I also have need for the core to adapt to the slope of the terrain. My experience is limited to only the M1 module, however.
With the M1 module, Yarbo is able to track a straight line UP a slope with little issue.
Downward or across with a downward aspect, however, is another story. All the weight is on a small section of the front of the tracks, and the weight of the M1 typically pulls the Yarbo off-course, directly downhill.
Demonstration video -
Camera is at top of fairly mild slope.
- Yarbo tracks wonderfully up the slope.
- Yarbo starts in the correct direction down the slope.
- M1 pulls the Yarbo off course, directly downhill.
- Core issues a course-correction when it is 3/4s of a Yarbo out of the lane.
- Yarbo heads back up the slope, perfectly in the lane.
The behavior is fairly consistent. If the direction of travel puts the M1 above the core on the slope, the lane will be mostly perfect. If not, the M1 will tend to pull the core toward the directly-downhill direction.
There is a related behavior with turning on a slope. When trying to (turn or) pivot up-hill, the core will drive the outside track forward, while parking the inside track. The M1 is typically too heavy for the pivot to take place on any appreciable slope, so the outside track will slip.
The tilt sensor could be utilized to predict this, instead parking the outside track and reversing the inside track. This would probably yield a higher success rate with less lawn damage. Maybe.
I suspect that some of this might be a side-effect of the M1 using the S1 treads when things were initially developed. Those S1 treads will horse the M1 in any direction you want, on any hillside. The new M1-specific treads are wonderful, but the algo needs some tweaks for them to succeed in non-flat areas.
The above photo is from a different slope, and demonstrates an extreme example of the traction issue when Yarbo slows itself down. Yarbo can CLIMB this hill like a champion. Down, however, is borderline “Hold My Beer”. At the point in the photo, the mass of the M1 dictates where the core is headed, and the tracks are merely a suggestion. The Yarbo typically ends up almost a full lane out of position before course correction takes place, if it even can take place. Trying to back up a hill produces this same effect.
I’ll suggest that the software stack should be able to predict this from the tilt sensor and change strategy, before this takes place. The slope angle where the tracks will stand up should probably be something the core understands, along with the slope angle where the M1 starts pulling the core off-course (as in the above video).
I personally don’t expect the adaptation to work on all slopes. Some will require a new mowing pattern that does not involve turning (like a dead-end that goes up, then down, all along the entire slope). With this pattern, the Yarbo would always climb, then back itself down to the next lane, using the tilt sensor to figure out what “straight up the hill” is for each lane. But I suspect this might be a different enhancement request. I do recall that someone discussed a “no turning at all, just back-and-forth” mowing pattern several months ago during a live-stream. It is also very similar to the “back and forth” pattern requested for the B1 module. The idea being offered here could be that pattern, but would require an option for “M1 always points uphill” by reacting to the tilt sensor.
See if your crew can have this all done by Saturday, ok?
Cheers
