← Back ICRA 2026

E2-BKI: Evidential Ellipsoidal Bayesian Kernel Inference for Uncertainty-Aware Gaussian Semantic Mapping

Junyoung Kim, Minsik Jeon, Jihong Min, Kiho Kwak, Junwon Seo

PDF

AI summary

Key figure (auto-extracted from paper)

E2-BKI unifies semantic, spatial, and observation uncertainty handling via anisotropic Gaussian primitives, significantly improving mapping accuracy, calibration, and robustness in complex outdoor environments.

Semantic mapping Uncertainty quantification Gaussian primitives Bayesian kernel inference Outdoor robotics Evidential deep learning

Problem

Existing continuous semantic mapping methods fail to account for multiple uncertainty sources in challenging outdoor settings, leading to geometric misalignment, unreliable predictions from sparse sensor data, and poor uncertainty calibration.

Approach

The method aggregates noisy semantic points into anisotropic Gaussian primitives to capture local geometry, then applies an evidential ellipsoidal kernel within Bayesian Kernel Inference to adaptively weight observations based on scene structure and prediction confidence.

Key results

Unified uncertainty-aware mapping framework with anisotropic Gaussian primitives
Evidential ellipsoidal kernel for geometry-aligned, uncertainty-aware fusion
Improved semantic accuracy, uncertainty calibration, and geometric completeness
Real-time efficiency with superior robustness to sparse and noisy data

Why it matters

Provides robots with a robust, uncertainty-calibrated semantic mapping solution essential for safe navigation and operation in unpredictable outdoor environments.

Abstract

Semantic mapping aims to construct a 3D se- mantic representation of the environment, providing essential knowledge for robots operating in complex outdoor settings. While Bayesian Kernel Inference (BKI) addresses discontinu- ities of map inference from sparse sensor data, existing semantic mapping methods suffer from various sources of uncertainties in challenging outdoor environments. To address these issues, we propose an uncertainty-aware semantic mapping framework that handles multiple sources of uncertainties, which signifi- cantly degrade mapping performance. Our method estimates uncertainties in semantic predictions using Evidential Deep Learning and incorporates them into BKI for robust semantic inference. It further aggregates noisy observations into coherent Gaussian representations to mitigate the impact of unreliable points, while employing geometry-aligned kernels that adapt to complex scene structures. These Gaussian primitives effectively fuse local geometric and semantic information, enabling ro- bust, uncertainty-aware mapping in complex outdoor scenarios. Comprehensive evaluation across diverse off-road and urban outdoor environments demonstrates consistent improvements in mapping quality, uncertainty calibration, representational flex- ibility, and robustness, while maintaining real-time efficiency. Our project website: https://e2-bki.github.io/

Index terms

Mapping Semantic Scene Understanding Deep Learning for Visual Perception