Concepts

The conceptual foundations referenced throughout this technical report. Each page is a self-contained treatment of one idea. Pages may be read in any order; the model lineage and evaluation pages link back to these when a concept is introduced by name.

Page	Subject
Bootstrap Perception	The system-level idea that surviving valid sensor pixels calibrate a learned monocular depth model to fill in for the failed regions.
ToF Failure Modes	The mechanism by which the Femto Bolt loses 77.79 % of its pixels in the corridor environment, and the surface classes that produce the failure.
Scale Calibration	The median-scale alignment step that anchors the student’s geometry to metric units using the surviving ToF pixels.
Confidence-Gated Fusion	The per-pixel runtime decision rule combining hardware ToF where valid and scaled student depth elsewhere.
Four-Layer Sensing Hierarchy	The graceful-degradation chain spanning LiDAR, hardware depth, learned depth, and semantic segmentation.
Specification and Deployment Realization	The relationship between the formal pipeline specification (used for evaluation and reporting) and the embedded deployment realization (used for on-vehicle inference).
Knowledge Distillation	The teacher–student training paradigm and the rationale for off-board teacher inference combined with on-board student deployment.