pith. sign in

arxiv: 2508.07267 · v1 · submitted 2025-08-10 · 💻 cs.RO

Bio-Inspired Topological Autonomous Navigation with Active Inference in Robotics

Daria de Tinguy , Tim Verbelen , Emilio Gamba , Bart Dhoedt This is my paper

Pith reviewed 2026-05-18 23:39 UTC · model grok-4.3

classification 💻 cs.RO
keywords active inferencetopological mappingautonomous navigationrobot explorationROS2 architecturebio-inspired roboticsdynamic environments
0
0 comments X

The pith

A bio-inspired active inference model enables real-time topological mapping and adaptive navigation in unknown environments without pre-training.

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

The paper introduces an agent based on the Active Inference Framework that unifies localization, mapping, and decision-making for autonomous robot navigation. It creates and maintains a topological map in real time to plan trajectories for exploration or reaching goals. This approach avoids reliance on pre-training or large datasets, making it suitable for dynamic and unknown settings. A sympathetic reader would care because it offers a transparent and adaptable alternative to computationally heavy AI methods or rigid rule-based systems.

Core claim

The Active Inference Framework can be implemented in a modular ROS2 architecture to create and update a topological map of the environment in real-time, enabling goal-directed planning for exploration and navigation that adapts to dynamic obstacles and sensor drift, performing comparably to existing strategies like Gbplanner, FAEL, and Frontiers in simulated large-scale environments.

What carries the argument

Active Inference Framework (AIF) for probabilistic reasoning that unifies mapping, localization, and adaptive decision-making to maintain a usable topological map.

If this is right

  • The agent can explore large-scale simulated environments in real time.
  • It adapts to dynamic obstacles and sensor drift without retraining.
  • The modular ROS2 design integrates with existing navigation systems.
  • It provides interpretable probabilistic reasoning for navigation decisions.
  • Performance matches or approaches that of Gbplanner, FAEL, and Frontiers in tests.

Where Pith is reading between the lines

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

  • Such a system might lower computational demands compared to deep learning based navigation.
  • Extensions could include handling multi-robot coordination using shared topological maps.
  • Further testing in unstructured real-world settings could reveal scalability limits.
  • Integration with other sensory modalities might enhance robustness to uncertainty.

Load-bearing premise

The Active Inference Framework can be directly translated into a real-time, modular ROS2 implementation that maintains a usable topological map and produces adaptive decisions in dynamic, unknown environments without any pre-training or large datasets.

What would settle it

Observing that the agent fails to generate coherent trajectories or maintain map consistency when faced with moving obstacles or localization drift in a large simulated environment would falsify the central claim.

Figures

Figures reproduced from arXiv: 2508.07267 by Bart Dhoedt, Daria de Tinguy, Emilio Gamba, Tim Verbelen.

Figure 2
Figure 2. Figure 2: Final map of exploration in a) a real-world en￾vironment, b) Amazon simulated warehouse. Coloured points signify visited locations, where the same colour attributions mean the same observation. The thickness of the lines depicts the agent’s believed probability of tran￾sitioning between two states given an action. To recognise locations, in this specific architecture, the agent builds a 360° panorama from … view at source ↗
Figure 3
Figure 3. Figure 3: Obstacle movement and its impact on the agent’s internal map during exploration. (state 1 instead of 3) because the agent cannot correct its belief by moving to that location. A new state (state 20) is created at the former position of the obstacle, and new transitions are established. This change does not affect any of the other existing states. 3.5 Agent Architecture With our method established, we now t… view at source ↗
Figure 4
Figure 4. Figure 4: Overview of the system architecture. In grey, we have modules that any ROS-compatible solution can re￾place. Modules interact through belief propagation, In￾ferring and planning (localisation, mapping and action selection) rely on the AIF framework. The perceptual and motion planning still use traditional approaches. Be￾lieved odometry takes precedence over sensor odometry. Preferences (goal) are expected … view at source ↗
Figure 5
Figure 5. Figure 5: Exploration efficiency in distance over the whole area of the warehouse by our model, Frontiers, FAEL and Gbplanner. The agent explored the 280m2 warehouse from five starting points using four different approaches: (1) our proposed method combining mapping, localisation, and planning with Nav2 for motion control; (2) the Frontier￾based exploration algorithm [1], implemented with Nav2 SLAM [36]; (3) FAEL [3… view at source ↗
Figure 7
Figure 7. Figure 7 [PITH_FULL_IMAGE:figures/full_fig_p006_7.png] view at source ↗
Figure 6
Figure 6. Figure 6: Drift between the real odometry (dashed red) and the agent-believed odometry (continuous blue) during a home exploration. The drift is due to a furniture collision and location approximation. 4.2 Reaching given objective The model can be given goals either as positions or RGB observations. It first determines whether the goal is familiar, then either explores or navigates directly to it. In these trials, R… view at source ↗
read the original abstract

Achieving fully autonomous exploration and navigation remains a critical challenge in robotics, requiring integrated solutions for localisation, mapping, decision-making and motion planning. Existing approaches either rely on strict navigation rules lacking adaptability or on pre-training, which requires large datasets. These AI methods are often computationally intensive or based on static assumptions, limiting their adaptability in dynamic or unknown environments. This paper introduces a bio-inspired agent based on the Active Inference Framework (AIF), which unifies mapping, localisation, and adaptive decision-making for autonomous navigation, including exploration and goal-reaching. Our model creates and updates a topological map of the environment in real-time, planning goal-directed trajectories to explore or reach objectives without requiring pre-training. Key contributions include a probabilistic reasoning framework for interpretable navigation, robust adaptability to dynamic changes, and a modular ROS2 architecture compatible with existing navigation systems. Our method was tested in simulated and real-world environments. The agent successfully explores large-scale simulated environments and adapts to dynamic obstacles and drift, proving to be comparable to other exploration strategies such as Gbplanner, FAEL and Frontiers. This approach offers a scalable and transparent approach for navigating complex, unstructured environments.

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 / 2 minor

Summary. The paper introduces a bio-inspired autonomous navigation system based on the Active Inference Framework (AIF) that unifies mapping, localisation, decision-making and motion planning. It creates and updates a topological map in real time to enable exploration and goal-directed trajectories without pre-training or large datasets. The approach is implemented as a modular ROS2 architecture and is tested in simulated large-scale environments and real-world settings, with claims of adaptability to dynamic obstacles and sensor drift, and performance comparable to Gbplanner, FAEL and Frontiers.

Significance. If the real-time performance and experimental claims hold, the work would offer a meaningful contribution by demonstrating a unified, interpretable AIF-based framework for navigation in unknown and dynamic environments that avoids data-intensive pre-training. The modular ROS2 design is a practical strength for compatibility with existing systems. The absence of detailed quantitative metrics and scaling analysis in the evaluation sections, however, limits the current strength of the comparability and feasibility assertions.

major comments (2)
  1. [Results] Results section: The claim that the agent 'proves to be comparable' to Gbplanner, FAEL and Frontiers is unsupported by quantitative metrics (e.g., exploration coverage, time to goal, success rates), error bars, or statistical comparisons. This directly weakens the evaluation of the method against baselines.
  2. [Methods] Methods/Implementation: No details are provided on per-cycle timing, computational scaling of variational free-energy minimisation with growing topological node count, or the specific factorisation of the generative model over map states. This leaves the central real-time feasibility claim for large-scale environments unverified.
minor comments (2)
  1. [Abstract] Abstract: The phrasing 'proving to be comparable' should be softened to 'demonstrating performance comparable to' pending quantitative support.
  2. [Experiments] Ensure that all experimental environments (simulated and real-world) are described with sufficient detail for reproducibility, including map sizes and obstacle dynamics.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive and detailed feedback on our manuscript. We have carefully reviewed each major comment and provide point-by-point responses below. We agree that additional quantitative support and implementation details will strengthen the paper and plan to incorporate these in the revised version.

read point-by-point responses

  1. Referee: [Results] Results section: The claim that the agent 'proves to be comparable' to Gbplanner, FAEL and Frontiers is unsupported by quantitative metrics (e.g., exploration coverage, time to goal, success rates), error bars, or statistical comparisons. This directly weakens the evaluation of the method against baselines.

    Authors: We agree that the comparability claim would be more robust with explicit quantitative metrics. The current manuscript focuses on demonstrating real-time adaptability to dynamic obstacles and sensor drift through qualitative results in simulation and real-world tests. In the revised manuscript, we will add a dedicated comparison table reporting exploration coverage, time to goal, success rates across repeated trials, standard deviations, and basic statistical comparisons against the baselines to better substantiate the evaluation. revision: yes

  2. Referee: [Methods] Methods/Implementation: No details are provided on per-cycle timing, computational scaling of variational free-energy minimisation with growing topological node count, or the specific factorisation of the generative model over map states. This leaves the central real-time feasibility claim for large-scale environments unverified.

    Authors: We acknowledge that more explicit implementation details are needed to verify real-time performance. The topological map is intentionally kept sparse to support scalability, but we will expand the Methods section in the revision to report measured per-cycle timings from our experiments, an analysis of how variational free-energy minimisation scales with node count (including observed growth rates in large environments), and the precise factorisation of the generative model over map states and actions. These additions will directly address the feasibility claims. revision: yes

Circularity Check

0 steps flagged

No circularity: application of established AIF framework to topological navigation

full rationale

The paper applies the Active Inference Framework—an externally established formalism—to a robotics navigation task by describing a modular ROS2 implementation that builds and updates a topological map in real time. No equations or derivations in the provided text reduce a claimed prediction or result to a fitted parameter or self-referential definition by construction. The central claims rest on implementation details and empirical comparison to baselines (Gbplanner, FAEL, Frontiers) rather than on any load-bearing self-citation chain or ansatz smuggled from prior author work. The derivation chain is therefore self-contained against external benchmarks and receives the default non-circularity finding.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim depends on the Active Inference Framework being sufficient to unify mapping, localisation and decision-making in robotics; this is drawn from prior literature rather than derived here.

axioms (1)
  • domain assumption Active Inference Framework unifies mapping, localisation, and adaptive decision-making for autonomous navigation
    Invoked in the abstract as the basis for the bio-inspired agent.

pith-pipeline@v0.9.0 · 5737 in / 1190 out tokens · 32408 ms · 2026-05-18T23:39:42.093804+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

50 extracted references · 50 canonical work pages · 2 internal anchors

  1. [1]

    Frontier based exploration for autonomous robot,

    A. Topiwala, P. Inani, and A. Kathpal, “Frontier based exploration for autonomous robot,” 2018

    work page 2018

  2. [2]

    Past, present and future of path- planning algorithms for mobile robot navigation in dynamic environments,

    H. S. Hewawasam, M. Y . Ibrahim, and G. K. Ap- puhamillage, “Past, present and future of path- planning algorithms for mobile robot navigation in dynamic environments,” IEEE Open Journal of the Industrial Electronics Society , vol. 3, pp. 353–365, 2022

    work page 2022

  3. [3]

    ORB-SLAM3: An accurate open-source library for visual, visual-inertial and multi-map SLAM,

    C. Campos, R. Elvira, J. J. Gomez, J. M. M. Mon- tiel, and J. D. Tardos, “ORB-SLAM3: An accurate open-source library for visual, visual-inertial and multi-map SLAM,”IEEE Transactions on Robotics, vol. 37, no. 6, pp. 1874–1890, 2021

    work page 2021

  4. [4]

    Gaussian splatting slam,

    H. Matsuki, R. Murai, P. H. J. Kelly, and A. J. Davi- son, “Gaussian splatting slam,” 2024

    work page 2024

  5. [5]

    Learning to explore using active neural slam,

    D. S. Chaplot, D. Gandhi, S. Gupta, A. Gupta, and R. Salakhutdinov, “Learning to explore using active neural slam,” inInternational Conference on Learn- ing Representations (ICLR), 2020

    work page 2020

  6. [6]

    Byol-explore: Exploration by boot- strapped prediction,

    Z. D. Guo, S. Thakoor, M. P ˆıslar, B. A. Pires, F. Altch´e, C. Tallec, A. Saade, D. Calandriello, J.- B. Grill, Y . Tang, M. Valko, R. Munos, M. G. Azar, and B. Piot, “Byol-explore: Exploration by boot- strapped prediction,” 2022

    work page 2022

  7. [7]

    Rapid exploration for open-world navi- gation with latent goal models,

    D. Shah, B. Eysenbach, G. Kahn, N. Rhinehart, and S. Levine, “Rapid exploration for open-world navi- gation with latent goal models,” 2023

    work page 2023

  8. [8]

    Ratslam: a hippocampal model for simultaneous localization and mapping,

    M. Milford, G. Wyeth, and D. Prasser, “Ratslam: a hippocampal model for simultaneous localization and mapping,” inIEEE International Conference on Robotics and Automation, 2004. Proceedings. ICRA ’04. 2004, vol. 1, pp. 403–408 V ol.1, 2004

    work page 2004

  9. [9]

    Safron, O

    A. Safron, O. C ¸ atal, and T. Verbelen, “Generalized simultaneous localization and mapping (g-slam) as unification framework for natural and artificial in- telligences: towards reverse engineering the hip- pocampal/entorhinal system and principles of high- level cognition,”Frontiers in Systems Neuroscience, vol. V olume 16 - 2022, 2022

    work page 2022

  10. [10]

    Learn- ing dynamic cognitive map with autonomous nav- igation,

    D. de Tinguy, T. Verbelen, and B. Dhoedt, “Learn- ing dynamic cognitive map with autonomous nav- igation,” Frontiers in Computational Neuroscience, vol. 18, Dec. 2024

    work page 2024

  11. [11]

    The dynamic window approach to collision avoidance,

    D. Fox, W. Burgard, and S. Thrun, “The dynamic window approach to collision avoidance,” IEEE Robotics and Automation Magazine , vol. 4, no. 1, pp. 23–33, 1997

    work page 1997

  12. [12]

    FAST- LIO2: fast direct lidar-inertial odometry,

    W. Xu, Y . Cai, D. He, J. Lin, and F. Zhang, “FAST- LIO2: fast direct lidar-inertial odometry,” CoRR, vol. abs/2107.06829, 2021

    work page arXiv 2021

  13. [13]

    Autonomous navigation of agvs in unknown cluttered environ- ments: Log-mppi control strategy,

    I. S. Mohamed, K. Yin, and L. Liu, “Autonomous navigation of agvs in unknown cluttered environ- ments: Log-mppi control strategy,” IEEE Robotics and Automation Letters , vol. 7, no. 4, pp. 10240– 10247, 2022

    work page 2022

  14. [14]

    Towards efficient mppi trajectory genera- tion with unscented guidance: U-mppi control strat- egy,

    I. S. Mohamed, J. Xu, G. S. Sukhatme, and L. Liu, “Towards efficient mppi trajectory genera- tion with unscented guidance: U-mppi control strat- egy,” 2024

    work page 2024

  15. [15]

    Etpnav: Evolving topological planning for vision-language navigation in continu- ous environments,

    D. An, H. Wang, W. Wang, Z. Wang, Y . Huang, K. He, and L. Wang, “Etpnav: Evolving topological planning for vision-language navigation in continu- ous environments,” 2024

    work page 2024

  16. [16]

    T. Parr, G. Pezzulo, and K. Friston, Active Infer- ence: The Free Energy Principle in Mind, Brain, and Behavior. The MIT Press, 03 2022

    work page 2022

  17. [17]

    Planning and navigation as active inference,

    R. Kaplan and K. Friston, “Planning and navigation as active inference,”bioRxiv, 12 2017

    work page 2017

  18. [18]

    Computational mechanisms of curiosity and goal- directed exploration,

    P. Schwartenbeck, J. Passecker, T. U. Hauser, T. H. FitzGerald, M. Kronbichler, and K. J. Friston, “Computational mechanisms of curiosity and goal- directed exploration,” eLife, vol. 8, p. e41703, may 2019

    work page 2019

  19. [19]

    Robot navigation as hierarchical ac- tive inference,

    O. C ¸ atal, T. Verbelen, T. Van de Maele, B. Dhoedt, and A. Safron, “Robot navigation as hierarchical ac- tive inference,”Neural Networks, vol. 142, pp. 192– 204, 2021

    work page 2021

  20. [20]

    Clone- structured graph representations enable flexible learning and vicarious evaluation of cognitive maps,

    D. George, R. Rikhye, N. Gothoskar, J. S. Guntu- palli, A. Dedieu, and M. L ´azaro-Gredilla, “Clone- structured graph representations enable flexible learning and vicarious evaluation of cognitive maps,” Nature Communications, vol. 12, 04 2021

    work page 2021

  21. [21]

    Structuring knowledge with cogni- tive maps and cognitive graphs,

    M. Peer, I. K. Brunec, N. S. Newcombe, and R. A. Epstein, “Structuring knowledge with cogni- tive maps and cognitive graphs,” Trends in Cogni- tive Sciences, vol. 25, no. 1, pp. 37–54, 2021

    work page 2021

  22. [22]

    Do humans integrate routes into a cognitive map? map- versus landmark-based navigation of novel short- cuts.,

    P. Foo, W. Warren, A. Duchon, and M. Tarr, “Do humans integrate routes into a cognitive map? map- versus landmark-based navigation of novel short- cuts.,” Journal of experimental psychology. Learn- ing, memory, and cognition , vol. 31, pp. 195–215, 04 2005

    work page 2005

  23. [23]

    The cognitive map in humans: Spatial navigation and beyond,

    R. Epstein, E. Z. Patai, J. Julian, and H. Spiers, “The cognitive map in humans: Spatial navigation and beyond,” Nature Neuroscience, vol. 20, pp. 1504– 1513, 10 2017

    work page 2017

  24. [24]

    A survey of monte carlo tree search methods,

    C. B. Browne, E. Powley, D. Whitehouse, S. M. Lucas, P. I. Cowling, P. Rohlfshagen, S. Tavener, D. Perez, S. Samothrakis, and S. Colton, “A survey of monte carlo tree search methods,” IEEE Trans- actions on Computational Intelligence and AI in Games, vol. 4, no. 1, pp. 1–43, 2012

    work page 2012

  25. [25]

    World Models

    D. Ha and J. Schmidhuber, “World models,” CoRR, vol. abs/1803.10122, 2018

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  26. [26]

    Active in- ference and learning,

    K. Friston, T. FitzGerald, F. Rigoli, P. Schwarten- beck, J. O. Doherty, and G. Pezzulo, “Active in- ference and learning,” Neuroscience and Biobehav- ioral Reviews, vol. 68, pp. 862–879, 2016

    work page 2016

  27. [27]

    aws-robomaker-small-warehouse- world,

    aws-robotics, “aws-robomaker-small-warehouse- world,” 2020. Accessed: 2024-08-01

    work page 2020

  28. [28]

    Neural topological SLAM for visual navigation.CoRR, abs/2005.12256, 2020

    D. S. Chaplot, R. Salakhutdinov, A. Gupta, and S. Gupta, “Neural topological SLAM for visual navigation,”CoRR, vol. abs/2005.12256, 2020

    work page arXiv 2005

  29. [29]

    World model learning and inference,

    K. Friston, R. J. Moran, Y . Nagai, T. Taniguchi, H. Gomi, and J. Tenenbaum, “World model learning and inference,”Neural Networks, vol. 144, pp. 573– 590, 2021

    work page 2021

  30. [30]

    Ros2 humble,

    ROS2, “Ros2 humble,” 2022. Accessed: 2025-05- 21

    work page 2022

  31. [31]

    Turtlebot versions,

    Turtlebot, “Turtlebot versions,” 2024. Accessed: 2024-12-16

    work page 2024

  32. [32]

    rosbotxl,

    Husarion, “rosbotxl,” 2025. Accessed: 2025-05-21

    work page 2025

  33. [33]

    Accessed: 2024-12-01

    nav2, “nav2,” 2021. Accessed: 2024-12-01

    work page 2021

  34. [34]

    Potential field meth- ods and their inherent limitations for mobile robot navigation,

    Y . Koren and J. Borenstein, “Potential field meth- ods and their inherent limitations for mobile robot navigation,” vol. 2, pp. 1398 – 1404 vol.2, 05 1991

    work page 1991

  35. [35]

    aws-robomaker-small-house- world,

    aws-robotics, “aws-robomaker-small-house- world,” 2021. Accessed: 2024-10-01

    work page 2021

  36. [36]

    Au- tonomous mobile vehicle using ros2 and 2d- lidar and slam navigation,

    P. Gyanani, M. Agarwal, R. Osari, et al. , “Au- tonomous mobile vehicle using ros2 and 2d- lidar and slam navigation,” Research Square , vol. Preprint (Version 1), May 2024. Available at Research Square

    work page 2024

  37. [37]

    Fael: Fast autonomous exploration for large-scale environments with a mobile robot,

    J. Huang, B. Zhou, Z. Fan, Y . Zhu, Y . Jie, L. Li, and H. Cheng, “Fael: Fast autonomous exploration for large-scale environments with a mobile robot,” IEEE Robotics and Automation Letters , vol. 8, pp. 1667–1674, 2023

    work page 2023

  38. [38]

    UFOMap: An efficient probabilistic 3D mapping framework that embraces the unknown,

    D. Duberg and P. Jensfelt, “UFOMap: An efficient probabilistic 3D mapping framework that embraces the unknown,” IEEE Robotics and Automation Let- ters, vol. 5, no. 4, pp. 6411–6418, 2020

    work page 2020

  39. [39]

    Graph-based subter- ranean exploration path planning using aerial and legged robots,

    T. Dang, M. Tranzatto, S. Khattak, F. Mascarich, K. Alexis, and M. Hutter, “Graph-based subter- ranean exploration path planning using aerial and legged robots,” Journal of Field Robotics , vol. 37, no. 8, pp. 1363–1388, 2020. Wiley Online Library

    work page 2020

  40. [40]

    Team cer- berus wins the darpa subterranean challenge: Tech- nical overview and lessons learned,

    M. Tranzatto, M. Dharmadhikari, L. Bernreiter, M. Camurri, S. Khattak, F. Mascarich, P. Pfre- undschuh, D. Wisth, S. Zimmermann, M. Kulka- rni, V . Reijgwart, B. Casseau, T. Homberger, P. D. Petris, L. Ott, W. Tubby, G. Waibel, H. Nguyen, C. Cadena, R. Buchanan, L. Well- hausen, N. Khedekar, O. Andersson, L. Zhang, T. Miki, T. Dang, M. Mattamala, M. Monte...

    work page 2022

  41. [41]

    Voxblox: Incremental 3D Euclidean Signed Distance Fields for On-Board MAV Planning

    H. Oleynikova, Z. Taylor, M. Fehr, J. I. Nieto, and R. Siegwart, “V oxblox: Building 3d signed distance fields for planning,” CoRR, vol. abs/1611.03631, 2016

    work page internal anchor Pith review Pith/arXiv arXiv 2016

  42. [42]

    A goal-directed spatial navigation model using forward trajectory planning based on grid cells,

    U. M. Erdem and M. Hasselmo, “A goal-directed spatial navigation model using forward trajectory planning based on grid cells,” European Journal of Neuroscience, vol. 35, no. 6, pp. 916–931, 2012

    work page 2012

  43. [43]

    Active inference and intentional be- haviour,

    K. J. Friston, T. Salvatori, T. Isomura, A. Tschantz, A. Kiefer, T. Verbelen, M. Koudahl, A. Paul, T. Parr, A. Razi, B. Kagan, C. L. Buckley, and M. J. D. Ramstead, “Active inference and intentional be- haviour,” 2023

    work page 2023

  44. [44]

    Over- coming exploration: Deep reinforcement learning for continuous control in cluttered environments from temporal logic specifications,

    M. Cai, E. Aasi, C. Belta, and C.-I. Vasile, “Over- coming exploration: Deep reinforcement learning for continuous control in cluttered environments from temporal logic specifications,” IEEE Robotics and Automation Letters , vol. 8, no. 4, pp. 2158– 2165, 2023

    work page 2023

  45. [45]

    How to build a cogni- tive map,

    J. C. R. Whittington, D. McCaffary, J. J. W. Baker- mans, and T. E. J. Behrens, “How to build a cogni- tive map,” Nature Neuroscience, vol. 25, pp. 1257– 1272, October 2022. Epub 2022 Sep 26

    work page 2022

  46. [46]

    An active inference approach to model- ing structure learning: Concept learning as an ex- ample case,

    R. Smith, P. Schwartenbeck, T. Parr, and K. J. Friston, “An active inference approach to model- ing structure learning: Concept learning as an ex- ample case,” Frontiers in Computational Neuro- science, vol. 14, 2020

    work page 2020

  47. [47]

    Bayesian model reduction,

    K. Friston, T. Parr, and P. Zeidman, “Bayesian model reduction,” 2019

    work page 2019

  48. [48]

    Ex- ploring and learning structure: Active inference ap- proach in navigational agents,

    D. de Tinguy, T. Verbelen, and B. Dhoedt, “Ex- ploring and learning structure: Active inference ap- proach in navigational agents,” 2024

    work page 2024

  49. [49]

    Object goal navigation using goal-oriented semantic exploration,

    D. S. Chaplot, D. Gandhi, A. Gupta, and R. Salakhutdinov, “Object goal navigation using goal-oriented semantic exploration,” 2020

    work page 2020

  50. [50]

    Spatial and tempo- ral hierarchy for autonomous navigation using ac- tive inference in minigrid environment,

    de Tinguy, Daria and Van de Maele, Toon and Ver- belen, Tim and Dhoedt, Bart, “Spatial and tempo- ral hierarchy for autonomous navigation using ac- tive inference in minigrid environment,”ENTROPY, vol. 26, no. 1, p. 32, 2024

    work page 2024