D3 Cliff Sensor Sensitivity Issue

[cherokee180c]

Is anyone else seeing this behavior? My floor is a medium dark Cherry hardwood. The X-21 has no issues at all, but the new D3 will go into full reverse when traveling full speed forward in the middle of the room with nothing around it. It will then normally continue straight or act to move around a ghost object. It will only happen a few times per cleaning cycle, but it must be related to the cliff sensor underneath not getting a reflection back momentarily. It is still a few too many times as the butt end flies up and then slams back down to the floor. Hopefully Neato can simply increase to "true" time before the robot reacts to the signal to reduce false positives if it can't actually dial in the sensitivity level of the sensor itself through code. The other thing I would like to see is the fine tuning of Extra care mode. It works well right now to not hit anything, but appears to be about twice the distance away from objects than necessary. Lastly the navigation appears very screwed up under the regular setting. My robot literally spent 50% of its battery under a single couch, and then it would come out, go around a doorway under a chair, and then repeat the pattern continuously back and forth. I finally shut down and pulled it out, but it was never getting stuck. Once I turned on Extra Care mode everything was fine. Other posts seem to suggest weird navigation issues since last firmware update, is that correct?

[someoneclever]

The 2.2.0 post is for the Botvac Connected-the D3 and D5 run different firmware.


Sent from my iPhone using Tapatalk

[glnc222]

1. The back flipping up is an effect of the suspension pivot in the rear and is a complement to lifting the front. Yet I always found it too much and limited the tilt in thread "tilt reduction mod". It is mainly a problem beneath certain cabinets.

2. The cliff or drop sensors have a complication. They must not trigger on drops small enough to navigate, requiring a specific calibration. Sensitivity adjustments are therefore not very practical. Time delays are tricky in that the robot must respond quickly to detected cliffs not to run over. (It is possible to add simple electronics to the sensors to add delays, for the technically skilled.)
Black colors are an old story, and if you do not have any stairway cliffs etc., the solution is to disable the cliff sensors with a false floor, light colored tape or paper over the sensor windows. The problem has previously occurred on black patterns on carpets besides solid dark materials. A variety of materials, wood molding strips, plastic corner guard angle strips, can be used for physical barriers, as well as mag strips, when the cliff sensors are disabled.
The robots are hardly perfect and designs continue to evolve slowly to handle furniture and flooring features.
The cliff sensor problem with black colors affects all robot brands relying on the optical sensing, and they can all use the tape fix. I have wondered whether they should use a mechanical feeler of some sort instead, like a spring mounted roller or sensors on existing rollers. Yet there usually needs to be two or even three sensors, often in front of any rollers.

[JEfromCanada]

So, cliff sensors are strictly optical? What if they used the same technology as LIDAR, so they could measure distances? Wouldn't that allow the unit to compute whether the "drop" was manageable for navigation?

[glnc222]

Given how some black colors have blocked very bright IR proximity detectors, I can imagine weaker lidar pulses also being absorbed. Also more expensive I think. Lasers are weak for eye safety. Also, the timing of returns made by lidars may not be practical at very short distances, frequencies too high, the velocity of light... There are laser computer mouse devices, but do not measure timing of reflections. There have been a few references to using these on the bottom for navigation, tracking motion, maybe LG?.

[JEfromCanada]

I think that a laser-based system aimed downward at a 20 - 30 degree angle would be sufficient to get a reasonable reflection, and would also provide sufficient "advance notice" of an impending drop off before getting too close to a ledge.

[Dejac]

Is this the behavior at 1:10 in the below video the same as the one you’re experiencing? Then you’re not alone.

[JEfromCanada]

Looks like a bump sensor issue to me (not cliff sensor). If it was a cliff sensor, I'd expect to see more backward movement. I thought it might be a wheel drive problem, except when it turns around and heads back to the door opening, it is able to turn counter-clockwise without issues.

[cherokee180c]

I'll try to video mine when I get the chance. It reverses hard while in full forward for no reason, but only a few times per run