pith. sign in

arxiv: 2606.08136 · v1 · pith:GQ33Y4KJnew · submitted 2026-06-06 · 💻 cs.RO

Learning Predictive Control with Deep Koopman Operators for Autonomous Vehicle Motion Planning

Xinglong Zhang , Yongqian Xiao , Haotian Cao , Xing Zhou , Xin Yin , Xin Xu This is my paper

Pith reviewed 2026-06-27 19:27 UTC · model grok-4.3

classification 💻 cs.RO
keywords learning predictive controldeep Koopman operatorsautonomous vehicle motion planningactor-critic reinforcement learningnonconvex constraintssafety-aware policy learningreceding-horizon controlobstacle avoidance
0
0 comments X

The pith

The LPC framework uses deep Koopman operators to produce closed-loop state-feedback policies for real-time autonomous vehicle motion planning under nonconvex constraints.

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

The paper proposes a learning predictive control framework that integrates deep Koopman operators with receding-horizon actor-critic learning to generate closed-loop policies for autonomous vehicle motion planning. Traditional MPC requires solving nonlinear nonconvex optimizations online and depends on accurate models, while actor-critic methods often lack explicit control structure. The approach lifts nonlinear and uncertain vehicle dynamics into a linear observable space via a data-driven deep Koopman predictor. Safety is addressed by constructing convex local obstacle surrogates and embedding their potential-field functions and gradients directly into the actor-critic structure. Simulations and real-world tests on the HongQi-EHS3 platform show gains in safety, efficiency, and comfort over methods like CBF-MPC and LMPCC.

Core claim

The LPC framework yields a closed-loop state-feedback policy within each prediction interval through receding-horizon actor-critic learning, enabling efficient, safety-aware policy learning under nonconvex environmental constraints, by using a deep-Koopman-based predictor to lift the nonlinear and uncertain vehicle dynamics into an interpretable linear observable space in a data-driven manner and by embedding convex local surrogate representations of obstacles along with their potential-field functions.

What carries the argument

Deep-Koopman-based predictor that lifts nonlinear vehicle dynamics to a linear observable space, combined with receding-horizon actor-critic learning that directly embeds potential-field safety functions derived from convex obstacle surrogates.

If this is right

  • Closed-loop policies are computed within each interval without solving full nonlinear optimizations at runtime.
  • Nonconvex environmental constraints are handled through embedded convex surrogates and potential fields for safety.
  • The method achieves improved computational efficiency and driving comfort relative to CBF-MPC and LMPCC in obstacle-avoidance scenarios.
  • Data-driven lifting allows the framework to operate without precise first-principles models of vehicle dynamics.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The linear lifted representation may permit direct application of linear control theory tools for stability or robustness analysis within the prediction horizon.
  • The framework could extend to other robotic platforms that face nonlinear dynamics and nonconvex constraints, such as manipulators or drones.
  • Embedding uncertainty estimates from the Koopman lifting step might further improve policy robustness in highly variable environments.
  • The approach reduces dependence on online optimization solvers, potentially enabling deployment on lower-power embedded hardware.

Load-bearing premise

The deep-Koopman-based predictor can accurately lift the nonlinear and uncertain vehicle dynamics into an interpretable linear observable space in a data-driven manner.

What would settle it

A real-world driving test on the HongQi-EHS3 where the trained deep Koopman predictor produces a policy that collides with obstacles or violates constraints when vehicle dynamics differ from the training distribution.

Figures

Figures reproduced from arXiv: 2606.08136 by Haotian Cao, Xinglong Zhang, Xing Zhou, Xin Xu, Xin Yin, Yongqian Xiao.

Figure 1
Figure 1. Figure 1: Deep Koopman vehicle dynamics modeling framework (MLP: Multi [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Illustration of convex representation for obstacle constraints. [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: LPC with deep Koopman models and actor-critic learning (Obst.: [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Open-loop multi-step prediction verification results of the learned deep Koopman model. overall root-mean-square error (RMSE) of the observable￾space multi-step prediction is defined as Eo = ¿ÁÁÁÀ 1 MN NpL MN ∑ j=1 Np ∑ k=1 L ∑ i=1 (ϕi(x [j] k ) − ϕˆ i(x [j] k ))2 , (26) where L is the observable-space dimension [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Vehicle obstacle avoidance path in Scenario I. The subfigures illustrate [PITH_FULL_IMAGE:figures/full_fig_p009_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Lateral displacement and heading angle errors of vehicle obstacle [PITH_FULL_IMAGE:figures/full_fig_p010_6.png] view at source ↗
Figure 9
Figure 9. Figure 9: Vehicle overtaking paths in Scenario III, the subfigures illustrate the [PITH_FULL_IMAGE:figures/full_fig_p010_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: The lateral displacement and heading angle deviation of the vehicle’s [PITH_FULL_IMAGE:figures/full_fig_p011_10.png] view at source ↗
Figure 13
Figure 13. Figure 13: Computational time of different methods per time step. [PITH_FULL_IMAGE:figures/full_fig_p011_13.png] view at source ↗
Figure 11
Figure 11. Figure 11: Vehicle overtaking paths in Scenario IV, the subfigures illustrate the [PITH_FULL_IMAGE:figures/full_fig_p011_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: The lateral displacement and heading angle deviation of the vehicle’s [PITH_FULL_IMAGE:figures/full_fig_p011_12.png] view at source ↗
Figure 15
Figure 15. Figure 15: Real-world experiments: vehicle obstacle avoidance trajectories on a straight road (Proposed vs. CBF-MPC). [PITH_FULL_IMAGE:figures/full_fig_p013_15.png] view at source ↗
Figure 16
Figure 16. Figure 16: In-vehicle and outside-view snapshots of the proposed method during the two real-world test scenarios. [PITH_FULL_IMAGE:figures/full_fig_p013_16.png] view at source ↗
Figure 17
Figure 17. Figure 17: Real-world experiments: vehicle obstacle avoidance trajectories on a curved road with surface transition (Proposed vs. CBF-MPC). [PITH_FULL_IMAGE:figures/full_fig_p014_17.png] view at source ↗
read the original abstract

Model Predictive Control (MPC) is widely used for autonomous-vehicle (AV) motion planning, but its real-time applicability is often limited by the need for accurate models and online solution of nonlinear, nonconvex optimization problems in dynamic road environments. Actor-critic reinforcement learning offers a promising alternative for online policy generation, yet its policy-learning process often lacks explicit control-theoretic structure. This article proposes a learning predictive control (LPC) framework with deep Koopman operators for efficient real-time motion planning under nonconvex constraints. To address nonlinear and uncertain vehicle dynamics, a deep-Koopman-based predictor is used to lift the system into an interpretable linear observable space in a data-driven manner. Unlike traditional MPC, which computes open-loop control sequences, the proposed LPC framework yields a closed-loop state-feedback policy within each prediction interval through receding-horizon actor-critic learning. To ensure safety under nonconvex environmental constraints, LPC constructs convex local surrogate representations of obstacles and defines corresponding potential-field functions. These functions and their gradients are directly embedded into the actor-critic structure, enabling efficient, safety-aware policy learning. Extensive simulations and real-world experiments on the HongQi-EHS3 platform demonstrate favorable performance in diverse obstacle-avoidance scenarios in terms of safety, computational efficiency, and driving comfort, compared with benchmark methods such as CBF-MPC and LMPCC.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 0 minor

Summary. The paper proposes a Learning Predictive Control (LPC) framework that integrates deep Koopman operators to lift nonlinear and uncertain vehicle dynamics into a linear observable space, combined with receding-horizon actor-critic learning to generate closed-loop state-feedback policies for AV motion planning. Safety under nonconvex constraints is addressed by constructing convex local obstacle surrogates and embedding corresponding potential-field functions and gradients directly into the actor-critic structure. The approach is positioned as more efficient than traditional nonlinear MPC for real-time use, with claims of favorable performance in safety, computational efficiency, and comfort demonstrated via extensive simulations and real-world experiments on the HongQi-EHS3 platform against benchmarks including CBF-MPC and LMPCC.

Significance. If the deep Koopman lifting accurately captures relevant dynamics and uncertainties over prediction horizons, and if the experimental claims are substantiated with quantitative evidence, the work could provide a structured hybrid method that combines data-driven linearization with explicit safety constraints in policy learning, potentially improving real-time applicability for AV planning compared to pure MPC or RL approaches. The direct embedding of potential fields into the actor-critic is a constructive element that merits attention.

major comments (2)
  1. [Abstract] Abstract: The central claim that 'extensive simulations and real-world experiments demonstrate favorable performance' in safety, efficiency, and comfort is load-bearing for the paper's contribution, yet the abstract provides no quantitative metrics, error bars, specific comparison values, or statistical details to support superiority over CBF-MPC and LMPCC; this absence prevents verification that the data actually backs the claims.
  2. [Abstract] Abstract (deep-Koopman predictor description): The framework's reliance on the deep-Koopman-based predictor to 'lift the system into an interpretable linear observable space in a data-driven manner' for use inside the receding-horizon actor-critic is a load-bearing assumption for both efficiency and safety claims, but no validation details (e.g., prediction error over horizons, handling of tire forces or friction variation) are supplied to address potential error accumulation.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback on the abstract. We address each major comment below and will revise the manuscript accordingly to strengthen the presentation of results and validation.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The central claim that 'extensive simulations and real-world experiments demonstrate favorable performance' in safety, efficiency, and comfort is load-bearing for the paper's contribution, yet the abstract provides no quantitative metrics, error bars, specific comparison values, or statistical details to support superiority over CBF-MPC and LMPCC; this absence prevents verification that the data actually backs the claims.

    Authors: We agree that the abstract would be strengthened by the inclusion of representative quantitative metrics. In the revised manuscript we will add concise highlights drawn from the simulation and experimental results (e.g., average computation time, safety-margin statistics, and comfort indices relative to CBF-MPC and LMPCC) while preserving the abstract's brevity. revision: yes

  2. Referee: [Abstract] Abstract (deep-Koopman predictor description): The framework's reliance on the deep-Koopman-based predictor to 'lift the system into an interpretable linear observable space in a data-driven manner' for use inside the receding-horizon actor-critic is a load-bearing assumption for both efficiency and safety claims, but no validation details (e.g., prediction error over horizons, handling of tire forces or friction variation) are supplied to address potential error accumulation.

    Authors: The manuscript contains closed-loop validation of the Koopman predictor through the overall planning performance; however, we acknowledge that explicit one-step and multi-step prediction-error statistics (including sensitivity to tire-force and friction variations) are not summarized in the abstract. We will add a short clause referencing these metrics from the experimental sections. revision: yes

Circularity Check

0 steps flagged

No circularity in derivation chain

full rationale

The abstract and description present a data-driven deep-Koopman lifting step followed by receding-horizon actor-critic policy learning and potential-field safety embedding. No equations, fitted parameters renamed as predictions, self-citations, or uniqueness theorems are supplied that would reduce any claimed result to its own inputs by construction. The method is described as relying on standard learning from data plus external benchmarks (CBF-MPC, LMPCC) and real-world experiments, making the chain self-contained against external validation rather than internally tautological.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Based on abstract only; no specific free parameters, axioms, or invented entities are identifiable.

pith-pipeline@v0.9.1-grok · 5780 in / 1119 out tokens · 25763 ms · 2026-06-27T19:27:15.023967+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

46 extracted references · 1 canonical work pages

  1. [1]

    Receding-horizon reinforcement learning approach for kinodynamic motion planning of autonomous vehicles.IEEE Transactions on Intelligent Vehicles, 7(3):556–568, 2022

    Xinglong Zhang, Yan Jiang, Yang Lu, and Xin Xu. Receding-horizon reinforcement learning approach for kinodynamic motion planning of autonomous vehicles.IEEE Transactions on Intelligent Vehicles, 7(3):556–568, 2022

    2022

  2. [2]

    The convex feasible set algorithm for real time optimization in motion planning

    Changliu Liu, Chung-Yen Lin, and Masayoshi Tomizuka. The convex feasible set algorithm for real time optimization in motion planning. SIAM Journal on Control and optimization, 56(4):2712–2733, 2018

    2018

  3. [3]

    A survey of the state- of-the-art localization techniques and their potentials for autonomous vehicle applications.IEEE Internet of Things Journal, 5:829–846, 2018

    Sampo Kuutti, Saber Fallah, Konstantinos Katsaros, Mehrdad Dianati, Francis Mccullough, and Alexandros Mouzakitis. A survey of the state- of-the-art localization techniques and their potentials for autonomous vehicle applications.IEEE Internet of Things Journal, 5:829–846, 2018

    2018

  4. [4]

    Autonomous driving motion planning with constrained iterative lqr.IEEE Transactions on Intelligent Vehicles, 4(2):244–254, 2019

    Jianyu Chen, Wei Zhan, and Masayoshi Tomizuka. Autonomous driving motion planning with constrained iterative lqr.IEEE Transactions on Intelligent Vehicles, 4(2):244–254, 2019

    2019

  5. [5]

    Safety-critical model predictive control with discrete-time control barrier function

    Jun Zeng, Bike Zhang, and Koushil Sreenath. Safety-critical model predictive control with discrete-time control barrier function. In2021 American Control Conference (ACC), pages 3882–3889. IEEE, 2021

    2021

  6. [6]

    Model predictive contouring control for collision avoidance in unstructured dy- namic environments.IEEE Robotics and Automation Letters, 4(4):4459– 4466, 2019

    Bruno Brito, Boaz Floor, Laura Ferranti, and Javier Alonso-Mora. Model predictive contouring control for collision avoidance in unstructured dy- namic environments.IEEE Robotics and Automation Letters, 4(4):4459– 4466, 2019

    2019

  7. [7]

    Multi-kernel online reinforcement learning for path tracking control of intelligent vehicles.IEEE Transactions on Systems, Man, and Cybernetics: Systems, 51(11):6962–6975, 2020

    Jiahang Liu, Zhenhua Huang, Xin Xu, Xinglong Zhang, Shiliang Sun, and Dazi Li. Multi-kernel online reinforcement learning for path tracking control of intelligent vehicles.IEEE Transactions on Systems, Man, and Cybernetics: Systems, 51(11):6962–6975, 2020

    2020

  8. [8]

    Lateral control for autonomous land vehicles via dual heuristic programming

    Zhenhua Huang, Chuanqiang Lian, Xin Xu, and Jian Wang. Lateral control for autonomous land vehicles via dual heuristic programming. International Journal of Robotics and Automation, 31(6), 2016

    2016

  9. [9]

    Xinglong Zhang, Jiahang Liu, Xin Xu, Shuyou Yu, and Hong Chen. Robust learning-based predictive control for discrete-time nonlinear sys- tems with unknown dynamics and state constraints.IEEE Transactions on Systems, Man, and Cybernetics: Systems, 52(12):7314–7327, 2022

    2022

  10. [10]

    Functional nonlinear model predictive control based on adaptive dynamic program- ming.IEEE transactions on Cybernetics, 49(12):4206–4218, 2018

    Lu Dong, Jun Yan, Xin Yuan, Haibo He, and Changyin Sun. Functional nonlinear model predictive control based on adaptive dynamic program- ming.IEEE transactions on Cybernetics, 49(12):4206–4218, 2018

    2018

  11. [11]

    Linear predictors for nonlinear dynamical systems: Koopman operator meets model predictive control.Automatica, 93:149–160, 2018

    Milan Korda and Igor Mezi ´c. Linear predictors for nonlinear dynamical systems: Koopman operator meets model predictive control.Automatica, 93:149–160, 2018

    2018

  12. [12]

    Koopman operator applications in signalized traffic systems.IEEE Transactions on Intelligent Transportation Systems, 23(4):3214–3225, 2020

    Esther Ling, Liyuan Zheng, Lillian J Ratliff, and Samuel Coogan. Koopman operator applications in signalized traffic systems.IEEE Transactions on Intelligent Transportation Systems, 23(4):3214–3225, 2020

    2020

  13. [13]

    Physically analyzable ai-based nonlinear platoon dynamics modeling during traffic oscillation: A koopman approach.IEEE Transactions on Intelligent Transportation Systems, 2025

    Kexin Tian, Haotian Shi, Yang Zhou, and Sixu Li. Physically analyzable ai-based nonlinear platoon dynamics modeling during traffic oscillation: A koopman approach.IEEE Transactions on Intelligent Transportation Systems, 2025

    2025

  14. [14]

    Differential high order control barrier function- based safe reinforcement learning.IEEE Robotics and Automation Letters, 2025

    Xiangyu Kong, Yuanqing Xia, Zhongqi Sun, Di-Hua Zhai, Yunshan Deng, and Sihua Zhang. Differential high order control barrier function- based safe reinforcement learning.IEEE Robotics and Automation Letters, 2025

    2025

  15. [15]

    Cbf-based hierar- chical quadratic programs with guaranteed feasibility for safety-critical systems.IEEE Transactions on Automation Science and Engineering, 22:23687–23699, 2025

    Junjun Xie, Liang Hu, Yunzhe Tan, and Jun Yang. Cbf-based hierar- chical quadratic programs with guaranteed feasibility for safety-critical systems.IEEE Transactions on Automation Science and Engineering, 22:23687–23699, 2025

    2025

  16. [16]

    Safe and fast tracking on a robot manipulator: Robust MPC and neural network control.IEEE Robotics and Automation Letters, 5(2):3050–3057, 2020

    Jakob Nubert, Jonas K ¨ohler, Vincent Berenz, Klaus L ¨utjens, and J ¨org Raisch. Safe and fast tracking on a robot manipulator: Robust MPC and neural network control.IEEE Robotics and Automation Letters, 5(2):3050–3057, 2020

    2020

  17. [17]

    MPC-based haptic shared steering system: A driver modeling approach for symbiotic driving

    Ander M Lazcano, Tianyi Niu, Xabier C Akutain, Mercedes Delgado, Oihana Otaegui, Unai Jauregi, and et al. MPC-based haptic shared steering system: A driver modeling approach for symbiotic driving. IEEE/ASME Transactions on Mechatronics, 26(3):1201–1211, 2021

    2021

  18. [18]

    A potential field-based model predictive path-planning con- troller for autonomous road vehicles.IEEE Transactions on Intelligent Transportation Systems, 18(5):1255–1267, 2016

    Yadollah Rasekhipour, Amir Khajepour, Shih-Ken Chen, and Bakhtiar Litkouhi. A potential field-based model predictive path-planning con- troller for autonomous road vehicles.IEEE Transactions on Intelligent Transportation Systems, 18(5):1255–1267, 2016

    2016

  19. [19]

    Lateral vehicle trajectory optimization using constrained linear time-varying MPC

    Benjamin Gutjahr, Lutz Gr ¨oll, and Moritz Werling. Lateral vehicle trajectory optimization using constrained linear time-varying MPC. IEEE Transactions on Intelligent Transportation Systems, 18(6):1586– 1595, 2016

    2016

  20. [20]

    Path planning and tracking for vehicle collision avoidance based on model predictive control with multiconstraints.IEEE Transactions on Vehicular Technology, 66(2):952–964, 2016

    Jie Ji, Amir Khajepour, Wael William Melek, and Yanjun Huang. Path planning and tracking for vehicle collision avoidance based on model predictive control with multiconstraints.IEEE Transactions on Vehicular Technology, 66(2):952–964, 2016

    2016

  21. [21]

    Qianxiao Li, Felix Dietrich, Erik M Bollt, and Ioannis G Kevrekidis. Ex- tended dynamic mode decomposition with dictionary learning: A data- driven adaptive spectral decomposition of the Koopman operator.Chaos: An Interdisciplinary Journal of Nonlinear Science, 27(10):103111, 2017

    2017

  22. [22]

    Witherden, and Mykel J

    Jeremy Morton, Freddie D. Witherden, and Mykel J. Kochenderfer. Deep variational koopman models: Inferring koopman observations for uncertainty-aware dynamics modeling and control. InProceedings of the 28th International Joint Conference on Artificial Intelligence (IJCAI), IJCAI’19, page 3173–3179. AAAI Press, 2019. 15

    2019

  23. [23]

    Joglekar et al

    A. Joglekar et al. Data-driven modeling and experimental validation of autonomous vehicles using koopman operator. In2023 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pages 9442–9447, Detroit, MI, USA, 2023

    2023

  24. [24]

    Deep koopman traffic modeling for freeway ramp metering.IEEE Transactions on Intelligent Transportation Systems, 24(6):6001–6013, 2023

    Chuanye Gu, Tao Zhou, and Changzhi Wu. Deep koopman traffic modeling for freeway ramp metering.IEEE Transactions on Intelligent Transportation Systems, 24(6):6001–6013, 2023

    2023

  25. [25]

    Transformers for modeling physical systems.Neural Networks, 146:272–289, 2022

    Nicholas Geneva and Nicholas Zabaras. Transformers for modeling physical systems.Neural Networks, 146:272–289, 2022

    2022

  26. [26]

    Characterization of groundwater contamination: A transformer-based deep learning model.Advances in Water Resources, 164:104217, 2022

    Tao Bai and Pejman Tahmasebi. Characterization of groundwater contamination: A transformer-based deep learning model.Advances in Water Resources, 164:104217, 2022

    2022

  27. [27]

    Deepkoco: Efficient latent planning with a task-relevant koopman rep- resentation

    Bas van der Heijden, Laura Ferranti, Jens Kober, and Robert Babu ˇska. Deepkoco: Efficient latent planning with a task-relevant koopman rep- resentation. In2021 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), page 183–189. IEEE Press, 2021

    2021

  28. [28]

    Near- optimal rapid MPC using neural networks: A primal-dual policy learn- ing framework.IEEE Transactions on Control Systems Technology, 29(5):2102–2114, 2020

    Xiaojing Zhang, Monimoy Bujarbaruah, and Francesco Borrelli. Near- optimal rapid MPC using neural networks: A primal-dual policy learn- ing framework.IEEE Transactions on Control Systems Technology, 29(5):2102–2114, 2020

    2020

  29. [29]

    Autonomous driving using linear model predictive control with a Koopman operator based bilinear vehicle model.IFAC-PapersOnLine, 55(24):254–259, 2022

    Siyuan Yu, Congkai Shen, and Tulga Ersal. Autonomous driving using linear model predictive control with a Koopman operator based bilinear vehicle model.IFAC-PapersOnLine, 55(24):254–259, 2022

    2022

  30. [30]

    A review of end-to-end autonomous driving in urban environments.IEEE Access, 10:75296–75311, 2022

    Daniel Coelho and Miguel Oliveira. A review of end-to-end autonomous driving in urban environments.IEEE Access, 10:75296–75311, 2022

    2022

  31. [31]

    Deep reinforcement learning for autonomous driving: A survey.IEEE Transactions on Intelligent Transportation Systems, 23(6):4909–4926, 2021

    B Ravi Kiran, Ibrahim Sobh, Victor Talpaert, Patrick Mannion, Ahmad A Al Sallab, Senthil Yogamani, and Patrick P ´erez. Deep reinforcement learning for autonomous driving: A survey.IEEE Transactions on Intelligent Transportation Systems, 23(6):4909–4926, 2021

    2021

  32. [32]

    Real- time drift-driving control for an autonomous vehicle: Learning from nonlinear model predictive control via a deep neural network.Electron- ics, 11(17):2651, 2022

    Taekgyu Lee, Dongyoon Seo, Jinyoung Lee, and Yeonsik Kang. Real- time drift-driving control for an autonomous vehicle: Learning from nonlinear model predictive control via a deep neural network.Electron- ics, 11(17):2651, 2022

    2022

  33. [33]

    Accept synthetic objects as real: End-to-end training of attentive deep visuomotor policies for manipulation in clutter

    Pooya Abolghasemi and Ladislau B ¨ol¨oni. Accept synthetic objects as real: End-to-end training of attentive deep visuomotor policies for manipulation in clutter. In2020 IEEE International Conference on Robotics and Automation (ICRA), pages 6506–6512. IEEE, 2020

    2020

  34. [34]

    End-to-end steering controller with cnn-based closed-loop feedback for autonomous vehicles

    Junekyo Jhung, Il Bae, Jaeyoung Moon, Taewoo Kim, Jincheol Kim, and Shiho Kim. End-to-end steering controller with cnn-based closed-loop feedback for autonomous vehicles. In2018 IEEE Intelligent Vehicles Symposium (IV), pages 617–622. IEEE, 2018

    2018

  35. [35]

    Mixgail: Autonomous driving using demonstrations with mixed qualities

    Gunmin Lee, Dohyeong Kim, Wooseok Oh, Kyungjae Lee, and Songh- wai Oh. Mixgail: Autonomous driving using demonstrations with mixed qualities. In2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pages 5425–5430. IEEE, 2020

    2020

  36. [37]

    arXiv preprint arXiv:2307.07162

    Daocheng Fu, Xin Li, Licheng Wen, Min Dou, Pinlong Cai, Botian Shi, and Yu Qiao. Drive as you speak: Enabling human-like interaction with large language models in autonomous vehicles.arXiv preprint arXiv:2307.07162, 2023

    work page arXiv 2023

  37. [38]

    Model-free deep reinforcement learning for urban autonomous driving

    Jianyu Chen, Bodi Yuan, and Masayoshi Tomizuka. Model-free deep reinforcement learning for urban autonomous driving. In2019 IEEE Intelligent Transportation Systems Conference (ITSC), pages 2765–2771. IEEE, 2019

    2019

  38. [39]

    Uncertainty-aware model-based reinforcement learning: Methodology and application in autonomous driving.IEEE Transactions on Intelligent Vehicles, 8(1):194–203, 2022

    Jingda Wu, Zhiyu Huang, and Chen Lv. Uncertainty-aware model-based reinforcement learning: Methodology and application in autonomous driving.IEEE Transactions on Intelligent Vehicles, 8(1):194–203, 2022

    2022

  39. [40]

    Interpretable end-to-end urban autonomous driving with latent deep reinforcement learning.IEEE Transactions on Intelligent Transportation Systems, 2021

    Jianyu Chen, Shengbo Eben Li, and Masayoshi Tomizuka. Interpretable end-to-end urban autonomous driving with latent deep reinforcement learning.IEEE Transactions on Intelligent Transportation Systems, 2021

    2021

  40. [41]

    Self-learning cruise control using kernel-based least squares policy iter- ation.IEEE Transactions on Control Systems Technology, 22(3):1078– 1087, 2013

    Jian Wang, Xin Xu, Daxue Liu, Zhenping Sun, and Qingyang Chen. Self-learning cruise control using kernel-based least squares policy iter- ation.IEEE Transactions on Control Systems Technology, 22(3):1078– 1087, 2013

    2013

  41. [42]

    An approxi- mate dynamic programming approach for path following control of an autonomous vehicle

    Kun Zhao, Jian Wang, Xin Xu, and Zhenhua Huang. An approxi- mate dynamic programming approach for path following control of an autonomous vehicle. InProceeding of the 11th World Congress on Intelligent Control and Automation (WCICA), pages 1998–2004. IEEE, 2014

    1998

  42. [43]

    Parameterized batch reinforcement learning for longitudinal control of autonomous land vehicles.IEEE Transactions on Systems, Man, and Cybernetics: Systems, 49(4):730–741, 2019

    Zhenhua Huang, Xin Xu, Haibo He, Jun Tan, and Zhenping Sun. Parameterized batch reinforcement learning for longitudinal control of autonomous land vehicles.IEEE Transactions on Systems, Man, and Cybernetics: Systems, 49(4):730–741, 2019

    2019

  43. [44]

    Learning-based predictive control for discrete-time nonlinear systems with stochastic disturbances.IEEE Transactions on Neural Networks and Learning Systems, 29(12):6202–6213, 2018

    Xin Xu, Hong Chen, Chuanqiang Lian, and Dazi Li. Learning-based predictive control for discrete-time nonlinear systems with stochastic disturbances.IEEE Transactions on Neural Networks and Learning Systems, 29(12):6202–6213, 2018

    2018

  44. [45]

    Deep neural networks with Koopman operators for modeling and control of autonomous vehicles.IEEE Transactions on Intelligent Vehicles, 8(1):135–146, 2023

    Yongqian Xiao, Xinglong Zhang, Xin Xu, Xueqing Liu, and Jiahang Liu. Deep neural networks with Koopman operators for modeling and control of autonomous vehicles.IEEE Transactions on Intelligent Vehicles, 8(1):135–146, 2023

    2023

  45. [46]

    Model-based safe reinforcement learning with time-varying constraints: Applications to intelligent vehicles.IEEE Transactions on Industrial Electronics, 71(10):12744–12753, 2024

    Xinglong Zhang, Yaoqian Peng, Biao Luo, Wei Pan, Xin Xu, and Haibin Xie. Model-based safe reinforcement learning with time-varying constraints: Applications to intelligent vehicles.IEEE Transactions on Industrial Electronics, 71(10):12744–12753, 2024

    2024

  46. [47]

    Toward scalable multirobot control: Fast policy learning in distributed mpc.IEEE Transactions on Robotics, 41:1491– 1512, 2025

    Xinglong Zhang, Wei Pan, Cong Li, Xin Xu, Xiangke Wang, Ronghua Zhang, and Dewen Hu. Toward scalable multirobot control: Fast policy learning in distributed mpc.IEEE Transactions on Robotics, 41:1491– 1512, 2025. APPENDIX A. Learning Convergence in Each Prediction Interval In the following, we analyze the convergence of the policy learning procedure in ea...

    2025