Hyperdimensional Uncertainty Quantification for Multimodal Uncertainty Fusion in Autonomous Vehicles Perception

2025-03-25

AI Summary

Existing uncertainty quantification (UQ) methods for autonomous driving perception predominantly focus on task-level outputs, neglecting epistemic uncertainty arising during multimodal feature fusion; moreover, Bayesian approaches incur prohibitive computational overhead, hindering real-time deployment. This paper proposes an efficient deterministic uncertainty quantification (DUM) framework—the first to explicitly model and suppress multi-source epistemic uncertainty at the feature fusion layer. Its core innovations include: (i) a channel- and spatial-granularity uncertainty projection and bundling mechanism grounded in hyperdimensional computing; and (ii) an uncertainty-aware adaptive weighting strategy for feature fusion. Evaluated on 3D object detection, DUM improves mAP by 2.01% and 1.27% under two benchmarks; for semantic segmentation, it boosts mIoU by 1.29%. Crucially, DUM reduces computational cost to 42% and parameter count to just 2.6% of current state-of-the-art methods, achieving unprecedented balance between accuracy and efficiency.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Reasoning under Uncertainty: Uncertainty Representations
• Computer Vision: Multi-modal Vision

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

Uncertainty Quantification (UQ) is crucial for ensuring the reliability of machine learning models deployed in real-world autonomous systems. However, existing approaches typically quantify task-level output prediction uncertainty without considering epistemic uncertainty at the multimodal feature fusion level, leading to sub-optimal outcomes. Additionally, popular uncertainty quantification methods, e.g., Bayesian approximations, remain challenging to deploy in practice due to high computational costs in training and inference. In this paper, we propose HyperDUM, a novel deterministic uncertainty method (DUM) that efficiently quantifies feature-level epistemic uncertainty by leveraging hyperdimensional computing. Our method captures the channel and spatial uncertainties through channel and patch -wise projection and bundling techniques respectively. Multimodal sensor features are then adaptively weighted to mitigate uncertainty propagation and improve feature fusion. Our evaluations show that HyperDUM on average outperforms the state-of-the-art (SOTA) algorithms by up to 2.01%/1.27% in 3D Object Detection and up to 1.29% improvement over baselines in semantic segmentation tasks under various types of uncertainties. Notably, HyperDUM requires 2.36x less Floating Point Operations and up to 38.30x less parameters than SOTA methods, providing an efficient solution for real-world autonomous systems.

Problem

• Quantify feature-level epistemic uncertainty in autonomous vehicles perception
• Address computational inefficiency in existing uncertainty quantification methods
• Improve multimodal sensor feature fusion by adaptive uncertainty weighting

Innovation

• Hyperdimensional computing for uncertainty quantification
• Channel and patch-wise projection techniques
• Adaptive weighting for multimodal feature fusion