ill make a try asap with your settings.

Oh the robot has no rear view. A little less than 50% of the scan is missing. The trouble with how slam works is gonna be a problem. Slam tries to determine where the robot is on the map by looking for distance measurements that position it within the previous map.

when a huge part of the data is missing, there’s no easy way to identify the location. Specifically because there isn’t enough stored map data to compare against.

use a Dremmel and cut a hole in your holder behind the lidar.

when I get home I’ll take pics of my setup on one of my robots that has a camera and sensors above the lidar. That will give you ideas.

ok. I will see your design and modify mine.

I’ve modified my robot’s design based on the example shown in the video Add LIDAR Navigation to a Robot. I’ve attached some photos for your reference.

At the moment, I’m unable to test the navigation because of a power issue-my battery shut off unexpectedly, and now Windows won’t boot. I’ll need to resolve this problem first before I can begin testing the navigation features.

I really hope the new design is correct, but please feel free to point out anything

That does look closer to what DJ might have been referring to. It is a holiday weekend, so no one is really around.

I do have two concerns about your change.

1. make sure your realsense offset reflects its location relative to the center of the sensor. Every single fraction of a mm matters - and there’s no internal mathematical interpolation that can resolve it. The dimensions are measured and used for the slam navigation which requires incredibly high precision.  Your sensor is offset from the lidar.

2. the opening will help,  but the angle of the lidar laser against the leg will cut off a significant amount of measurement data. It would be best if the leg was exactly right and left centered of the lidar. That way, it’s only cutting off 2-5 degrees based on thickness. The configuration you have now is still blocking significant measurement.

many thanks for your help . I will modify according to your recommendations. I always thought the position of the t265 is not that critical, just facing the front.

Oh yeah that’s good suggestion from support. It’s crazy how important the measurements are - until you realize that a small error in the math continues and grows over time.

The t265 is okay but I don’t use it often. I find it really starts to drift. Sometimes I’ve seen a drift of 20-30 cm over a short period of time. The problem with the drift can usually be resolved if using the last option (I think) in the pose hint type in the better navigate. I think it’s the last, I can’t recall top of my head.

but the last option uses the difference between the slam and the external (t265) and keeps the difference for the next time.

Yeah - Dynamic is what it's called. This is from the Better Navigator robot skill manual page.

However, I'd first ensure the lidar works correctly by using "Hector" as the pose hint type because that will depend entirely on the lidar's positioning and scanning. The lidar's map will be the best option for determining where the robot is on the map, without using external sensor data. Once you get that dialed in correctly, you can add additional sensors. For example, only add one sensor at a time to ensure they complement each other.

If you do use Hector as the pose hint type, don't have any other odometer pose data being sent into the NMS, or it will confuse it.