Implement ZED 2i within RTABMap

[zbrotherton]

RTABMap (Real Time Appearance Based Mapping) is the SLAM software we are using; check below for documentation and an introduction to the software:
https://github.com/introlab/rtabmap
https://github.com/introlab/rtabmap/wiki

We're changing cameras to one I'm unfamiliar with, the ZED 2i
I don't know how to use this with RTABMap, and while it doesn't look to be difficult, I need you to look into it

Here's an example:
https://github.com/stereolabs/zed-ros-examples/tree/master/examples/zed_rtabmap_example
Here's a list of all possible RTABMap arguments:
https://github.com/RPI-IGVC-2025/2026RobotCode/wiki/RTABMap-Arguments

Your task is to

1. Implement the ZED2i within our RTABMap configuration by editing or adding parameters
2. Examine all possible RTABMap vision processing configurations (stereo, rgbd sync, etc) and generate a report on the differences
3. Investigate how to compose ZED2i depth images, since the computing is not done onboard the camera

[Roy-sb123]

Passive Stereo (e.g., ZED 2i): Uses two synchronized RGB sensors to calculate depth via triangulation (disparity). It mimics human vision and relies on ambient light.
Active RGB-D (e.g., RealSense D435): Uses a single RGB camera paired with an Infrared (IR) Projector. It calculates depth by projecting a known IR pattern and measuring its deformation or Time-of-Flight (ToF).

table of SLAM camera comparison

https://docs.google.com/spreadsheets/d/1yZ3f9reM813SJJj18z-Q1HFeTB8y30ciJ8zTWEKLgM0/edit?usp=sharing

For the IGVC (Intelligent Ground Vehicle Competition), which occurs in outdoor grass environments:
Choice: ZED 2i (Passive Stereo) is the superior choice.
Reasoning: Active RGB-D sensors lose depth data under direct sunlight, making them unreliable for outdoor navigation.
Hardware Leverage: By utilizing the NVIDIA Jetson GPU, we can offload the ZED’s heavy "Neural Depth" processing, ensuring high-accuracy SLAM and obstacle detection without a significant CPU bottleneck.