Pith. sign in

REVIEW 4 major objections 4 minor 63 references

Drive Fast, Learn Faster: On-Board RL for High Performance Autonomous Racing

T0 review · 4 major / 4 minor · reviewed 2026-08-15 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read A reinforcement-learning race car trained only on the physical track beat its classical state-of-the-art baseline by up to 11.5% in lap time.

desk verdict Real on-board residual RL that beats its own classical baselines; the 'SotA' headline is softer than it looks because the baselines are in-house. read the letter →

arxiv 2505.07321 v1 pith:KGTFGMOP submitted 2025-05-12 cs.RO

Benedict Hildisch , Edoardo Ghignone , Nicolas Baumann , Cheng Hu , Andrea Carron , Michele Magno This is my paper · ORCID
classification cs.RO
keywords autonomousracingreinforcementlearningon-boardresidualpolicysoftactor-criticF1TENTHsim-to-realsampleefficiency
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper claims that a reinforcement-learning agent can be trained entirely on a physical race car, with no simulator, and still beat carefully tuned classical controllers. On the F1TENTH 1:10 platform, a residual Soft Actor-Critic policy that adds small corrections on top of a classical Model- and Acceleration-based Pursuit controller reached a 9.26-second minimum lap time versus the baseline's 10.41 seconds on the C-track with Turbo tires, an 11.5% improvement, after about 20 minutes of on-track training. The same architecture improved every classical baseline it was paired with (PP, MAP, FTG), and an end-to-end version trained without any base controller also beat the classical baseline after 82 minutes. The significance is that it removes the simulation-to-reality transfer step that often limits RL in real-time autonomous systems, learning directly from the platform's own dynamics.

What carries the argument

The load-bearing mechanism is the residual policy structure: the RL agent learns only a small additive correction $u_{RL}$ to the base controller's command $u_{base}$, with steering corrections bounded to $\pm 0.15$ rad and speed corrections in $[-0.5, 2]$ m/s, so an untrained agent starts near the base behavior and improves it. Three supporting mechanisms carry the fast on-board training: asynchronous acting/training (the acting policy is synced once per second while training runs at 32 Hz), multi-step TD learning over three steps to handle delayed rewards, and HDRA, which linearly redistributes the crash penalty $p$ over the $N$ previous transitions to discourage the whole sequence leading to failure. A curriculum safety filter with an adaptive heading threshold gates unsafe commands and triggers a recovery state machine. Together these let a physical car collect and learn from its own experience at racing speed without simulation.

What would settle it

Run the same training pipeline on the same physical track against an independently implemented, publicly benchmarked MAP or MPC baseline that has itself been tuned to the track, and compare minimum lap times over 20 laps; if the classical controller matches or beats the trained RL policy, or the improvement disappears against an equally strong baseline, the central claim would be refuted. Replicating the Y-track zero-shot experiment and finding that no seed reliably completes 20 laps would likewise weaken the generalization claim.

Watch

Extended reading notes

Core claim

The discovery is that a deliberately small residual policy can learn the delta between what a classical controller does and what is actually fastest, and that this delta is learnable on-board and in real time. Rather than learning the full control mapping, the agent outputs bounded corrections to steering and velocity that are added to the base controller's commands; the base controller keeps the car safe while the residual policy explores faster lines. With Soft Actor-Critic as the learner, the paper adds three mechanisms: multi-step temporal-difference updates to spread delayed reward signal, an asynchronous acting/training pipeline that decouples data collection at 10 Hz from policy updates at 32 Hz, and a Heuristic Delayed Reward Adjustment that retroactively amplifies penalties on the transitions leading to a crash. The result is consistent improvement: across C-track and Y-track, Turbo and TPU tires, and three random seeds, residual RL MAP improved minimum lap times by 5.4% to 11.5% and mean lap times by 5.9% to 9.0% over its baseline, with speeds above 6.5 m/s.

Load-bearing premise

The claim hinges on the baselines labeled state-of-the-art (MAP and MPC) being strong, fairly tuned representatives of classical racing control; if they are under-tuned or conservative, the measured lap-time win over them overstates the advantage over genuinely best classical controllers.

Editorial extensions

If this is right

  • If the result holds, on-board RL can replace sim-to-real transfer for agile racing robots: the learned controller outperforms the classical baseline it is attached to, so classical controllers can be kept as safety nets while RL refines them in real time.
  • The asynchronous pipeline means a resource-constrained robot without a GPU can train on-board at racing speeds, so the approach should transfer to other real-time platforms with limited compute.
  • The residual structure's consistency across the PP, MAP, and FTG baselines suggests the method is controller-agnostic: any classical controller, even a simple one, can be upgraded by the same learned residual.
  • The end-to-end result, reaching a 9.65-second minimum lap time after 82 minutes with no base controller, shows the same framework can eventually learn racing from scratch, though more slowly.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If this recipe generalizes beyond F1TENTH, the same residual-plus-safety-filter structure should work for drone racing or legged locomotion: start from any stabilizing base policy, train the residual directly on the platform, and let the filter act as a curriculum; the paper does not test this, but nothing in the architecture is platform-specific.
  • The ablation suggests the asynchronous pipeline and multi-step TD are the main drivers of convergence speed, while HDRA mostly reduces variance across seeds, so a user with limited time might drop HDRA without losing much average lap-time performance.
  • The zero-shot transfer results (mixed, with two of the three seeds unstable) imply that the practical recipe is few-shot fine-tuning rather than direct transfer, and an independent comparison against an externally tuned classical racing controller is the natural next test of the 'state-of-the-art' claim.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 4 minor

Summary. The paper presents an on-board reinforcement learning framework for a 1:10 F1TENTH autonomous racing car. The method is a residual Soft Actor-Critic policy that augments a classical base controller (PP, MAP, or FTG), trained entirely on the physical vehicle without simulation pre-training. Additional components include multi-step TD learning, an asynchronous acting/training pipeline, a curriculum safety filter, and a Heuristic Delayed Reward Adjustment (HDRA) mechanism. Experiments on a real car across two tracks and two tire types show that the residual policy consistently improves its base controller, with the RL MAP configuration reducing the minimum lap time by up to 11.5% relative to the MAP baseline and achieving a 9.26 s minimum lap on the C-track with Turbo tires. An end-to-end variant, trained for 82 minutes, also surpasses the MAP baseline. A simulator ablation studies the contribution of the asynchronous pipeline, multi-step TD, and HDRA. The paper claims that this is the fastest F1TENTH algorithm reported to date and that on-board RL can surpass classical state-of-the-art racing controllers without sim-to-real transfer.

Significance. If the empirical claims hold, this is a meaningful advance for physical robot learning in autonomous racing: it demonstrates that a policy trained directly on a real vehicle, with no simulation pre-training and no external GPU, can consistently beat the classical controller it is built on, and can do so within a single battery charge. The paper provides real physical experiments on multiple tracks, two tire compounds, and multiple random seeds, with per-seed tables, training curves, and speed-profile visualizations. The within-paper comparison to the base controllers is convincing and is the core strength of the paper. The claim to surpass the state of the art, however, rests on in-house implementations of the comparison controllers, and the evidence for that external-facing claim is currently incomplete. The simulator ablation is a useful addition, though it is limited to one setting and does not establish a statistically significant benefit for HDRA.

major comments (4)
  1. [Section 4.2, Table 1, Contribution 2] The state-of-the-art comparison is not established. The MAP baseline is from the authors' own group (Becker et al. 2023), the MPC baseline is an in-house implementation of Vázquez et al. (2020) that Appendix J admits is conservative and 'induced unnecessary conservativity' on high-grip tires, and no quantitative comparison is provided to independent on-board RL systems such as RLPP (Ghignone et al. 2025) or FastRLAP (Stachowicz et al. 2023). The abstract and Contribution 2 claim the fastest F1TENTH algorithm, but the 'up to 11.5% vs SotA' headline measures the gap to these in-house baselines. Please either provide external comparisons on comparable tracks or carefully rephrase the SotA claims to refer to the provided baseline controllers.
  2. [Section 5.1, Table 1] The statistical evidence for repeatability and robustness is thin. Only three random seeds are used, no significance tests or confidence intervals are reported, and the lap-time statistics in Table 1 are computed only over the 20 violation-free deployment laps, while boundary violations during deployment (up to 2 on C-Track Turbo, 1 on C-Track TPU, and 5 on Y-Track) are excluded from those statistics. The claim that the method is robust across seeds and environments would be substantially strengthened by reporting per-seed pairwise comparisons, effect sizes, or confidence intervals, and by stating the sensitivity of the reported lap times to the violation-exclusion rule.
  3. [Section 5.2, Table 4] The zero-shot generalization claim is overstated. The text states that zero-shot results were mixed and that Seeds 2 and 3 experienced significant instability, and Table 4 shows that Seed 2 required 14 boundary violations and Seed 3 required 7 before completing 20 violation-free laps, with 22 violations in the combined deployment. The conclusion that 'a single policy can be tuned to drive consistently on a different track' is supported by the few-shot results, not by the zero-shot results. The section should be reworded to distinguish the two settings clearly and to present zero-shot transfer as a limitation rather than as demonstrated capability.
  4. [Section 5.3, Figure 5, Table 5] The ablation does not support HDRA as a core algorithmic improvement. Table 5 shows that HDRA+TD3 Async achieves an average lap time of 9.53 s versus 9.68 s for TD3 Async, and the text itself says this is not a statistically relevant improvement, with the effect appearing mainly as reduced lap-time spread. This is acceptable for an engineering paper, but the contribution list and abstract should present HDRA as a variance-reduction heuristic rather than as one of the main reasons for the performance gain, or the authors should provide statistical evidence for a benefit.
minor comments (4)
  1. [Abstract and Section 5.1] The abstract states 'up to an 11.5% reduction ... with only 20 min of training,' but the 11.5% improvement appears to refer to the Y-Track from-scratch comparison in Table 1, which used 21–25 minutes of training, while the 20-minute C-Track Turbo result is about 11.05%. Please make the training-time and scenario attribution of the headline number explicit.
  2. [Figure 2] The legend for the RL MAP panel repeats 'RL MAP' for the training curve and for what appears to be the deployed policy or baseline reference; please clarify the legend labels and use distinct names for the training and deployment curves.
  3. [Table 6 and Section 3] The paper relies on several hand-chosen hyperparameters (reward multiplier lambda, penalty p, HDRA steps N, safety filter threshold range, residual action limits, curriculum speed increment) without sensitivity analysis. A short discussion of which parameters are critical and how sensitive performance is to them would improve reproducibility.
  4. [Appendix G and Section 3.6] The pseudocode in Algorithm 1 has an indentation ambiguity in the condition 'if terminal_state is True and replay_buffer.reward[current_step] = 0', and the relation between the pseudocode variable 'adjustment_steps N' and Eq. (1) should be stated explicitly. Also, there is a minor typo in the abstract: '20 minof training' should be '20 min of training'.

Circularity Check

0 steps flagged · score 2.0 of 10

No circularity in the derivation chain; lap times are physically measured, but the 'SotA' label relies on the authors' own MAP baseline, a benchmark-selection concern rather than a circular derivation.

full rationale

The paper's central results are empirical: RL MAP achieves 9.26 s versus 10.41 s for MAP on the C-track, and these lap times are measured on the physical F1TENTH platform, not computed from the method's assumptions. The residual policy u = u_base + u_RL is trained with SAC and evaluated by wall-clock lap times; no fitted parameter is renamed as a prediction, and no theoretical derivation is claimed for the measured improvements. The only load-bearing self-reference is the choice of 'state-of-the-art' baselines: MAP is the authors' own controller and the MPC is manually tuned in-house, so the headline '11.5% vs SotA' is only as strong as those implementations. This is a benchmark-fairness limitation, not a circularity by construction: the residual improvement over each provided base controller is independently measured and would stand even if the SotA label were withdrawn. The ablation study is similarly empirical. No uniqueness theorem, ansatz-as-citation, or definitional equivalence is present. Score 2 reflects the minor self-citation in the baseline selection without any circular derivation step.

Assumptions & free parameters 7 free parameters · 5 assumptions · 0 invented entities

The method introduces algorithmic heuristics (HDRA, adaptive safety filter, async training) but no new physical entities, forces, or dimensions. The free parameters are all hyperparameters or safety-curriculum choices; none are fitted to the reported deployment lap times. The main hidden assumptions are about state-estimation accuracy, stationarity of the platform, the racing objective, simulator representativeness, and the fairness of the SotA baselines.

free parameters (7)
  • Reward progress multiplier lambda = 10
    Hand-chosen in Section 3.5 and Table 6; scales the progress reward and affects training speed and final risk-taking.
  • Terminal penalty p = 10
    Hand-chosen in Table 6; used both for terminal reward and as the base penalty propagated by HDRA.
  • HDRA adjustment steps N = 10
    Chosen by hand in Table 6; controls how far the retroactive penalty reaches into the replay buffer.
  • Safety filter threshold range and increment = psi_filter in [pi/6, pi/2], epsilon = 0.05 rad
    Adaptive curriculum bounds and step size chosen by hand in Section 3.2 and Table 6; they determine how much off-heading exploration is allowed.
  • SAC hyperparameters = lr=0.003, gamma=0.96, batch=256, hidden=256, replay=1e6, TD steps=3
    Selected in Table 6 and partly motivated by prior work; these affect sample efficiency and stability but are not fitted to the reported lap times.
  • Residual action limits = delta_RL in [-0.15,0.15] rad, v_RL in [-0.5,2] m/s
    Hand-chosen in Section 3.3; they bound how much the RL agent can deviate from the base controller and shape the training regime.
  • E2E curriculum speed increment = alpha increases by 0.5 m/s after 3 clean laps
    Hand-chosen for the end-to-end experiments in Section 4.2; affects how quickly the E2E agent reaches high speeds.
assumptions (5)
  • domain assumption SLAM-based state estimates (velocities, yaw rate, lateral deviation) are accurate enough for both control and lap-time measurement.
    The observation space and evaluation metrics rely on onboard localization estimates; no ground-truth validation is provided. Introduced in Section 3.
  • domain assumption The F1TENTH vehicle dynamics are stationary enough within one battery and across the reported deployment runs.
    Training and deployment assume tire wear, battery state, and temperature do not materially change the learned policy's validity. Stated implicitly in Section 4.
  • domain assumption Progress along the reference line and track-boundary violations define the correct racing objective.
    The reward uses lambda * delta_s and terminal penalties; this assumes the reference raceline is a good objective and that boundary violations are the right failure criterion. Sections 3.5 and 4.
  • domain assumption The F1TENTH simulator used for ablations is representative enough to choose between architectural components.
    The ablation study in Section 5.3 is run in simulation, not on the physical car; the conclusions about async training, multi-step TD, and HDRA assume sim results transfer to real training.
  • domain assumption MAP and MPC baselines are representative of the state of the art in F1TENTH racing.
    The SotA comparison in Table 1 uses in-house implementations of MAP and MPC. No independent benchmark or direct prior-RL lap-time table is included, so the representativeness of these baselines is assumed. Section 4.2.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Drive Fast, Learn Faster: On-Board RL for High Performance Autonomous Racing." pith.science (2026). https://pith.science/paper/KGTFGMOP

@misc{pith2026250507321,
  author       = {Pith},
  title        = {Pith review of: Drive Fast, Learn Faster: On-Board RL for High Performance Autonomous Racing},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/KGTFGMOP}},
  note         = {Machine review of arXiv:2505.07321}
}
read the original abstract

Autonomous racing presents unique challenges due to its non-linear dynamics, the high speed involved, and the critical need for real-time decision-making under dynamic and unpredictable conditions. Most traditional Reinforcement Learning (RL) approaches rely on extensive simulation-based pre-training, which faces crucial challenges in transfer effectively to real-world environments. This paper introduces a robust on-board RL framework for autonomous racing, designed to eliminate the dependency on simulation-based pre-training enabling direct real-world adaptation. The proposed system introduces a refined Soft Actor-Critic (SAC) algorithm, leveraging a residual RL structure to enhance classical controllers in real-time by integrating multi-step Temporal-Difference (TD) learning, an asynchronous training pipeline, and Heuristic Delayed Reward Adjustment (HDRA) to improve sample efficiency and training stability. The framework is validated through extensive experiments on the F1TENTH racing platform, where the residual RL controller consistently outperforms the baseline controllers and achieves up to an 11.5 % reduction in lap times compared to the State-of-the-Art (SotA) with only 20 min of training. Additionally, an End-to-End (E2E) RL controller trained without a baseline controller surpasses the previous best results with sustained on-track learning. These findings position the framework as a robust solution for high-performance autonomous racing and a promising direction for other real-time, dynamic autonomous systems.

Figures

Figures reproduced from arXiv: 2505.07321 by the authors.

Figure 1
Figure 1. Overview of the proposed RL architecture, which processes observations Ocar, provid￾ing proprioceptive information such as position and velocity (vx, vy), and Otrack, describing track boundaries and a given reference trajectory to generate a residual control command uRL. Such a command is additively combined with the output of a conventional controller ubase. Since the agent starts untrained and may encounter unsafe… view at source ↗
Figure 2
Figure 2. Lap Time vs. Wall Time during training for different residual controllers, with respective [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figure 3
Figure 3. Comparison of the driven trajectories, speed profiles, and the time delta for [PITH_FULL_IMAGE:figures/full_fig_p010_3.png] view at source ↗
Figures from the paper (13 more)
Figure 4
Figure 4. Figure 4: Stacked histogram of lap times achieved during zero-shot and few-shot transfer from the [PITH_FULL_IMAGE:figures/full_fig_p010_4.png]
Figure 5
Figure 5. Figure 5: Lap times during training averaged over 5 runs for ablation regarding multi-step TD over one (TD1) and three (TD3) steps, a synchronous (sync) and asynchronous (async) architecture as well as with HDRA and without. Lines only start when the first collision-free lap is …
Figure 6
Figure 6. Figure 6: Image of the F1TENTH car used for evaluation equipped with the Intel NUC 10 computer [PITH_FULL_IMAGE:figures/full_fig_p020_6.png]
Figure 7
Figure 7. Figure 7: Image of C-track with an overlay of the car state observation Ocar = [vx, vy, ψ,˙ ∆d, ∆ψ, δbase, vbase, δRL, prev, vRL, prev] T , denoted as shortened state vector in purple. The observation of the track information Otrack = [p 0:J ref , p 0:J left , p 0:J right] T is …
Figure 8
Figure 8. Figure 8: Images of the two tire types used during the experiments. [PITH_FULL_IMAGE:figures/full_fig_p021_8.png]
Figure 9
Figure 9. Figure 9: Visualizations of the trajectory and speed profile of the [PITH_FULL_IMAGE:figures/full_fig_p023_9.png]
Figure 10
Figure 10. Figure 10: Visualizations of the trajectory and speed profile of the [PITH_FULL_IMAGE:figures/full_fig_p024_10.png]
Figure 11
Figure 11. Figure 11: Histogram of lap times of the E2E controller after 29 min and 82 min . Orange lines represent the MAP baseline (Min. Lap Time: dashed, Avg. Lap Time: solid). 23 [PITH_FULL_IMAGE:figures/full_fig_p024_11.png]
Figure 12
Figure 12. Figure 12: Stacked histogram of lap times across the different tire-track combinations and weight [PITH_FULL_IMAGE:figures/full_fig_p025_12.png]
Figure 13
Figure 13. Figure 13: Lap Time vs. Wall Time during training for [PITH_FULL_IMAGE:figures/full_fig_p026_13.png]
Figure 14
Figure 14. Figure 14: Visualizations of the trajectory, speed profile and time delta of [PITH_FULL_IMAGE:figures/full_fig_p027_14.png]
Figure 15
Figure 15. Figure 15: Visualizations of the trajectory, speed profile and time delta of [PITH_FULL_IMAGE:figures/full_fig_p028_15.png]
Figure 16
Figure 16. Figure 16: Visualizations of the trajectories during zero-shot transfer from [PITH_FULL_IMAGE:figures/full_fig_p030_16.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

63 extracted references · 31 canonical work pages

  1. [1]

    Hindsight Experience Replay

    Marcin Andrychowicz, Filip Wolski, Alex Ray, Jonas Schneider, Rachel Fong, Peter Welinder, Bob McGrew, Josh Tobin, OpenAI Pieter Abbeel, and Wojciech Zaremba. Hindsight Experience Replay . In Advances in Neural Information Processing Systems, volume 30. Curran Associates, Inc., 2017

    work page 2017

  2. [2]

    Hoffman, David Budden, Will Dabney, Dan Horgan, Dhruva TB, Alistair Muldal, Nicolas Heess, and Timothy Lillicrap

    Gabriel Barth-Maron, Matthew W. Hoffman, David Budden, Will Dabney, Dan Horgan, Dhruva TB, Alistair Muldal, Nicolas Heess, and Timothy Lillicrap. Distributional policy gradients. In International Conference on Learning Representations, 2018. URL https://openreview.net/forum?id=SyZipzbCb

    work page 2018

  3. [3]

    ForzaETH Race Stack -- Scaled Autonomous Head-to-Head Racing on Fully Commercial off-the-Shelf Hardware

    Nicolas Baumann, Edoardo Ghignone, Jonas Kühne, Niklas Bastuck, Jonathan Becker, Nadine Imholz, Tobias Kränzlin, Tian Yi Lim, Michael Lötscher, Luca Schwarzenbach, Luca Tognoni, Christian Vogt, Andrea Carron, and Michele Magno. ForzaETH Race Stack -- Scaled Autonomous Head-to-Head Racing on Fully Commercial off-the-Shelf Hardware . Journal of Field Roboti...

    doi:10.1002/rob.22429 2024

  4. [4]

    Model- and Acceleration-based Pursuit Controller for High-Performance Autonomous Racing

    Jonathan Becker, Nadine Imholz, Luca Schwarzenbach, Edoardo Ghignone, Nicolas Baumann, and Michele Magno. Model- and Acceleration-based Pursuit Controller for High-Performance Autonomous Racing . In 2023 IEEE International Conference on Robotics and Automation (ICRA), pp.\ 5276--5283, 2023. doi:10.1109/ICRA48891.2023.10161472

    arXiv 2023

  5. [5]

    Curriculum learning

    Yoshua Bengio, J \'e r \^o me Louradour, Ronan Collobert, and Jason Weston. Curriculum learning . In Proceedings of the 26th annual international conference on machine learning, pp.\ 41--48, 2009

    work page 2009

  6. [6]

    Tum autonomous motorsport: An autonomous racing software for the indy autonomous challenge

    Johannes Betz, Tobias Betz, Felix Fent, Maximilian Geisslinger, Alexander Heilmeier, Leonhard Hermansdorfer, Thomas Herrmann, Sebastian Huch, Phillip Karle, Markus Lienkamp, Boris Lohmann, Felix Nobis, Levent Ögretmen, Matthias Rowold, Florian Sauerbeck, Tim Stahl, Rainer Trauth, Frederik Werner, and Alexander Wischnewski. Tum autonomous motorsport: An au...

    2023

  7. [7]

    Train in Austria, Race in Montecarlo: Generalized RL for Cross-Track F1TENTH LIDAR-Based Races

    Michael Bosello, Rita Tse, and Giovanni Pau. Train in Austria, Race in Montecarlo: Generalized RL for Cross-Track F1TENTH LIDAR-Based Races . In 2022 IEEE 19th Annual Consumer Communications & Networking Conference (CCNC), pp.\ 290--298, 2022. doi:10.1109/CCNC49033.2022.9700730

    arXiv 2022

  8. [8]

    Automatic Vehicle Guidance

    Alberto Broggi, Massimo Bertozzi, Alessandra Fascioli, and Gianni Conte. Automatic Vehicle Guidance . WORLD SCIENTIFIC, 1999. doi:10.1142/3986

    work page doi:10.1142/3986 1999

Show all 63 references
  1. [10]

    Learning from simulation, racing in reality

    Eugenio Chisari, Alexander Liniger, Alisa Rupenyan, Luc Van Gool, and John Lygeros. Learning from simulation, racing in reality . In 2021 IEEE International Conference on Robotics and Automation (ICRA), pp.\ 8046--8052. IEEE, 2021

    2021

  2. [11]

    Implementation of the Pure Pursuit Path Tracking Algorithm

    RC Coulter. Implementation of the Pure Pursuit Path Tracking Algorithm . DTIC Document, 1992

    1992

  3. [12]

    Pilco: a model-based and data-efficient approach to policy search

    Marc Peter Deisenroth and Carl Edward Rasmussen. Pilco: a model-based and data-efficient approach to policy search. In Proceedings of the 28th International Conference on International Conference on Machine Learning, ICML'11, pp.\ 465–472, Madison, WI, USA, 2011. Omnipress. IS...

    2011

  4. [13]

    Learning-based on-track system identification for scaled autonomous racing in under a minute

    Onur Dikici, Edoardo Ghignone, Cheng Hu, Nicolas Baumann, Lei Xie, Andrea Carron, Michele Magno, and Matteo Corno. Learning-based on-track system identification for scaled autonomous racing in under a minute. IEEE Robotics and Automation Letters, 10 0 (2): 0 1984--1991, 2025. ...

    1984

  5. [14]

    Evans, Hendrik W

    Benjamin D. Evans, Hendrik W. Jordaan, and Herman A. Engelbrecht. Safe reinforcement learning for high-speed autonomous racing . Cognitive Robotics, 3: 0 107--126, 2023 a . ISSN 2667-2413. doi:https://doi.org/10.1016/j.cogr.2023.04.002

    2023 doi

  6. [15]

    Engelbrecht, Rahul Mangharam, and Hendrik W

    Benjamin David Evans, Johannes Betz, Hongrui Zheng, Herman A. Engelbrecht, Rahul Mangharam, and Hendrik W. Jordaan. Bypassing the Simulation-to-Reality Gap: Online Reinforcement Learning Using a Supervisor . In 2023 21st International Conference on Advanced Robotics (ICAR), pp...

    2023

  7. [16]

    High-Speed Autonomous Racing Using Trajectory-Aided Deep Reinforcement Learning

    Benjamin David Evans, Herman Arnold Engelbrecht, and Hendrik Willem Jordaan. High-Speed Autonomous Racing Using Trajectory-Aided Deep Reinforcement Learning . IEEE Robotics and Automation Letters, 8 0 (9): 0 5353--5359, 2023 c . doi:10.1109/LRA.2023.3295252

    2023

  8. [17]

    Comparing deep reinforcement learning architectures for autonomous racing

    Benjamin David Evans, Hendrik Willem Jordaan, and Herman Arnold Engelbrecht. Comparing deep reinforcement learning architectures for autonomous racing . Machine Learning with Applications, 14: 0 100496, 2023 d . ISSN 2666-8270. doi:https://doi.org/10.1016/j.mlwa.2023.100496

    2023

  9. [18]

    Super-Human Performance in Gran Turismo Sport Using Deep Reinforcement Learning

    Florian Fuchs, Yunlong Song, Elia Kaufmann, Davide Scaramuzza, and Peter Durr. Super-Human Performance in Gran Turismo Sport Using Deep Reinforcement Learning . IEEE Robotics and Automation Letters, 6: 0 4257--4264, 7 2021. ISSN 23773766. doi:10.1109/LRA.2021.3064284

    2021

  10. [19]

    Tc-driver: A trajectory-conditioned reinforcement learning approach to zero-shot autonomous racing

    Edoardo Ghignone, Nicolas Baumann, and Michele Magno. Tc-driver: A trajectory-conditioned reinforcement learning approach to zero-shot autonomous racing. IEEE Transactions on Field Robotics, 1: 0 527--536, 2024. doi:10.1109/TFR.2024.3499912

    2024

  11. [20]

    Rlpp: A residual method for zero-shot real-world autonomous racing on scaled platforms, 2025

    Edoardo Ghignone, Nicolas Baumann, Cheng Hu, Jonathan Wang, Lei Xie, Andrea Carron, and Michele Magno. Rlpp: A residual method for zero-shot real-world autonomous racing on scaled platforms, 2025. URL https://arxiv.org/abs/2501.17311

    2025 arXiv

  12. [21]

    Region-Based Convolutional Networks for Accurate Object Detection and Segmentation

    Ross Girshick, Jeff Donahue, Trevor Darrell, and Jitendra Malik. Region-Based Convolutional Networks for Accurate Object Detection and Segmentation . IEEE Transactions on Pattern Analysis and Machine Intelligence, 38 0 (1): 0 142--158, 2016. doi:10.1109/TPAMI.2015.2437384

    2016

  13. [22]

    A survey of deep learning techniques for autonomous driving

    Sorin Grigorescu, Bogdan Trasnea, Tiberiu Cocias, and Gigel Macesanu. A survey of deep learning techniques for autonomous driving . Journal of Field Robotics, 37 0 (3): 0 362--386, 2020. doi:https://doi.org/10.1002/rob.21918

    2020 doi

  14. [23]

    Soft Actor-Critic: Off-Policy Maximum Entropy Deep Reinforcement Learning with a Stochastic Actor

    Tuomas Haarnoja, Aurick Zhou, Pieter Abbeel, and Sergey Levine. Soft Actor-Critic: Off-Policy Maximum Entropy Deep Reinforcement Learning with a Stochastic Actor . In Jennifer Dy and Andreas Krause (eds.), Proceedings of the 35th International Conference on Machine Learning, v...

    2018

  15. [24]

    Lillicrap

    Danijar Hafner, Jurgis Pasukonis, Jimmy Ba, and Timothy P. Lillicrap. Mastering diverse domains through world models. CoRR, abs/2301.04104, 2023. URL https://doi.org/10.48550/arXiv.2301.04104

    2023 doi arXiv

  16. [25]

    Minimum curvature trajectory planning and control for an autonomous race car

    Alexander Heilmeier, Alexander Wischnewski, Leonhard Hermansdorfer, Johannes Betz, Markus Lienkamp, and Boris Lohmann. Minimum curvature trajectory planning and control for an autonomous race car . Vehicle System Dynamics, 58 0 (10): 0 1497--1527, 10 2020. ISSN 0042-3114. doi:...

    2020

  17. [26]

    Stable Baselines

    Ashley Hill, Antonin Raffin, Maximilian Ernestus, Adam Gleave, Anssi Kanervisto, Rene Traore, Prafulla Dhariwal, Christopher Hesse, Oleg Klimov, Alex Nichol, Matthias Plappert, Alec Radford, John Schulman, Szymon Sidor, and Yuhuai Wu. Stable Baselines . https://github.com/hill...

    2018

  18. [27]

    Anymal parkour: Learning agile navigation for quadrupedal robots

    David Hoeller, Nikita Rudin, Dhionis Sako, and Marco Hutter. Anymal parkour: Learning agile navigation for quadrupedal robots. Science Robotics, 9 0 (88): 0 eadi7566, 2024. doi:10.1126/scirobotics.adi7566. URL https://www.science.org/doi/abs/10.1126/scirobotics.adi7566

    2024 doi

  19. [28]

    When to trust your model: Model-based policy optimization

    Michael Janner, Justin Fu, Marvin Zhang, and Sergey Levine. When to trust your model: Model-based policy optimization. In H. Wallach, H. Larochelle, A. Beygelzimer, F. d Alch\' e -Buc, E. Fox, and R. Garnett (eds.), Advances in Neural Information Processing Systems, volume 32....

    2019

  20. [29]

    Residual reinforcement learning for robot control

    Tobias Johannink, Shikhar Bahl, Ashvin Nair, Jianlan Luo, Avinash Kumar, Matthias Loskyll, Juan Aparicio Ojea, Eugen Solowjow, and Sergey Levine. Residual reinforcement learning for robot control . In 2019 international conference on robotics and automation (ICRA), pp.\ 6023--...

    2019

  21. [30]

    Champion-level drone racing using deep reinforcement learning

    Elia Kaufmann, Leonard Bauersfeld, Antonio Loquercio, Matthias M \"u ller, Vladlen Koltun, and Davide Scaramuzza. Champion-level drone racing using deep reinforcement learning . Nature, 620 0 (7976): 0 982--987, 2023

    2023

  22. [31]

    Lang, Sourabh Vora, Holger Caesar, Lubing Zhou, Jiong Yang, and Oscar Beijbom

    Alex H. Lang, Sourabh Vora, Holger Caesar, Lubing Zhou, Jiong Yang, and Oscar Beijbom. PointPillars: Fast Encoders for Object Detection From Point Clouds . In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), June 2019

    2019

  23. [32]

    Wei Liu, Dragomir Anguelov, Dumitru Erhan, Christian Szegedy, Scott Reed, Cheng-Yang Fu, and Alexander C. Berg. SSD: Single Shot MultiBox Detector . In Computer Vision -- ECCV 2016, pp.\ 21--37, Cham, 2016. Springer International Publishing. ISBN 978-3-319-46448-0

    2016

  24. [33]

    Ng, Daishi Harada, and Stuart J

    Andrew Y. Ng, Daishi Harada, and Stuart J. Russell. Policy invariance under reward transformations: Theory and application to reward shaping. In Proceedings of the Sixteenth International Conference on Machine Learning, ICML '99, pp.\ 278–287, San Francisco, CA, USA, 1999. Mor...

    1999

  25. [34]

    F1/10: An Open-Source Autonomous Cyber-Physical Platform

    Matthew O'Kelly, Varundev Sukhil, Houssam Abbas, Jack Harkins, Chris Kao, Yash Vardhan Pant, Rahul Mangharam, Dipshil Agarwal, Madhur Behl, Paolo Burgio, and Marko Bertogna. F1/10: An Open-Source Autonomous Cyber-Physical Platform . ArXiv, abs/1901.08567, 2019. URL https://api...

    1901 arXiv

  26. [35]

    F1TENTH: An Open-source Evaluation Environment for Continuous Control and Reinforcement Learning

    Matthew O'Kelly, Hongrui Zheng, Dhruv Karthik, and Rahul Mangharam. F1TENTH: An Open-source Evaluation Environment for Continuous Control and Reinforcement Learning . In Proceedings of the NeurIPS 2019 Competition and Demonstration Track, volume 123 of Proceedings of Machine L...

    2019

  27. [36]

    Hans B. Pacejka. Chapter 1 - tire characteristics and vehicle handling and stability. In Hans B. Pacejka (ed.), Tire and Vehicle Dynamics (Third Edition), pp.\ 1--58. Butterworth-Heinemann, Oxford, third edition edition, 2012. ISBN 978-0-08-097016-5. doi:https://doi.org/10.101...

    2012 doi

  28. [37]

    Imitation learning for agile autonomous driving

    Yunpeng Pan, Ching-An Cheng, Kamil Saigol, Keuntaek Lee, Xinyan Yan, Evangelos A Theodorou, and Byron Boots. Imitation learning for agile autonomous driving . The International Journal of Robotics Research, 39 0 (2-3): 0 286--302, 2020

    2020

  29. [38]

    Sim-to-real transfer of robotic control with dynamics randomization

    Xue Bin Peng, Marcin Andrychowicz, Wojciech Zaremba, and Pieter Abbeel. Sim-to-real transfer of robotic control with dynamics randomization . In 2018 IEEE international conference on robotics and automation (ICRA), pp.\ 3803--3810. IEEE, 2018

    2018

  30. [39]

    You Only Look Once: Unified, Real-Time Object Detection

    Joseph Redmon, Santosh Divvala, Ross Girshick, and Ali Farhadi. You Only Look Once: Unified, Real-Time Object Detection . In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), June 2016

    2016

  31. [40]

    Bridging the sim-to-real gap with bayesian inference

    Jonas Rothfuss, Bhavya Sukhija, Lenart Treven, Florian Dörfler, Stelian Coros, and Andreas Krause. Bridging the sim-to-real gap with bayesian inference. In 2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pp.\ 10784--10791, 2024. doi:10.1109/IRO...

    2024

  32. [41]

    Learning to walk in minutes using massively parallel deep reinforcement learning

    Nikita Rudin, David Hoeller, Philipp Reist, and Marco Hutter. Learning to walk in minutes using massively parallel deep reinforcement learning. In 5th Annual Conference on Robot Learning, 2021. URL https://openreview.net/forum?id=wK2fDDJ5VcF

    2021

  33. [42]

    Follow the Gap Method

    Volkan Sezer and Metin Gokasan. A novel obstacle avoidance algorithm: "Follow the Gap Method" . Robotics and Autonomous Systems, 60 0 (9): 0 1123–1134, September 2012. ISSN 0921-8890. doi:10.1016/j.robot.2012.05.021

    2012 doi

  34. [43]

    Complex-YOLO: An Euler-Region-Proposal for Real-time 3D Object Detection on Point Clouds

    Martin Simony, Stefan Milzy, Karl Amendey, and Horst-Michael Gross. Complex-YOLO: An Euler-Region-Proposal for Real-time 3D Object Detection on Point Clouds . In Proceedings of the European Conference on Computer Vision (ECCV) Workshops, September 2018

    2018

  35. [44]

    Reaching the limit in autonomous racing: Optimal control versus reinforcement learning

    Yunlong Song, Angel Romero, Matthias Müller, Vladlen Koltun, and Davide Scaramuzza. Reaching the limit in autonomous racing: Optimal control versus reinforcement learning. Science Robotics, 8 0 (82): 0 eadg1462, 2023. doi:10.1126/scirobotics.adg1462. URL https://www.science.or...

    2023 doi

  36. [45]

    Hong, and Sehoon Hal

    Nitish Sontakke, Hosik Chae, Sangjoon Lee, Tianle Huang, Dennis W. Hong, and Sehoon Hal. Residual physics learning and system identification for sim-to-real transfer of policies on buoyancy assisted legged robots. In 2023 IEEE/RSJ International Conference on Intelligent Robots...

    2023

  37. [46]

    FastRLAP: A System for Learning High-Speed Driving via Deep RL and Autonomous Practicing

    Kyle Stachowicz, Dhruv Shah, Arjun Bhorkar, Ilya Kostrikov, and Sergey Levine. FastRLAP: A System for Learning High-Speed Driving via Deep RL and Autonomous Practicing . In Proceedings of The 7th Conference on Robot Learning, volume 229 of Proceedings of Machine Learning Resea...

    2023

  38. [47]

    Sutton and Andrew G

    Richard S. Sutton and Andrew G. Barto. Reinforcement Learning: An Introduction. The MIT Press, second edition, 2018. URL http://incompleteideas.net/book/the-book-2nd.html

    2018

  39. [48]

    Wabersich, Andrea Carron, and Melanie N

    Ben Tearle, Kim P. Wabersich, Andrea Carron, and Melanie N. Zeilinger. A Predictive Safety Filter for Learning-Based Racing Control . IEEE Robotics and Automation Letters, 6 0 (4): 0 7635--7642, 2021. doi:10.1109/LRA.2021.3097073

    2021

  40. [49]

    Thorpe, M.H

    C. Thorpe, M.H. Hebert, T. Kanade, and S.A. Shafer. Vision and navigation for the Carnegie-Mellon Navlab . IEEE Transactions on Pattern Analysis and Machine Intelligence, 10 0 (3): 0 362--373, 1988. doi:10.1109/34.3900

    1988 doi

  41. [50]

    Stanley: The robot that won the DARPA Grand Challenge

    Sebastian Thrun, Mike Montemerlo, Hendrik Dahlkamp, David Stavens, Andrei Aron, James Diebel, Philip Fong, John Gale, Morgan Halpenny, Gabriel Hoffmann, Kenny Lau, Celia Oakley, Mark Palatucci, Vaughan Pratt, Pascal Stang, Sven Strohband, Cedric Dupont, Lars-Erik Jendrossek, C...

    2006

  42. [51]

    Domain randomization for transferring deep neural networks from simulation to the real world

    Josh Tobin, Rachel Fong, Alex Ray, Jonas Schneider, Wojciech Zaremba, and Pieter Abbeel. Domain randomization for transferring deep neural networks from simulation to the real world . 2017 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pp.\ 23--30, 2017

    2017

  43. [52]

    Residual Policy Learning for Vehicle Control of Autonomous Racing Cars

    Raphael Trumpp, Denis Hoornaert, and Marco Caccamo. Residual Policy Learning for Vehicle Control of Autonomous Racing Cars . In 2023 IEEE Intelligent Vehicles Symposium (IV), pp.\ 1--6, 2023. doi:10.1109/IV55152.2023.10186744

    2023

  44. [53]

    Wurman, and Peter Stone

    Miguel Vasco, Takuma Seno, Kenta Kawamoto, Kaushik Subramanian, Peter R. Wurman, and Peter Stone. A Super-human Vision-based Reinforcement Learning Agent for Autonomous Racing in Gran Turismo . Reinforcement Learning Journal, 4: 0 1674--1710, 2024

    2024

  45. [54]

    acados -- a modular open-source framework for fast embedded optimal control

    Robin Verschueren, Gianluca Frison, Dimitris Kouzoupis, Jonathan Frey, Niels van Duijkeren, Andrea Zanelli, Branimir Novoselnik, Thivaharan Albin, Rien Quirynen, and Moritz Diehl. acados -- a modular open-source framework for fast embedded optimal control. Mathematical Program...

    2021

  46. [55]

    Vázquez, Marius Brühlmeier, Alexander Liniger, Alisa Rupenyan, and John Lygeros

    José L. Vázquez, Marius Brühlmeier, Alexander Liniger, Alisa Rupenyan, and John Lygeros. Optimization-based hierarchical motion planning for autonomous racing. In 2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pp.\ 2397--2403, 2020. doi:10.110...

    2020

  47. [56]

    A Survey on Curriculum Learning

    Xin Wang, Yudong Chen, and Wenwu Zhu. A Survey on Curriculum Learning . IEEE Transactions on Pattern Analysis and Machine Intelligence, 44 0 (9): 0 4555--4576, 2022. doi:10.1109/TPAMI.2021.3069908

    2022

  48. [57]

    DayDreamer: World Models for Physical Robot Learning

    Philipp Wu, Alejandro Escontrela, Danijar Hafner, Pieter Abbeel, and Ken Goldberg. DayDreamer: World Models for Physical Robot Learning . In Karen Liu, Dana Kulic, and Jeff Ichnowski (eds.), Proceedings of The 6th Conference on Robot Learning, volume 205 of Proceedings of Mach...

    2023

  49. [58]

    Dolan, and Guanya Shi

    Wenli Xiao, Haoru Xue, Tony Tao, Dvij Kalaria, John M. Dolan, and Guanya Shi. Anycar to anywhere: Learning universal dynamics model for agile and adaptive mobility, 2024. URL https://arxiv.org/abs/2409.15783

    2024 arXiv

  50. [59]

    Rupam Mahmood

    Yufeng Yuan and A. Rupam Mahmood. Asynchronous Reinforcement Learning for Real-Time Control of Physical Robots . In 2022 International Conference on Robotics and Automation (ICRA), pp.\ 5546--5552, 2022. doi:10.1109/ICRA46639.2022.9811771

    2022

  51. [60]

    PolarNet: An Improved Grid Representation for Online LiDAR Point Clouds Semantic Segmentation

    Yang Zhang, Zixiang Zhou, Philip David, Xiangyu Yue, Zerong Xi, Boqing Gong, and Hassan Foroosh. PolarNet: An Improved Grid Representation for Online LiDAR Point Clouds Semantic Segmentation . In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition...

    2020

  52. [61]

    Jingyuan Zhao, Wenyi Zhao, Bo Deng, Zhenghong Wang, Feng Zhang, Wenxiang Zheng, Wanke Cao, Jinrui Nan, Yubo Lian, and Andrew F. Burke. Autonomous driving system: A comprehensive survey . Expert Systems with Applications, 242: 0 122836, 2024. ISSN 0957-4174. doi:https://doi.org...

    2024

  53. [62]

    Sim-to-Real Transfer in Deep Reinforcement Learning for Robotics: a Survey

    Wenshuai Zhao, Jorge Peña Queralta, and Tomi Westerlund. Sim-to-Real Transfer in Deep Reinforcement Learning for Robotics: a Survey . In 2020 IEEE Symposium Series on Computational Intelligence (SSCI), pp.\ 737--744, 2020. doi:10.1109/SSCI47803.2020.9308468

    2020

  54. [63]

    A Comparative Analysis of LiDAR SLAM-Based Indoor Navigation for Autonomous Vehicles

    Qin Zou, Qin Sun, Long Chen, Bu Nie, and Qingquan Li. A Comparative Analysis of LiDAR SLAM-Based Indoor Navigation for Autonomous Vehicles . IEEE Transactions on Intelligent Transportation Systems, 23 0 (7): 0 6907--6921, 2022. doi:10.1109/TITS.2021.3063477

    2022

  55. [64]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 gl...

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.