Autonomous racing researcher at TUM: MPC-guided reinforcement learning and edge-deployable in-vehicle intelligence
2
papers here
1
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
2
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-17) 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—
Not yet scored: the underlying publication-volume data for this axis is insufficient in the current snapshot. Output would measure total and per-year publication counts, authorship positions, and venue spread; only early counts (6 works in 2025, 19 in 2026) exist so far.
Impact—
Not yet scored: citation-based impact data is insufficient — the snapshot lists h-index 2 and i10-index 0, too thin to rank. Impact would measure citation counts, h- and i10-index trajectories, and uptake by other research groups.
Focus100/100
Scores 100/100. Topic distribution is tightly concentrated: Autonomous Vehicle Technology and Safety leads with 13 records, followed by Reinforcement Learning in Robotics (6), Vehicle Dynamics and Control Systems (6), Safety Systems Engineering in Autonomy (5) and Robotic Path Planning Algorithms (4) — all within one autonomous-systems cluster, and the field distribution lists a single field, autonomous driving systems / human-robot interaction.
Collaboration50/100
Scores 50/100. The OpenAlex snapshot records 11 distinct co-authors across 25 works in the last three years, a moderate-size, team-driven pattern rather than a solo or very large-network profile. The frequent-coauthor list is led by recurring partners such as Daniel Henel, Frederik Werner and Jasper Hoffmann, with external ties to Sébastien Gros (MPC/optimization) and Davide Scaramuzza (robotics), consistent with the last-author slots on both 2026 catalog papers.
Activity95/100
Scores 95/100. Year-by-year counts rise from 6 works in 2025 to 19 in 2026 — more than triple in a single year — and the snapshot attributes 25 works to the last three years, indicating a very high and accelerating publication cadence, with the two reviewed 2026 entries both appearing as arXiv preprints.
Reproducibility55/100
Scores 55/100, computed from a reproducibility mean of 55 over only one assessed paper, so the signal is thin. Both reviewed 2026 papers carry medium repro-readiness: the open-source, edge-deployable Jarvis framework helps replication, but its evaluation uses synthetic LLM-augmented data; the SG-RL paper reports up to 70.6% fewer samples yet rests on single-seed, simulation-only statistics.
Research summary
Johannes Betz is a researcher at the Technical University of Munich whose OpenAlex-resolved profile centers on Autonomous Vehicle Technology and Safety (13 topic records), with supporting activity in Reinforcement Learning in Robotics (6), Vehicle Dynamics and Control Systems (6), Safety Systems Engineering in Autonomy (5) and Robotic Path Planning Algorithms (4). The citation footprint is still early-stage — h-index 2, i10-index 0 — but output is climbing fast: 6 works in 2025 rising to 19 in 2026, 25 works over the last three years, and 11 distinct co-authors. Two arXiv papers reviewed in the Evolve catalog, both with Betz in the last author position, indicate the current direction of the work: 'Talk to Me, Jarvis' (2026) is an open-source, edge-deployable voice assistant for autonomous racecars built on a fine-tuned Mistral-7B, reporting 97.63% command accuracy at 1.39 s; SG-RL (2026) couples PPO with MPC solver-gradient guidance, reporting up to 70.6% fewer training samples in simulation. Evolve's verdicts flag the shared methodological caveats — synthetic-only evaluation for the assistant, single-seed and simulation-only statistics for the RL work — with medium repro-readiness on both and a reproducibility mean of 55 from one assessed paper. A 95 activity score and a 100 focus score reflect this concentrated, fast-moving profile; the frequent-coauthor list points to a stable core team with external ties to Sébastien Gros and Davide Scaramuzza.
Collaboration pattern
Betz works in compact, recurring teams: 11 distinct co-authors across 25 recent-3y works. The frequent-coauthor list (Daniel Henel, Frederik Werner, Alexander Langmann, Baha Zarrouki, Arslan Thobani, Jasper Hoffmann, Mattia Piccinini, Rudolf Reiter, Felix Jahncke) suggests a stable core group, extended by Sébastien Gros (MPC/optimization) and Davide Scaramuzza (robotics/vision). On both reviewed 2026 arXiv papers he appears as last author without a corresponding-author flag.
Research keywords
Highlights
Open-source edge voice assistant for autonomous racecars
MPC solver gradients guide reinforcement learning
Publication output more than tripled in a single year
OpenAlex counts rise from 6 works in 2025 to 19 in 2026 (25 works over the last three years, 11 distinct co-authors), with both 2026 Evolve-catalog entries posted as arXiv preprints and Betz in the last author position.
Timeline
- 2025OpenAlex records 6 works for Betz this year — the earliest year covered by the available snapshot.
- 2026Output jumps to 19 works, more than triple the prior year; the three-year total reaches 25 works with 11 distinct co-authors.
- 2026'Talk to Me, Jarvis' — an open-source, edge-deployable voice assistant for autonomous racecars — appears on arXiv with Betz as last author.
- 2026SG-RL, coupling PPO with MPC solver-gradient guidance for weights-varying MPC, appears on arXiv with Betz as last author.
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.
- Bibliometric counts come from an OpenAlex-resolved identity; citation coverage may lag or split works across name variants.
- Citation footprint is minimal (h-index 2, i10-index 0), so impact rankings are not meaningful yet.
- Only 2 papers are reviewed by Evolve; the reproducibility mean (55) rests on a single assessed paper.
- Both reviewed papers are 2026 arXiv preprints; peer-review venue status is not reflected in the snapshot.