arXiv subjects

Heejun Park

Publications and source records attributed to Heejun Park.

7 recordsLinked to original sources

arXiv2026

VIPS: Vehicle-Infrastructure Cooperative Planning Benchmark via Pseudo-Simulation

End-to-end autonomous driving in urban environments requires robust decision-making under partial observability and complex multi-agent interactions. Severe occlusions and dense traffic at intersections limit the perception capability of single-agent systems, motivating recent efforts on Vehicle-to-Infrastructure (V2I) cooperation for perception and planning. However, existing evaluation protocols face a fundamental trade-off: open-loop evaluation fails to capture error accumulation and recovery from deviations, while closed-loop evaluation is costly, difficult to scale, and often relies on simulated environments that may suffer from domain gaps. To bridge this gap, we propose VIPS, a benchmark for cooperative autonomous driving in V2I settings based on pseudo-simulation. VIPS extends pseudo-simulation by integrating vehicle and infrastructure observations. This enables scalable yet realistic evaluation of robustness and error propagation without full simulation. We further present CoS-V2X, a cooperative planning framework based on sparse representations. CoS-V2X models vehicle-infrastructure interactions using compact features for efficient communication and robust decision-making under heterogeneous observations. Code and dataset are available at https://vips2026.github.io.

cs.CV↗

arXiv2026

RAF: Reliability-Aware Fusion of Camera, LiDAR, and 4D RADAR for Robust 3D Object Detection in Adverse Weather

Robust 3D object detection in adverse weather conditions is challenging due to sensor limitations. Although combining complementary modalities such as LiDAR and 4D RADAR has shown promise, the sparsity of these sensors becomes apparent in adverse weather with reduced reflections, leading to objects with few or no point cloud returns. To address this limitation, camera sensors provide visual cues even when LiDAR and RADAR signals are weakened. However, cameras themselves are also vulnerable to adverse weather, where some regions become unreliable due to snow or rain occluding the camera lens. While some camera-fusion methods designed for adverse weather learn to weigh image regions via confidence maps, these maps receive no direct supervision and are learned solely through the detection loss. We introduce Reliability-Aware Fusion (RAF), which explicitly supervises per-pixel reliability estimation and provides a direct learning signal for identifying and suppressing unreliable visual cues. Our framework leverages pretrained LiDAR-RADAR networks, keeping their backbones frozen while only training the added camera branch, BEV fusion encoder, and detection head. Extensive experiments on the K-Radar and VoD datasets demonstrate that integrating RAF consistently improves detection accuracy over LiDAR-RADAR baselines, achieving up to +6.5 $AP_{BEV}$ and +7.4 $AP_{3D}$ gains. Code is available at https://github.com/parkie0517/RAF.

cs.CV↗

arXiv2026

HEAT: Heterogeneous End-to-End Autonomous Driving via Trajectory-Guided World Models

End-to-end autonomous driving has emerged as a compelling alternative to traditional modular pipelines by directly mapping raw sensor data to driving actions. While recent approaches achieve strong performance on single-domain datasets, their performance degrades significantly when trained jointly across multiple heterogeneous domains. In practice, however, autonomous systems must operate across diverse environments with heterogeneous distributions, including different cities, sensor configurations, and traffic patterns, without domain-specific retraining. This gap highlights a key challenge in multi-domain learning: domain-specific variations across heterogeneous domains introduce conflicting learning signals, driving models toward compromised solutions that are suboptimal across domains. To address this, we propose a trajectory-driven learning paradigm that organizes training around planning trajectories, enabling the model to capture domain-invariant representations of driving intent. Furthermore, we incorporate a world model that predicts future latent features conditioned on ego actions, improving feature consistency and mitigating domain-induced biases. We evaluate our approach on three benchmarks, nuScenes, NAVSIM, and the Waymo end-to-end dataset, and show substantial improvements over existing methods across all domains. Our results demonstrate that a single unified model can be trained on heterogeneous datasets while maintaining strong performance within each domain, highlighting a step toward scalable real-world deployment. We will make our code publicly available.

cs.RO↗

arXiv2025

VR-Drive: Viewpoint-Robust End-to-End Driving with Feed-Forward 3D Gaussian Splatting

End-to-end autonomous driving (E2E-AD) has emerged as a promising paradigm that unifies perception, prediction, and planning into a holistic, data-driven framework. However, achieving robustness to varying camera viewpoints, a common real-world challenge due to diverse vehicle configurations, remains an open problem. In this work, we propose VR-Drive, a novel E2E-AD framework that addresses viewpoint generalization by jointly learning 3D scene reconstruction as an auxiliary task to enable planning-aware view synthesis. Unlike prior scene-specific synthesis approaches, VR-Drive adopts a feed-forward inference strategy that supports online training-time augmentation from sparse views without additional annotations. To further improve viewpoint consistency, we introduce a viewpoint-mixed memory bank that facilitates temporal interaction across multiple viewpoints and a viewpoint-consistent distillation strategy that transfers knowledge from original to synthesized views. Trained in a fully end-to-end manner, VR-Drive effectively mitigates synthesis-induced noise and improves planning under viewpoint shifts. In addition, we release a new benchmark dataset to evaluate E2E-AD performance under novel camera viewpoints, enabling comprehensive analysis. Our results demonstrate that VR-Drive is a scalable and robust solution for the real-world deployment of end-to-end autonomous driving systems.

cs.CV↗

arXiv2024

Search for Dark Matter Axions with Tunable TM_020 mode

Axions are hypothesized particles believed to potentially resolve two major puzzles in modern physics: the strong CP problem and the nature of dark matter. Cavity-based axion haloscopes represent the most sensitive tools for probing their theoretically favored couplings to photons in the microelectronvolt range. However, as the search mass (or frequency) increases, the detection efficiency decreases, largely due to a decrease in cavity volume. Despite the potential of higher-order resonant modes to preserve experimental volume, their practical application in searches has been limited by the challenge of maintaining a high form factor over a reasonably wide search bandwidth. We introduce an innovative tuning method that uses the unique properties of auxetic materials, designed to effectively tune higher modes. This approach was applied to the TM_020 mode for a dark matter axion search exploring a mass range from 21.38 to 21.79 ueV, resulting in the establishment of new exclusion limits for axion-photon coupling greater than approximately 10^-13 GeV^-1. These findings signify a breakthrough, demonstrating that our tuning mechanism facilitates the practical utilization of higher-order modes for cavity haloscope searches.

hep-ex↗

arXiv2022

Axion Haloscope Using an 18 T High Temperature Superconducting Magnet

We report details on the axion dark matter search experiment that uses the innovative technologies of a High-Temperature Superconducting (HTS) magnet and a Josephson Parametric Converter (JPC). An 18 T HTS solenoid magnet is developed for this experiment. The JPC is used as the first stage amplifier to achieve a near quantum-limited low-noise condition. The first dark matter axion search was performed with the 18 T axion haloscope. The scan frequency range is from 4.7789 GHz to 4.8094 GHz (30.5 MHz range). No significant signal consistent with Galactic dark matter axion is observed. Our results set the best limit of the axion-photon-photon coupling ($g_{a\gamma\gamma}$) in the axion mass range of 19.764 to 19.890 $\mu$eV. Using the Bayesian method, the upper bounds of $g_{a\gamma\gamma}$ are set at 0.98$\times|g_{a\gamma\gamma}^{\text{KSVZ}}|$ (1.11$\times|g_{a\gamma\gamma}^{\text{KSVZ}}|$) in the mass ranges of 19.764 to 19.771 $\mu$eV (19.863 to 19.890 $\mu$eV), and at 1.76 $\times|g_{a\gamma\gamma}^{\text{KSVZ}}|$ in the mass ranges of 19.772 to 19.863 $\mu$eV with 90\% confidence level, respectively. We report design, construction, operation, and data analysis of the 18 T axion haloscope experiment.

hep-ex↗

arXiv2022

Searching for Invisible Axion Dark Matter with an 18T Magnet Haloscope

We report the first search results for axion dark matter using an 18\,T high-temperature superconducting magnet haloscope. The scan frequency ranges from 4.7789 to 4.8094\,GHz. No significant signal consistent with the Galactic halo dark matter axion is observed. The results set the best upper bound of axion-photon-photon coupling ($g_{a\gamma\gamma}$) in the mass ranges of 19.764 to 19.771\,$\mu$eV (19.863 to 19.890\,$\mu$eV) at 1.5$\times|g_{a\gamma\gamma}^{\text{KSVZ}}|$ (1.7$\times|g_{a\gamma\gamma}^{\text{KSVZ}}|$), and 19.772 to 19.863\,$\mu$eV at 2.7 $\times|g_{a\gamma\gamma}^{\text{KSVZ}}|$ with 90\% confidence level, respectively. This remarkable sensitivity in the high mass region of dark matter axion is achieved by using the strongest magnetic field among the existing haloscope experiments and realizing a low-noise amplification of microwave signals using a Josephson parametric converter.

hep-ex↗