Selected work

Representative Papers

Belief-Aware Scheduling for Predictive Wildfire Hazard Mapping under Sparse-Window Telemetry

Jun 05, 2026

This study addresses the challenge of jointly scheduling regional sensing, representation, and transmission under telemetry bandwidth constraints—modeled as sparse observation windows—to enable accurate prediction of future wildfire risk maps at the receiver. The problem is formulated as a partially observable sequential resource allocation task. To explicitly link scheduling decisions with predictive performance, the authors propose a prediction-oriented structured belief mechanism grounded in the input requirements of the forward operator. Evaluated in a physics-calibrated synthetic environment using a lightweight cross-regional attention encoder (only 40k parameters), the approach outperforms baseline methods by 28% and 11% on default and structured landscapes, respectively. Notably, deeper Transformer architectures yield no average loss improvement and exhibit higher training variance, underscoring the efficacy of the proposed lightweight design.

0 citationsRead paper

Towards Compact Autonomous Driving Perception with Balanced Learning and Multi-sensor Fusion

Jun 01, 2026

This work addresses key challenges in autonomous driving perception—namely model redundancy, imbalanced learning across tasks, and inefficient multi-sensor fusion—by proposing a compact deep multi-task learning architecture that simultaneously performs semantic segmentation, depth estimation, LiDAR segmentation, and bird’s-eye-view projection within a single forward pass. To mitigate inter-task optimization conflicts, an adaptive loss weighting mechanism is introduced, while a novel intermediate-level fusion strategy integrates RGB, dynamic vision sensor (DVS), and LiDAR modalities to enhance environmental understanding. Experimental results demonstrate that the proposed method outperforms existing approaches on both the CARLA simulation environment and the nuScenes-lidarseg real-world dataset, achieving higher accuracy and faster inference speeds while reducing model parameters and GPU memory consumption.

0 citationsRead paper

DeepIPCv3: Event-Aware Multi-Modal Sensor Fusion for Sudden Pedestrian Crossing Avoidance

May 31, 2026

This work addresses the safety limitations of existing end-to-end autonomous driving systems in sudden pedestrian crossing scenarios, where frame-based sensors suffer from perceptual latency and motion blur. To overcome these challenges, the authors propose a multimodal perception framework that fuses LiDAR point clouds with event streams from dynamic vision sensors (DVS). The core innovation lies in a Transformer-inspired cross-modal attention mechanism that dynamically aligns the two asynchronous modalities, coupled with a hybrid policy network to generate safe and formally verifiable local trajectories and control commands. The resulting system exhibits strong robustness to varying illumination conditions and achieves microsecond-level responsiveness. Evaluated on a newly curated multimodal dataset, the approach attains state-of-the-art performance, significantly reducing trajectory and control errors while entirely eliminating issues related to exposure failure and motion blur.

0 citationsRead paper

Tight Bounds on Window Size and Time for Single-Agent Graph Exploration under T-Interval Connectivity

Apr 06, 2026

This study investigates the minimal window size and time required for a single agent to deterministically explore an unknown T-interval-connected dynamic graph, without prior knowledge of T, the number of nodes n, or edges m. Under two local perception models—KT₀ (neighbor IDs invisible) and KT₁ (neighbor IDs visible)—the work establishes, for the first time, a necessary window size of Ω(m). It also proves lower bounds on exploration time: Ω((m−n+1)n) under KT₀ and Ω(m) under KT₁. The proposed deterministic algorithms achieve tight bounds when m = n^{1+Θ(1)}, attaining Θ(n³) time complexity under KT₀ and Θ(n²) under KT₁, with both window size and exploration time matching the respective theoretical lower bounds.

0 citationsRead paper

Mapping Social Media User Behaviors in Reciprocity Space

Jan 22, 2026

This study addresses the limitations of traditional discrete categorizations in modeling social media user behavior, which hinder unified interpretation and analysis. The authors propose the first continuous two-dimensional reciprocity space that encompasses all forms of social interaction, quantifying user engagement patterns through bidirectional connection ratios and naturally mapping diverse behaviors onto continuous regions within this space. Through large-scale empirical analysis of 48,830 Twitter users and 149 million connections, the research demonstrates that user attributes vary smoothly along the reciprocity dimensions, with conventional discrete behavioral types spontaneously clustering into interpretable regions. Beyond revealing a behavioral gradient, the model provides a quantifiable framework for assessing influence, offering a novel paradigm for platform design and user analytics.

0 citationsRead paper

Recent publications

Latest Papers

Belief-Aware Scheduling for Predictive Wildfire Hazard Mapping under Sparse-Window Telemetry

Jun 05, 2026

This study addresses the challenge of jointly scheduling regional sensing, representation, and transmission under telemetry bandwidth constraints—modeled as sparse observation windows—to enable accurate prediction of future wildfire risk maps at the receiver. The problem is formulated as a partially observable sequential resource allocation task. To explicitly link scheduling decisions with predictive performance, the authors propose a prediction-oriented structured belief mechanism grounded in the input requirements of the forward operator. Evaluated in a physics-calibrated synthetic environment using a lightweight cross-regional attention encoder (only 40k parameters), the approach outperforms baseline methods by 28% and 11% on default and structured landscapes, respectively. Notably, deeper Transformer architectures yield no average loss improvement and exhibit higher training variance, underscoring the efficacy of the proposed lightweight design.

0 citationsRead paper

Towards Compact Autonomous Driving Perception with Balanced Learning and Multi-sensor Fusion

Jun 01, 2026

This work addresses key challenges in autonomous driving perception—namely model redundancy, imbalanced learning across tasks, and inefficient multi-sensor fusion—by proposing a compact deep multi-task learning architecture that simultaneously performs semantic segmentation, depth estimation, LiDAR segmentation, and bird’s-eye-view projection within a single forward pass. To mitigate inter-task optimization conflicts, an adaptive loss weighting mechanism is introduced, while a novel intermediate-level fusion strategy integrates RGB, dynamic vision sensor (DVS), and LiDAR modalities to enhance environmental understanding. Experimental results demonstrate that the proposed method outperforms existing approaches on both the CARLA simulation environment and the nuScenes-lidarseg real-world dataset, achieving higher accuracy and faster inference speeds while reducing model parameters and GPU memory consumption.

0 citationsRead paper

DeepIPCv3: Event-Aware Multi-Modal Sensor Fusion for Sudden Pedestrian Crossing Avoidance

May 31, 2026

This work addresses the safety limitations of existing end-to-end autonomous driving systems in sudden pedestrian crossing scenarios, where frame-based sensors suffer from perceptual latency and motion blur. To overcome these challenges, the authors propose a multimodal perception framework that fuses LiDAR point clouds with event streams from dynamic vision sensors (DVS). The core innovation lies in a Transformer-inspired cross-modal attention mechanism that dynamically aligns the two asynchronous modalities, coupled with a hybrid policy network to generate safe and formally verifiable local trajectories and control commands. The resulting system exhibits strong robustness to varying illumination conditions and achieves microsecond-level responsiveness. Evaluated on a newly curated multimodal dataset, the approach attains state-of-the-art performance, significantly reducing trajectory and control errors while entirely eliminating issues related to exposure failure and motion blur.

0 citationsRead paper

Tight Bounds on Window Size and Time for Single-Agent Graph Exploration under T-Interval Connectivity

Apr 06, 2026

This study investigates the minimal window size and time required for a single agent to deterministically explore an unknown T-interval-connected dynamic graph, without prior knowledge of T, the number of nodes n, or edges m. Under two local perception models—KT₀ (neighbor IDs invisible) and KT₁ (neighbor IDs visible)—the work establishes, for the first time, a necessary window size of Ω(m). It also proves lower bounds on exploration time: Ω((m−n+1)n) under KT₀ and Ω(m) under KT₁. The proposed deterministic algorithms achieve tight bounds when m = n^{1+Θ(1)}, attaining Θ(n³) time complexity under KT₀ and Θ(n²) under KT₁, with both window size and exploration time matching the respective theoretical lower bounds.

0 citationsRead paper

Mapping Social Media User Behaviors in Reciprocity Space

Jan 22, 2026

This study addresses the limitations of traditional discrete categorizations in modeling social media user behavior, which hinder unified interpretation and analysis. The authors propose the first continuous two-dimensional reciprocity space that encompasses all forms of social interaction, quantifying user engagement patterns through bidirectional connection ratios and naturally mapping diverse behaviors onto continuous regions within this space. Through large-scale empirical analysis of 48,830 Twitter users and 149 million connections, the research demonstrates that user attributes vary smoothly along the reciprocity dimensions, with conventional discrete behavioral types spontaneously clustering into interpretable regions. Beyond revealing a behavioral gradient, the model provides a quantifiable framework for assessing influence, offering a novel paradigm for platform design and user analytics.

0 citationsRead paper