Model submission for Absolem sensor configurations 6,7,8 by CTU-CRAS-Norlab

[peci1]

Depends and builds on top of #859 (sensor configs 3,4,5), #877 and #878. That is why you see so much commits. If you want to see just changes from the sensor configurations, look e.g. at ctu-vras/subt@submitted_models/ctu_cras_norlab_absolem_improved...ctu-vras:submitted_models/ctu_cras_norlab_absolem/clone .

Everything that was written in #859 applies also here. The main difference of these sensor configs is that they do not have the Ladybug omnicamera, and instead they use our custom omnicamera solution based on 5 pcs of Basler ACE 2 Pro cameras. This is beneficial for occlusions by body parts, which are inevitable when using a centralized omnicam, but are nicely avoided by putting each camera on the edge of the robot body.

The field of view of the Basler cameras is experimentally measured. We took a ruler, shot a picture of it, and by using a bit of trigonometry, we ended up with horizontal FOV of 86.5°. For the trigonometrical computation we used the datasheet value of "Front focal length", which is 9.95 mm. We sketched a triangle from the point on lens axis that is 9.95 mm inside the lens (measured from the vertex of the outern-most lens). One side of the triangle we used for computation is distance of the ruler from the point inside lens, and the other side is the distance viewed on the ruler.

The theoretical value for a 4 mm lens on 1/2.3" chip is 79°, so I assume the difference to our measured value is done by distortion, which is not modeled in the simulator, and neither is it modeled in the theoretical equation.

The computation I used for horizontal FOV is:

degrees(2*atan(tan(0.0066)/(2*0.004))) = 79.04 

Where 0.004 is effective focal length and 0.0066 is the sensor width (both from datasheet). This theoretical value is valid for focusing into infinity. Closer focus can decrease the FOV by a few degrees.