VOXL 2 as external perception module for 2kg proprietary-flight-controller UAV, GPS-denied forest & confined spaces

[Leon Yu]

Hi ModalAI team,
I'm evaluating your VOXL 2 platform as the perception and autonomy module for our UAV, and would like to understand the most production-ready configuration that fits our constraints.
Platform
Proprietary flight controller​ – we intend to keep our FC as the attitude inner loop. VOXL 2 would serve as an external companion computer providing mid-loop perception and autonomy.
Operating environments
Dense forest / under-canopy:​ frequent GPS loss, dynamic vegetation, severe and rapid lighting changes
Pipelines & enclosed spaces:​ narrow passages, low-texture surfaces, potentially total darkness
Seamless transition between outdoor and indoor/tunnel conditions.
Required outputs (after GPS loss)
Self-state estimation:​ relative position and 3D velocity with respect to the environment
Obstacle perception:​ real-time detection and ranging of surroundings (trees, branches, walls, pipeline interiors)
All-light operation​ capability
We understand VOXL SDK provides GPS-denied navigation, SLAM, obstacle avoidance, and ROS 1/2 support. We would like to know:
Which stock VOXL 2 configuration​ (compute board + camera/ToF bundle such as the Immervision 190° wide-FOV module, or VOXL CAM) best matches these requirements?
What is the recommended way for VOXL 2 to output pose, velocity, and obstacle/depth data to our proprietary FC via UART (or alternatively CAN/I2C)? Is there a standard VOXL SDK node or message protocol for this, or would custom driver work be required on our side?
Our goal is to minimize Non-Recurring Engineering cost by using as much off-the-shelf hardware and software as possible. We leave the specific hardware composition to your recommendation based on our constraints.