KAIST researcher advancing end-to-end 4D radar preprocessing for autonomous driving perception
4
papers here
4
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
9
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-16) and cover the author's entire public record, not just the papers listed here. Reproducibility signal comes from Evolve's own reviews.
Research profile
Output—
The output axis is unscored for now: the record holds yearly work counts, but venue tier, venue type, and artifact-level metadata needed to grade research output are missing from the current OpenAlex profile.
Impact—
The impact axis is unscored: only coarse author-level indicators (h-index 9, i10-index 9) exist, while the per-paper citations and field-normalized influence fields required for impact scoring are not yet available.
Focus0/100
The focus axis currently scores 0, the lowest band on this radar. The raw distribution behind it is tightly clustered — Advanced Neural Network Applications (31), Autonomous Vehicle Technology and Safety (24), Radar Systems and Signal Processing (22), Advanced SAR Imaging Techniques (11) — and fieldDist splits 0.5/0.5 across two adjacent 4D-radar perception labels, so read this score as provisional pending recalibration.
Collaboration46/100
Collaboration scores 39, reflecting a compact working circle: 6 distinct coauthors across 31 recent works, with Seung-Hyun Kong, Seung-Hyun Song, Woo-Jin Jung, and Jeong-Su Park recurring. On both Evolve-reviewed 2026 papers Paek holds a middle author position without corresponding-author status, pointing to a contributor role inside a small KAIST-centered team rather than a broad multi-institution network.
Activity100/100
Activity is the strongest axis at 100. Recorded works rise from 2 in 2020 and 2 in 2021 to 8 in 2022, dip to 3 in 2023, recover to 7 in 2024, and peak at 12 in each of 2025 and 2026, with 31 works in the recent three-year window — an unusually steep publication tempo centered on 4D-radar perception.
Reproducibility55/100
Reproducibility scores 55, averaged over 2 Evolve-reviewed papers, both rated medium readiness. STAR reports +5.6 AP_BEV on K-Radar but evaluates on a single dataset without preprocessing baselines or variance data; the P3DP/MF-KDE paper shows consistent AP gains yet omits ablations and grounds real-time claims in RTX 3090 runs — code and config disclosure will decide the next step.
Research summary
Dong-Hee Paek is a researcher at the Korea Advanced Institute of Science and Technology (KAIST) working on 4D imaging-radar perception for autonomous driving. The OpenAlex-derived record (h-index 9, i10-index 9) shows a steep activity ramp: from 2 works in each of 2020 and 2021 to 8 in 2022 and 12 in each of 2025 and 2026, with 31 works in the recent three-year window. Topic distribution concentrates on Advanced Neural Network Applications (31), Autonomous Vehicle Technology and Safety (24), and Radar Systems and Signal Processing (22), with smaller clusters in SAR imaging (11) and 3D shape modeling (5); the field distribution splits 0.5/0.5 across two adjacent 4D-radar autonomous-driving perception labels. Two 2026 arXiv preprints reviewed by Evolve define the current agenda: STAR, a scene- and task-aware radar preprocessing network that gains +5.6 AP_BEV on K-Radar when trained end-to-end, and an accuracy- and real-time-aware framework introducing P3DP, MF-KDE, and an ENS metric. Both place Paek in a middle author position within a small coauthor circle of six that includes Seung-Hyun Kong, Seung-Hyun Song, Woo-Jin Jung, and Jeong-Su Park. Evolve rates both papers medium on reproducibility, citing single-dataset evaluation, missing ablations, and RTX 3090-based real-time claims.
Collaboration pattern
Collaboration is compact and stable: 6 distinct coauthors across 31 recent works, anchored by Seung-Hyun Kong, Seung-Hyun Song, Woo-Jin Jung, and Jeong-Su Park. On both Evolve-reviewed 2026 papers Paek appears in a middle author position without corresponding-author status, consistent with contributing inside a small KAIST-centered group rather than leading multi-institution consortia; the collaboration dimension scores 39.
Research keywords
Highlights
STAR: task-aware radar preprocessing trained end-to-end
P3DP and MF-KDE: the accuracy-real-time trade-off made explicit
A steep 2025-26 acceleration into learnable radar preprocessing
Annual output rose from 7 works in 2024 to 12 in each of 2025 and 2026 (31 works in the last three years), with the two reviewed preprocessing papers anchoring a focused 4D-radar agenda at KAIST.
Timeline
- 2020First works enter the OpenAlex record: 2 publications.
- 2022Output climbs to 8 works, the strongest year of the early phase.
- 20247 works recorded, rebounding from 3 in 2023.
- 2025Cadence doubles to 12 works, the highest annual count on record.
- 202612 more works; STAR and the P3DP/MF-KDE preprocessing papers enter the Evolve catalog as arXiv preprints.
Papers in the Evolve catalog
STAR: Scene- and Task-Aware 4D Radar Preprocessing Towards End-to-End Cognitive Radar
Accuracy- and Real-Time-Aware 4D Radar Preprocessing for Autonomous Driving Perception Systems
IMM-based Multiple Object Tracking using a State Prediction Neural Network
Frequency Estimation Based on SNR-adaptive Frequency Estimator Under Wide SNR Range
Frequent co-authors
About this profile
This profile is AI-structured by Evolve.Pub from public bibliographic data (OpenAlex) and Evolve's own paper reviews. It is not written or endorsed by the researcher unless the profile is claimed. Radar dimensions are computed from real bibliometrics with fixed formulas — raw figures are shown alongside; missing dimensions mean insufficient data, never zero. Errors? Claim the profile to correct or hide it.
- Bibliometric figures are OpenAlex-derived only; the output and impact dimensions are unscored for lack of underlying data.
- Quality judgments rest on just two Evolve-reviewed items, both 2026 arXiv preprints; no peer-reviewed venue record is on file.
- Identity is name-resolved from public bibliographic data (OpenAlex); the quoted h-index and i10-index are as-of-snapshot values.
- The reproducibility score averages only 2 papers and should be read as indicative, not definitive.