in the future? I was also considering the Drotek Pixhawk 3 that is on the open list. It actually costs a few dollars more than the Pixhawk 4.
I like the built-in plugs for the connections to the controller. Any suggestions or comments?
I am also starting to understand the importance of vibration isolation for longevity of the flight controller. Is that the reason that Kenny went with the Kakute F7 AIO in his new control box? Supposedly the old black hex cube 2.1 and the Kakute has a tripple isolation system on the IMU. I think this might be important for these mowers with a big gas engine vibrating everything and bouncing around over gopher holes in the big open fields.
Steve,
I went with the Kakute F7 AIO really just to try something new and the cost was so low at about $50.00. I didn’t really research too much. Using the F7 AIO is somewhat of a pain in that it does not have standard connectors as the Pixhawks do (the ones I know about anyway). I ended up building a carrier board to bring the signals out to standard connectors. Even then, I had to solder individual wires from some pads on the the F7 AIO to the carrier board. It is working fine. I have had it lose the mission a few times if I power down and up. I had that same issue with a Pixhawk 2.4.8 after the Pixhawk had been well used and abused. I have not tried to troubleshoot the issue. I have only used my mower maybe 10 time with the F7 AIO. I will be firing it back up soon.
Thanks for the information Kenny. I am considering using either the new Holybro Pixhawk 4 or the old black hex cube 2.1 for my flight controller. They both look easy to hookup. There is actually more technical reasons to use the old black hex cube, with the triple IMU dampening system, the rugged industrial design, and the ease of hookup up for a newbe. The 2.1 is a very proven platform. The Pixhawk 4 is newer and faster but probably not built as rugged. It doesn’t seem like the speed will buy anything. Do you have any suggestion on which one might work better.
Thanks again,
I am also considering mounting the flight controller as forward as and I can get it and up off the deck. There is not much in that area to mount on but I can build something. I also don’t want to use the flight controller as the front bumper either.
I think the Pixhawk4 and CubeBlack will have very similar performance in terms of processor speed because I think they use the same processor.
The big difference comes when moving to a board with an H7 processor like the Cube Orange, Holybro Durandal, CUAV Nora, etc. if price isn’t a major issue I would recommend any of these over either the Cube Black or the Pixhawk4 (or any of the F7 based boards).
