Safety-oriented end-to-end autonomous driving researcher at Tsinghua, focused on risk-aware occupancy for planning
2
papers here
2
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
—
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-15) 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—
No output score is available yet: this profile lacks publication-volume counts for this researcher, so output cannot be scored. The axis would measure total papers, output cadence, and authorship roles across the record.
Impact—
Impact is not scored here because citation counts and influence indicators are absent from the current bibliometric snapshot. The axis would measure citations, citation velocity, and uptake of the two reviewed papers.
Focus0/100
The radar focus score is 0. The underlying field distribution splits the two reviewed papers evenly, 0.5 to autonomous driving/computer vision and 0.5 to autonomous driving/machine learning, so the record spans two adjacent subfields rather than one label. With only two papers feeding the metric, this concentration reading is fragile and should be revisited as more output accumulates.
Collaboration39/100
The radar collaboration score is 39, derived from 6 distinct coauthors across the reviewed record. Recurring collaborators include Jiaxing Chen, Yiren Zhao, Bolin Gao, YuKai Qin, and Lidong Yu, indicating a compact, stable team. On both 2026 arXiv papers this researcher holds a middle authorship position without corresponding-author status, so the signature pattern suggests contribution within a group rather than independent leadership so far.
Activity—
Activity is unscored: the snapshot contains no per-year publication totals, so recent publishing cadence cannot be assessed. This axis would measure output frequency and research continuity across recent years.
Reproducibility55/100
The radar reproducibility score is 55, the mean over 2 papers with reproducibility assessments, and both are rated medium readiness. The Evolve verdicts flag what a replication would face: RRDrive's headline PDMS gain from 0.937 to 0.951 leans on oracle future-derived risk labels, and the GenAD collision reductions of 35-53% rest on small-count open-loop metrics without statistical testing.
Research summary
Hengduo Zou is a researcher at Tsinghua University whose Evolve catalog centers on safety-oriented end-to-end autonomous driving, specifically risk-aware occupancy representations. This is an early-stage, name-resolved profile built from public bibliographic data (ORCID 0009-0000-8306-8710) and two 2026 arXiv preprints reviewed by Evolve, on both of which Zou appears as a middle author rather than corresponding author. The first paper proposes risk-aware occupancy for end-to-end driving, reporting GenAD collision reductions of 35-53%; Evolve notes these open-loop, small-count results lack statistical testing. The companion paper, RRDrive, integrates risk-aware occupancy into a register-based planner on NAVSIM, lifting PDMS from 0.937 to 0.951, though the headline gain relies on oracle future-derived risk labels. Field distribution splits evenly between autonomous driving/computer vision and autonomous driving/machine learning, and the collaboration footprint covers 6 distinct coauthors including Jiaxing Chen, Yiren Zhao, Bolin Gao, YuKai Qin, and Lidong Yu. Radar scores are available for focus (0), collaboration (39), and reproducibility (55, medium readiness on both papers); output, impact, and activity axes lack data and are left unscored. The consistent thread is treating occupancy prediction as a risk signal for planning — a direction worth tracking, with the caveats that both papers are preprints and the reported gains carry methodological dependencies.
Collaboration pattern
Collaboration is compact: 6 distinct coauthors appear across the reviewed record, with recurring partners Jiaxing Chen, Yiren Zhao, Bolin Gao, YuKai Qin, and Lidong Yu (Yiren Zhao is listed twice in the frequency data). The radar collaboration score is 39. On both 2026 arXiv preprints Zou sits in a middle authorship position without corresponding-author status, consistent with contributing within a small Tsinghua-affiliated team rather than leading the author line so far.
Research keywords
Highlights
Risk-aware occupancy cuts GenAD collisions by 35-53%
RRDrive lifts NAVSIM PDMS from 0.937 to 0.951
Oracle-label dependency flagged in RRDrive's headline gain
Timeline
- 2026Released the arXiv preprint 'Risk-Aware Occupancy for Safety-Oriented End-to-End Autonomous Driving', proposing risk-aware occupancy and reporting GenAD collision reductions of 35-53%.
- 2026Released the companion preprint 'Driving on Registers, Reasoning on Risk' (RRDrive), lifting NAVSIM PDMS from 0.937 to 0.951 with risk-aware occupancy.
- 2026Both 2026 papers split evenly across autonomous driving/computer vision and autonomous driving/machine learning, marking a continued risk-aware occupancy research thread.
Papers in the Evolve catalog
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.
- Dimension scores come from a bibliometric snapshot of public records; the profile is name-resolved via ORCID 0009-0000-8306-8710 and public bibliographic data.
- Only the two Evolve-reviewed 2026 arXiv preprints feed this profile, both with middle authorship; no thesis, dataset, or other venue output is included.
- Output, impact, and activity dimensions are unscored: citation counts, h-index, and per-year publication totals were not available in the input data.
- Benchmark figures (PDMS 0.937→0.951; 35-53% collision reduction) are quoted from Evolve's verdicts on the papers' own claims and are not independently verified.