pith. sign in

arxiv: 2606.29222 · v1 · pith:TEX6UUSYnew · submitted 2026-06-28 · 💻 cs.RO

CORE Planner: Contextual-memory Oriented Reinforcement-learning in Unknown Environments for Robot Navigation

Jintao Kong , Zhihao Zhang , Weihuang Chen , Liming Chen , Zhongyu Guo , Shuaiyu Liu , Hongbin Sun This is my paper

Pith reviewed 2026-06-30 07:38 UTC · model grok-4.3

classification 💻 cs.RO
keywords robot navigationreinforcement learningunknown environmentssim-to-real transfervisibility graphcontextual memorytransformer networkautonomous navigation
0
0 comments X

The pith

The CORE planner integrates sparse visibility graphs with Transformer contextual memory in reinforcement learning to navigate unknown environments and achieve zero-shot sim-to-real transfer without fine-tuning.

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

The paper aims to show that a hybrid planner can combine the structure of traditional methods with the adaptability of learning to solve robot navigation in spaces with no prior map. It builds a sparse visibility graph to represent the environment efficiently instead of dense grids, then uses a Transformer to maintain contextual memory across observations so the agent avoids repeating mistakes or getting stuck. Training happens only in simulated image environments, yet the system deploys directly on physical robots. If the claim holds, robots could be trained at low cost in simulation and operate reliably in real unknown settings, cutting path lengths substantially compared with both rule-based and pure learning approaches.

Core claim

The CORE planner uses a visibility graph for structured environment representation together with graph sparsification and a contextual memory mechanism inside a reinforcement learning policy. A Transformer network processes the memory to produce a holistic understanding that improves decision making. This combination yields lower travel distances than the FAR Planner or learning baselines and supports direct transfer from image-only simulation training to real-robot execution without any fine-tuning or human intervention in representative environments.

What carries the argument

The CORE planner, which fuses a sparse visibility graph representation and graph sparsification with a Transformer-driven contextual memory module inside reinforcement learning.

If this is right

  • Travel distance drops 13 percent relative to the traditional FAR Planner across tested environments.
  • Performance gains reach up to 48 percent over learning-based baselines, with larger improvements in complex scenes.
  • Zero-shot deployment succeeds in real-world navigation tasks without further training or human assistance.
  • Graph sparsification keeps computation tractable even as environments grow larger.
  • The same trained policy works across both simulated and physical settings without domain-specific adjustments.

Where Pith is reading between the lines

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

  • Similar memory-augmented graph representations could reduce data requirements for other robotic skills that involve long-horizon exploration.
  • The approach points to a general pattern where sparse structured inputs help reinforcement learners generalize across the simulation-reality boundary more reliably than dense image or occupancy inputs alone.
  • If the memory mechanism scales, it may support deployment in progressively larger or more cluttered spaces where pure reactive policies tend to fail.

Load-bearing premise

That the visibility graph plus contextual memory learned from simulated images will continue to produce correct navigation actions on physical robots despite real sensor noise and dynamics.

What would settle it

A physical robot trial in which the trained policy produces repeated collisions or entrapment in local minima when sensor readings differ from the simulated visibility graphs used in training.

Figures

Figures reproduced from arXiv: 2606.29222 by Hongbin Sun, Jintao Kong, Liming Chen, Shuaiyu Liu, Weihuang Chen, Zhihao Zhang, Zhongyu Guo.

Figure 1
Figure 1. Figure 1: Illustration of the unknown-environment navigation [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Illustration of visibility graph construction [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: We explicitly note that this architecture is not [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 3
Figure 3. Figure 3: The overall pipeline of the proposed CORE planner for zero-shot navigation. A visibility graph (orange nodes [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: The training set includes both simple (a) and [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: As shown in the table, our method achieved a 100% [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗
Figure 5
Figure 5. Figure 5: Comparison of representative navigation trajec [PITH_FULL_IMAGE:figures/full_fig_p009_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Sequential navigation through waypoints in environments of varying complexity: (a) simple environment, (b) [PITH_FULL_IMAGE:figures/full_fig_p010_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Experimental Robotic Platforms: LYNX M20 (a) [PITH_FULL_IMAGE:figures/full_fig_p011_7.png] view at source ↗
Figure 9
Figure 9. Figure 9: Illustration of the LYNX M20 Quadruped Robot [PITH_FULL_IMAGE:figures/full_fig_p011_9.png] view at source ↗
Figure 8
Figure 8. Figure 8: Real-world navigation performance in a multi [PITH_FULL_IMAGE:figures/full_fig_p011_8.png] view at source ↗
read the original abstract

Autonomous navigation in unknown environments requires a robot to efficiently reach a predefined goal while exploring without prior maps. Although progress has been made in this area, most existing works still rely on traditional planning methods with hand-crafted rules, while learning-based methods often suffer from limited environmental memory and challenges in simulation-to-real (sim-to-real) transfer. To overcome these limitations, we propose a Contextual-memory Oriented Reinforcement-learning (CORE) planner for robot navigation in unknown environments. The proposed CORE planner effectively combines the core advantages of traditional and learning-based methods. Specifically, our method uses a sparse visibility graph for structured environment representation, reducing the computational overhead of dense grid maps, and employs a Transformer network to achieve a holistic environmental understanding, thereby significantly improving navigation efficiency. Moreover, we introduce a visibility graph-based graph sparsification method and a contextual memory mechanism, which alleviates local optima and enhances computational performance in large-scale scenes. Finally, our approach achieves zero-shot sim-to-real transfer after training solely on image-based environments, requiring no fine-tuning. Experimental results show that CORE Planner consistently outperforms state-of-the-art methods, including the traditional FAR Planner and all learning-based baselines, across representative environments, reducing travel distance by 13\% over traditional FAR Planner and by up to 48\% relative to learning-based baselines, with larger gains observed in more complex environments. In real-world scenarios, CORE successfully navigates without human intervention, showcasing zero-shot sim-to-real transfer. Code is available at https://github.com/BBD00/core_planner.

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

3 major / 2 minor

Summary. The manuscript proposes CORE Planner, a hybrid approach combining sparse visibility graphs with Transformer-based contextual memory in a reinforcement learning framework for autonomous robot navigation in unknown environments. It claims superior performance over traditional (FAR Planner) and learning-based methods, with 13% and up to 48% reductions in travel distance respectively, and demonstrates zero-shot sim-to-real transfer without fine-tuning, supported by real-world experiments.

Significance. If the performance improvements and zero-shot transfer hold under scrutiny, this work could advance the field by effectively merging the strengths of structured planning and learning-based methods, potentially improving efficiency in large-scale scenes and reducing reliance on fine-tuning for real-world deployment. The provision of code enhances the potential for reproducibility and further research.

major comments (3)
  1. [Abstract] Abstract: The central claim of zero-shot sim-to-real transfer after training solely on image-based environments is load-bearing, yet the manuscript supplies no quantitative characterization of the sim-to-real image gap, no noise model, and no ablation on visibility-graph quality under perturbation, leaving the robustness of the sparse graph extraction unverified.
  2. [§5 (Experiments)] §5 (Experiments) and real-world results paragraph: The reported 13% distance reduction over FAR Planner and up to 48% over learning baselines, plus successful real-world runs without intervention, are presented without details on trial counts, variance, statistical tests, or controls for sensor noise/calibration, undermining assessment of whether the gains generalize beyond the evaluated setups.
  3. [Methods] Methods section on visibility graph sparsification and contextual memory: The claim that these mechanisms alleviate local optima and enable large-scale performance is central, but without explicit equations or pseudocode showing how the Transformer memory integrates with the graph to produce noise-invariant plans, the zero-shot assertion rests on an untested assumption.
minor comments (2)
  1. [Abstract] Abstract: The statement 'larger gains observed in more complex environments' lacks a definition of complexity (e.g., obstacle density or map size); adding this would improve precision.
  2. [Abstract] The GitHub link is provided, but the manuscript does not specify which code components correspond to the graph extraction versus the RL policy, reducing immediate reproducibility.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive feedback. The comments highlight opportunities to strengthen the presentation of the zero-shot transfer claim and the experimental rigor. We respond to each major comment below and will incorporate revisions accordingly.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The central claim of zero-shot sim-to-real transfer after training solely on image-based environments is load-bearing, yet the manuscript supplies no quantitative characterization of the sim-to-real image gap, no noise model, and no ablation on visibility-graph quality under perturbation, leaving the robustness of the sparse graph extraction unverified.

    Authors: We agree that additional quantitative support would strengthen the zero-shot claim. The manuscript demonstrates successful real-world navigation without fine-tuning, but does not currently include explicit gap metrics or ablations. In revision we will add a characterization of the sim-to-real image differences (e.g., feature distribution statistics), a basic sensor noise model, and an ablation on visibility-graph robustness under synthetic perturbations. These additions will be placed in §5 or a new appendix subsection. revision: yes

  2. Referee: [§5 (Experiments)] §5 (Experiments) and real-world results paragraph: The reported 13% distance reduction over FAR Planner and up to 48% over learning baselines, plus successful real-world runs without intervention, are presented without details on trial counts, variance, statistical tests, or controls for sensor noise/calibration, undermining assessment of whether the gains generalize beyond the evaluated setups.

    Authors: The reported improvements are averages over repeated trials, yet the manuscript indeed omits explicit counts, variance, and statistical tests. We will revise §5 to report the exact number of trials per environment, include standard deviations, add statistical significance tests (paired t-tests), and describe sensor calibration and noise-handling procedures used in both simulation and real-world experiments. revision: yes

  3. Referee: [Methods] Methods section on visibility graph sparsification and contextual memory: The claim that these mechanisms alleviate local optima and enable large-scale performance is central, but without explicit equations or pseudocode showing how the Transformer memory integrates with the graph to produce noise-invariant plans, the zero-shot assertion rests on an untested assumption.

    Authors: The methods section describes the sparsification and Transformer memory integration in prose. To improve clarity we will add formal equations for the sparsification step and the memory-conditioned policy update, together with pseudocode for the full planning loop. These additions will make explicit how the contextual memory contributes to noise-invariant plan generation. revision: yes

Circularity Check

0 steps flagged

No circularity detected; method and claims rest on independent empirical results

full rationale

The provided abstract and description outline a hybrid planner using sparse visibility graphs, Transformer contextual memory, and graph sparsification for RL-based navigation, with zero-shot sim-to-real transfer asserted via experimental outcomes (13% and 48% gains). No equations, derivations, or self-citations appear that would reduce any prediction or uniqueness claim to fitted inputs or prior author work by construction. Performance metrics are presented as measured results rather than tautological consequences of the architecture definition, and no load-bearing self-referential steps are identifiable from the given text.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract provides no equations, training details, or explicit assumptions, so no free parameters, axioms, or invented entities can be extracted.

pith-pipeline@v0.9.1-grok · 5825 in / 1042 out tokens · 24328 ms · 2026-06-30T07:38:52.331632+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

39 extracted references · 4 canonical work pages · 1 internal anchor

  1. [1]

    A compre- hensive review on autonomous navigation,

    S. Nahavandi, R. Alizadehsani, D. Nahavandi, S. Mohamed, N. Mohajer, M. Rokonuzzaman, and I. Hossain, “A compre- hensive review on autonomous navigation,” ACM Computing Surveys, vol. 57, no. 9, pp. 1–67, 2025

    2025

  2. [2]

    Advancements and challenges in mobile robot nav- igation: A comprehensive review of algorithms and potential for self-learning approaches,

    S. Al Mahmud, A. Kamarulariffin, A. M. Ibrahim, and A. J. H. Mohideen, “Advancements and challenges in mobile robot nav- igation: A comprehensive review of algorithms and potential for self-learning approaches,” Journal of Intelligent & Robotic Systems, vol. 110, no. 3, p. 120, 2024

    2024

  3. [3]

    Online robot naviga- tion using discrete waypoints via time-varying guidance vector fields,

    J. Wang, L. Zhao, F. Liu, and K. Xia, “Online robot naviga- tion using discrete waypoints via time-varying guidance vector fields,” IEEE Transactions on Industrial Electronics, 2024

    2024

  4. [4]

    F AR plan- ner: Fast, attemptable route planner using dynamic visibility update,

    F. Yang, C. Cao, H. Zhu, J. Oh, and J. Zhang, “F AR plan- ner: Fast, attemptable route planner using dynamic visibility update,” in Proc. IEEE/RSJ Int. Conf. Intell. Robots Syst. (IROS). IEEE, 2022, pp. 9–16

    2022

  5. [5]

    NavRL: Learning safe flight in dynamic environments,

    Z. Xu, X. Han, H. Shen, H. Jin, and K. Shimada, “NavRL: Learning safe flight in dynamic environments,” IEEE Robotics and Automation Letters, 2025

    2025

  6. [6]

    You only plan once: A learning-based one-stage planner with guidance learning,

    J. Lu, X. Zhang, H. Shen, L. Xu, and B. Tian, “You only plan once: A learning-based one-stage planner with guidance learning,” IEEE Robotics and Automation Letters, vol. 9, no. 7, pp. 6083–6090, 2024

    2024

  7. [7]

    Goal-driven au- tonomous exploration through deep reinforcement learning,

    R. Cimurs, I. H. Suh, and J. H. Lee, “Goal-driven au- tonomous exploration through deep reinforcement learning,” IEEE Robotics and Automation Letters, vol. 7, no. 2, pp. 730– 737, 2021

    2021

  8. [8]

    Autonomous navigation by mobile robot with sensor fusion based on deep reinforcement learning,

    Y. Ou, Y. Cai, Y. Sun, and T. Qin, “Autonomous navigation by mobile robot with sensor fusion based on deep reinforcement learning,” Sensors, vol. 24, no. 12, p. 3895, 2024

    2024

  9. [9]

    Voronoi- based multi-robot autonomous exploration in unknown environ- ments via deep reinforcement learning,

    J. Hu, H. Niu, J. Carrasco, B. Lennox, and F. Arvin, “Voronoi- based multi-robot autonomous exploration in unknown environ- ments via deep reinforcement learning,” IEEE Transactions on Vehicular Technology, vol. 69, no. 12, pp. 14 413–14 423, 2020

    2020

  10. [10]

    Goal-guided transformer-enabled reinforcement learning for efficient au- tonomous navigation,

    W. Huang, Y. Zhou, X. He, and C. Lv, “Goal-guided transformer-enabled reinforcement learning for efficient au- tonomous navigation,” IEEE Transactions on Intelligent Trans- portation Systems, vol. 25, no. 2, pp. 1832–1845, 2023

    2023

  11. [11]

    Attention is all you need,

    A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, Ł. Kaiser, and I. Polosukhin, “Attention is all you need,” in Advances in Neural Information Processing Systems (NeurIPS), vol. 30, 2017

    2017

  12. [12]

    Language models are few-shot learners,

    T. Brown, B. Mann, N. Ryder, M. Subbiah, J. D. Kaplan, P. Dhariwal, A. Neelakantan, P. Shyam, G. Sastry, A. Askell et al., “Language models are few-shot learners,” in Advances in Neural Information Processing Systems (NeurIPS), vol. 33, 2020, pp. 1877–1901

    2020

  13. [13]

    CTSAC: Curriculum- based transformer soft actor-critic for goal-oriented robot ex- ploration,

    C. Yang, S. Bi, Y. Xu, and X. Zhang, “CTSAC: Curriculum- based transformer soft actor-critic for goal-oriented robot ex- ploration,” arXiv preprint arXiv:2503.14254, 2025

    work page arXiv 2025

  14. [14]

    Context-aware deep reinforcement learning for au- tonomous robotic navigation in unknown area,

    J. Liang, Z. Wang, Y. Cao, J. Chiun, M. Zhang, and G. A. Sartoretti, “Context-aware deep reinforcement learning for au- tonomous robotic navigation in unknown area,” in Proc. Conf. Robot Learn. (CoRL). PMLR, 2023, pp. 1425–1436

    2023

  15. [15]

    3D-Mem: 3D scene memory for embodied exploration and reasoning,

    Y. Yang, H. Yang, J. Zhou, P. Chen, H. Zhang, Y. Du, and C. Gan, “3D-Mem: 3D scene memory for embodied exploration and reasoning,” in Proc. IEEE/CVF Conf. Comput. Vis. Pat- tern Recognit. (CVPR), 2025, pp. 17 294–17 303

    2025

  16. [16]

    Move to understand a 3D scene: Bridging visual grounding and exploration for efficient and versatile embodied navigation,

    Z. Zhu, X. Wang, Y. Li, Z. Zhang, X. Ma, Y. Chen, B. Jia, W. Liang, Q. Yu, Z. Deng et al., “Move to understand a 3D scene: Bridging visual grounding and exploration for efficient and versatile embodied navigation,” in Proc. IEEE/CVF Int. Conf. Comput. Vis. (ICCV), 2025, pp. 8120–8132

    2025

  17. [17]

    A note on two problems in connexion with graphs,

    E. W. Dijkstra, “A note on two problems in connexion with graphs,” in Edsger Wybe Dijkstra: His Life, Work, and Legacy, 2022, pp. 287–290

    2022

  18. [18]

    A formal basis for the heuristic determination of minimum cost paths,

    P. E. Hart, N. J. Nilsson, and B. Raphael, “A formal basis for the heuristic determination of minimum cost paths,” IEEE Transactions on Systems Science and Cybernetics, vol. 4, no. 2, pp. 100–107, 1968

    1968

  19. [19]

    Optimal and efficient path planning for partially- known environments,

    A. Stentz, “Optimal and efficient path planning for partially- known environments,” in Proc. IEEE Int. Conf. Robot. Autom. (ICRA). IEEE, 1994, pp. 3310–3317

    1994

  20. [20]

    Lifelong planning A*,

    S. Koenig, M. Likhachev, and D. Furcy, “Lifelong planning A*,” Artificial Intelligence, vol. 155, no. 1-2, pp. 93–146, 2004

    2004

  21. [21]

    Fast replanning for navigation in unknown terrain,

    S. Koenig and M. Likhachev, “Fast replanning for navigation in unknown terrain,” IEEE Transactions on Robotics, vol. 21, no. 3, pp. 354–363, 2005

    2005

  22. [22]

    Sampling-based algorithms for op- timal motion planning,

    S. Karaman and E. Frazzoli, “Sampling-based algorithms for op- timal motion planning,” The International Journal of Robotics Research, vol. 30, no. 7, pp. 846–894, 2011

    2011

  23. [23]

    RRT-connect: An efficient approach to single-query path planning,

    J. J. Kuffner and S. M. LaValle, “RRT-connect: An efficient approach to single-query path planning,” in Proc. IEEE Int. Conf. Robot. Autom. (ICRA), vol. 2. IEEE, 2000, pp. 995– 1001

    2000

  24. [24]

    Batch informed trees (BIT*): Sampling-based optimal planning via the heuristically guided search of implicit random geometric graphs,

    J. D. Gammell, S. S. Srinivasa, and T. D. Barfoot, “Batch informed trees (BIT*): Sampling-based optimal planning via the heuristically guided search of implicit random geometric graphs,” in Proc. IEEE Int. Conf. Robot. Autom. (ICRA). IEEE, 2015, pp. 3067–3074

    2015

  25. [25]

    Chase and track: Toward safe and smooth trajectory planning for robotic navigation in dynamic environments,

    C. Wang, X. Chen, C. Li, R. Song, Y. Li, and M. Q.-H. Meng, “Chase and track: Toward safe and smooth trajectory planning for robotic navigation in dynamic environments,” IEEE Transactions on Industrial Electronics, vol. 70, no. 1, pp. 604–613, 2022

    2022

  26. [26]

    Rapidly-exploring random trees: Progress and prospects,

    S. M. LaValle and J. J. Kuffner, “Rapidly-exploring random trees: Progress and prospects,” Algorithmic and Computational Robotics, pp. 303–307, 2001

    2001

  27. [27]

    E-Planner: An efficient path planner on a visibility graph in unknown environments,

    C. Zhang, G. Wang, M. Chen, Y. Lin, K. Li, M. Wu, Z. Li, and Q. Wang, “E-Planner: An efficient path planner on a visibility graph in unknown environments,” IEEE Transactions on Instrumentation and Measurement, 2024

    2024

  28. [28]

    FPS: Fast path planner algorithm based on sparse visibility graph and bidirectional breadth-first search,

    Q. Li, F. Xie, J. Zhao, B. Xu, J. Yang, X. Liu, and H. Suo, “FPS: Fast path planner algorithm based on sparse visibility graph and bidirectional breadth-first search,” Remote Sensing, vol. 14, no. 15, p. 3720, 2022

    2022

  29. [29]

    A review of convolutional neural network architectures and their optimizations,

    S. Cong and Y. Zhou, “A review of convolutional neural network architectures and their optimizations,” Artificial Intelligence Review, vol. 56, no. 3, pp. 1905–1969, 2023

    1905

  30. [30]

    Velma: Verbalization embodiment of LLM agents for vision and language navigation in street view,

    R. Schumann, W. Zhu, W. Feng, T.-J. Fu, S. Riezler, and W. Y. Wang, “Velma: Verbalization embodiment of LLM agents for vision and language navigation in street view,” in Proc. AAAI Conf. Artif. Intell. (AAAI), vol. 38, no. 17, 2024, pp. 18 924– 18 933

    2024

  31. [31]

    Qwen Technical Report

    J. Bai, S. Bai, Y. Chu, Z. Cui, K. Dang, X. Deng, Y. Fan, W. Ge, Y. Han, F. Huang et al., “Qwen technical report,” arXiv preprint arXiv:2309.16609, 2023

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  32. [32]

    Explore until confident: Efficient exploration for em- bodied question answering,

    A. Z. Ren, J. Clark, A. Dixit, M. Itkina, A. Majumdar, and D. Sadigh, “Explore until confident: Efficient exploration for em- bodied question answering,” arXiv preprint arXiv:2403.15941, 2024

    work page arXiv 2024

  33. [33]

    Expand your SCOPE: Semantic cognition over JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2021 13 potential-based exploration for embodied visual navigation,

    N. Wang, W. Chen, L. Chen, H. Ji, Z. Guo, X. Zhang, and H. Sun, “Expand your SCOPE: Semantic cognition over JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2021 13 potential-based exploration for embodied visual navigation,” arXiv preprint arXiv:2511.08935, 2025

    work page arXiv 2021

  34. [34]

    Pointer networks,

    O. Vinyals, M. Fortunato, and N. Jaitly, “Pointer networks,” in Advances in Neural Information Processing Systems (NeurIPS), vol. 28, 2015

    2015

  35. [35]

    ARiADNE: A reinforcement learning approach using attention-based deep networks for exploration,

    Y. Cao, T. Hou, Y. Wang, X. Yi, and G. Sartoretti, “ARiADNE: A reinforcement learning approach using attention-based deep networks for exploration,” in Proc. IEEE Int. Conf. Robot. Autom. (ICRA). IEEE, 2023, pp. 10 219–10 225

    2023

  36. [36]

    Soft actor- critic: Off-policy maximum entropy deep reinforcement learning with a stochastic actor,

    T. Haarnoja, A. Zhou, P. Abbeel, and S. Levine, “Soft actor- critic: Off-policy maximum entropy deep reinforcement learning with a stochastic actor,” in Proc. Int. Conf. Mach. Learn. (ICML). PMLR, 2018, pp. 1861–1870

    2018

  37. [37]

    Self-learning explo- ration and mapping for mobile robots via deep reinforcement learning,

    F. Chen, S. Bai, T. Shan, and B. Englot, “Self-learning explo- ration and mapping for mobile robots via deep reinforcement learning,” in Proc. AIAA Scitech Forum, 2019, p. 0396

    2019

  38. [38]

    Autonomous exploration development environment and the planning algorithms,

    C. Cao, H. Zhu, F. Yang, Y. Xia, H. Choset, J. Oh, and J. Zhang, “Autonomous exploration development environment and the planning algorithms,” in Proc. IEEE Int. Conf. Robot. Autom. (ICRA). IEEE, 2022, pp. 8921–8928. Jintao Kong received the B.E. degree in Con- trol Science and Engineering from the College of Automation Engineering, Nanjing Univer- sity o...

    2022

  39. [39]

    candidate in the College of Artificial Intelligence at Xi’an Jiao- tong University

    He is currently a Ph.D. candidate in the College of Artificial Intelligence at Xi’an Jiao- tong University. His research interests include LLM-based RAG and VLM. Hongbin Sun (M’11-SM’19) received the B.S. and Ph.D. degree in electrical engineering from Xi’an Jiaotong University, Xi’an, China in 2003 and 2009, respectively. Currently he is a Professor of t...

    2003