Adaptive LiDAR Scanning: Harnessing Temporal Cues for Efficient 3D Object Detection via Multi-Modal Fusion

2025-08-02

AI Summary

Traditional LiDAR dense scanning overlooks temporal continuity in dynamic scenes, leading to redundant perception and high power consumption. To address this, we propose a history-aware adaptive sparse-dense hybrid scanning framework. Our method jointly leverages LiDAR and RGB temporal sequences, incorporating a lightweight prediction network and a differentiable binary mask generator; region-of-interest (ROI)-guided dynamic scanning is realized via Gumbel-Softmax relaxation. Crucially, we explicitly model temporal continuity as a guiding signal for scan policy optimization—a novel formulation in LiDAR scanning. Evaluated on nuScenes and Lyft, our approach reduces LiDAR energy consumption by over 65% while maintaining or surpassing the detection accuracy of full-scan baselines. This yields substantial improvements in energy efficiency and practical deployability on resource-constrained platforms.

Technology Category

• Planning, Routing, and Scheduling: Optimization of Spatio-temporal Systems
• Search and Optimization: Mixed Discrete/Continuous Search
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines
• Systems and Infrastructure for Web, Mobile and WoT: Energy management for devices in mobile Web and WoT environments
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

Multi-sensor fusion using LiDAR and RGB cameras significantly enhances 3D object detection task. However, conventional LiDAR sensors perform dense, stateless scans, ignoring the strong temporal continuity in real-world scenes. This leads to substantial sensing redundancy and excessive power consumption, limiting their practicality on resource-constrained platforms. To address this inefficiency, we propose a predictive, history-aware adaptive scanning framework that anticipates informative regions of interest (ROI) based on past observations. Our approach introduces a lightweight predictor network that distills historical spatial and temporal contexts into refined query embeddings. These embeddings guide a differentiable Mask Generator network, which leverages Gumbel-Softmax sampling to produce binary masks identifying critical ROIs for the upcoming frame. Our method significantly reduces unnecessary data acquisition by concentrating dense LiDAR scanning only within these ROIs and sparsely sampling elsewhere. Experiments on nuScenes and Lyft benchmarks demonstrate that our adaptive scanning strategy reduces LiDAR energy consumption by over 65% while maintaining competitive or even superior 3D object detection performance compared to traditional LiDAR-camera fusion methods with dense LiDAR scanning.

Problem

• Reduces LiDAR energy consumption via adaptive scanning
• Improves 3D object detection with temporal context
• Minimizes sensing redundancy in multi-modal fusion

Innovation

• Predictive adaptive LiDAR scanning using temporal cues
• Lightweight predictor network for ROI identification
• Differentiable Mask Generator with Gumbel-Softmax sampling