Smart Fusion for Adaptive Visual-IMU Weighting in cuVSLAM Sparse Bundle Adjustment

Key points

• We introduce a scalar $α\geq1$ that upscales the visual information matrix relative to the IMU preintegration constraint.1
• The IMU-TRUST regime contains $α\in\{1,2\}$ near the calibrated baseline, while the VISION-TRUST regime contains $α\in\{10,20\}$.1
• We train a lightweight Visual Quality Estimator (VQE), implemented as a multilayer perceptron with 914 trainable parameters.1
• It maps a 10-dimensional visual quality feature vector assembled from statistics already maintained by the tracker to a per-SBA regime prediction.1

Sources (1)

• [1]Smart Fusion for Adaptive Visual-IMU Weighting in cuVSLAM Sparse Bundle Adjustment
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 14, 02:09 PM

  In cuVSLAM, visual and inertial measurement unit (IMU) residuals are fused in sparse bundle adjustment (SBA) with a weighting fixed by pre-calibrated noise covariances, leaving no mechanism to adapt the visual-inertial balance at runtime.

  We introduce a scalar $α\geq1$ that upscales the visual information matrix relative to the IMU preintegration constraint.

Extractive summary: sentences quoted from the sources.