Andrew Capodieci
Papers
2
Total Citations
21
H-Index
2
About
Andrew Capodieci is a leading researcher at the intersection of robotic perception, probabilistic AI, and 3D semantic mapping. His work addresses a critical challenge in modern robotics: bridging the gap between efficient, data-driven deep learning and the mathematical rigor of classical probabilistic methods. Capodieci’s major contributions center on developing novel frameworks that are both interpretable and trustworthy. He is best known for introducing **Convolutional Bayesian Kernel Inference (ConvBKI)**, a modular neural network that enables real-time (>10 Hz) probabilistic semantic mapping with quantifiable uncertainty. This approach, detailed in his most-cited paper (2023, 14 citations) and its follow-up (2024, 7 citations), explicitly updates per-voxel probability distributions within a neural network layer, allowing robots to understand not just *what* an object is, but *how certain* they are about it. By fusing the reliability of classical algorithms with the speed of modern architectures, Capodieci’s work is paving the way for safer, more robust autonomous systems in unstructured environments. His research is essential reading for anyone interested in building AI that is both powerful and principled.
Research Focus
Key Achievements
Top Papers
- 1Convolutional Bayesian Kernel Inference for 3D Semantic Mapping14 citations · 2023
- 2