Pith. sign in

REVIEW 3 major objections 6 minor 31 references

Navigating the Crowd: Non-linear MPC with Social Forces Dynamics for Human-Aware Robot Navigation

T0 review · 3 major / 6 minor · reviewed 2026-07-14 · grok-4.5

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

Pith's one-line read Embedding the Social Force Model inside nonlinear MPC lets a robot jointly predict human and robot motion and stay socially compliant at 20 Hz.

desk verdict Solid real-time engineering that jointly rolls out robot and SFM humans inside NMPC; the SW_step win is partly by construction, but the rest of the evidence still holds. read the letter →

arxiv 2607.10374 v1 pith:YKIJRFPA submitted 2026-07-11 cs.RO cs.HC

Stefano Trepella , Andrea Ostuni , Mauro Martini , Pablo Pueyo , Noé Pérez-Higueras , Marcello Chiaberge , Fernando Caballero , Luis Merino This is my paper · ORCID
classification cs.ROcs.HC
keywords socialnavigationmodelpredictivecontrolForcehuman-awareroboticsproxemicscrowdnonlinearMPC
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

Robots that share space with people must do more than avoid collisions: they must anticipate how people will move and keep a comfortable distance. This paper argues that a nonlinear model-predictive controller can do that job if the Social Force Model is placed inside the prediction itself rather than treated as a fixed external forecast. At every step the optimizer rolls out both the robot and the surrounding people under the same force dynamics, then chooses velocity commands that also minimize intrusion into personal space, mutual social force, and awkward head-on or crossing headings. The resulting controller, SFM-NMPC, runs at 20 Hz and, in crowded simulated rooms and corridors, beats a range of standard planners on social-work and minimum-distance metrics while still reaching the goal reliably. The claim matters because it shows that a classical, interpretable human model can be made predictive and real-time when it is co-evolved with the robot inside the control loop.

What carries the argument

SFM-NMPC: the Social Force Model appears both as the closed-loop dynamics that propagate every human agent inside the prediction horizon and as the source of the social-work cost term; the robot’s planned pose therefore directly shapes the predicted human reactions that the optimizer must satisfy.

What would settle it

Run the same open-space and mixed-corridor scenarios with real pedestrians whose motion is not generated by a force model; if measured social work and intimate-space intrusion rise above the simulated baselines while success rate drops, the central claim fails.

Watch

Extended reading notes

Core claim

By writing the Social Force Model into the multi-agent dynamics of a nonlinear MPC and adding three soft social costs (proxemics, social work, and heading), the robot can jointly forecast its own trajectory and the reactive trajectories of nearby people, producing socially compliant commands that improve measured comfort metrics over strong baselines while remaining real-time at 20 Hz.

Load-bearing premise

The Social Force Model is assumed to be a good enough closed-loop predictor of how real people will react to the robot’s planned motion over the next two seconds.

Editorial extensions

If this is right

  • Social navigation need not rely on precomputed human trajectories or large learned predictors; a classical force model can be co-optimized with the robot in real time.
  • The three social cost terms (proxemics, social work, heading) are modular: any other MPC dynamics model can adopt them and expect measurable comfort gains.
  • Real-time execution at 20 Hz with a multi-agent force model shows that increased prediction complexity need not force a lower control rate.
  • Visual trajectory evidence and ablation results become standard diagnostics for claiming social compliance, not just collision-free success.

Reading between the lines

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

  • Because the simulator’s humans are themselves force-based, the reported gains may shrink once perception noise and non-force human behavior appear on a physical robot.
  • Co-optimizing robot and human trajectories (the paper’s stated future direction) could turn the current open-loop force prediction into a true interactive game inside the horizon.
  • The same embedded-dynamics pattern could be reused with any other interpretable pedestrian model (e.g., ORCA or a simple constant-velocity plus social potential) without redesigning the cost structure.
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

3 major / 6 minor

Summary. The paper proposes SFM-NMPC, a nonlinear MPC local planner that embeds Helbing-style Social Force Model dynamics of surrounding agents inside the prediction model, so robot and human states are jointly propagated over a short horizon while optimizing unicycle velocity commands. A composite cost adds obstacle/costmap, goal/path, dynamic smoothness, and social terms (proxemics, social-work force magnitude, and heading long/cross penalties). The controller is implemented with Ceres, warm-started from regulated pure pursuit, and reported to run at 20 Hz. Evaluation is entirely in HuNavSim/Gazebo on open-space and mixed indoor maps (30 trials per scenario), comparing against Nav2 MPPI/DWB with Gaussian social costmaps, ORCA, SARL, pure SFM, and a non-social NMPC. The authors claim superior social compliance (especially SW_step and AMD) with high success rate, support this with Tukey HSD tests, proxemics zone occupancy, trajectory visualizations, and an ablation of social cost terms, and conclude suitability for real-world social navigation.

Significance. If the independent performance claims hold, this is a useful systems contribution to human-aware local planning: closed-loop multi-agent SFM prediction inside real-time NMPC is a concrete step beyond precomputed pedestrian forecasts, the reported 20 Hz rate with multi-agent dynamics is practically relevant, and the public code, multi-map protocol, statistical tests, and ablation strengthen reproducibility. Success rate in constrained mixed scenes, AMD, proxemics occupancy, and qualitative trajectory smoothness are meaningful even if SW_step is partly aligned with the objective. The advance is incremental engineering rather than a new theory of social navigation, but it is a credible addition to the MPC social-navigation literature provided the metric framing and sim-to-real caveats are tightened.

major comments (3)
  1. [Sec. IV-B, Eq. (9), Table I, Fig. 6] SW_step is defined (Sec. IV-B) as the average of robot-generated SFM social forces along the trajectory, using the same Helbing force structure that appears in the dynamics (Eqs. 3–5) and is explicitly minimized by J_work (Eq. 9). Table I and Sec. V-A then present the overall SW_step win (10.70 vs ~12.26 for the next-best methods) as a primary social-compliance advantage, and the ablation (Fig. 6) shows that adding J_work improves SW_step as expected. This makes the headline social-work result partly by construction rather than an external measure of comfort. AMD, proxemics (Fig. 5), success rate, path length/time, and visual interference with human paths are more independent and already favor the method in places; the manuscript should either demote SW_step to a diagnostic of the objective, report it only as an optimization residual, or add a clearly external social score (e.g., human-m
  2. [Sec. III-A, Sec. IV (HuNavSim), Abstract] All quantitative evidence is generated in HuNavSim, whose agents are themselves force-based, while the controller’s prediction constraints and J_work use the same SFM family (Sec. III-A, IV). Under that match, joint SFM prediction and social-work minimization are expected to look strong; the evaluation does not stress-test the weakest modeling assumption—that SFM is an adequate closed-loop predictor of how humans react to the robot over Tp = 2 s. The paper should (i) state this model-match limitation explicitly, (ii) ideally re-evaluate against at least one non-SFM pedestrian model or recorded trajectories, and (iii) avoid language that treats the current suite as sufficient confirmation of real-world social compliance. Without that, the central claim that the method is demonstrably better at social navigation remains only partially supported.
  3. [Abstract, Sec. VI] The Abstract and Conclusions assert effectiveness “for real-world social navigation” and suitability for real deployment, yet the manuscript contains no real-robot experiments, no perception noise study, and no human-subject comfort evaluation (real-robot work is deferred to future work in Sec. VI). For a journal claim of real-world readiness this is overreach. Either add hardware or human-in-the-loop results, or rewrite the abstract/conclusion claims to match the evidence: real-time simulated social navigation under HuNavSim with improved independent metrics (SR, AMD, proxemics, trajectory quality).
minor comments (6)
  1. [Sec. III-B] Cost weights α_obs, α_ρ, α_work, α_sθ, α_cross, etc., and several SFM scalars are deferred entirely to the public repository. For reproducibility in the archival version, include a table of the values used for the reported experiments (or a clear default set) in the main text or appendix.
  2. [Sec. III-A–B] Notation for agent sets and indices is dense (p^R_i, p^{A_k}_i, F_soc, etc.). A short symbol table would help readers track which quantities are optimized versus predicted.
  3. [Sec. II] Related work cites [25] as embedding SFM in MPC but treating humans as fixed initial predictions. A short quantitative or qualitative comparison to that class of methods (or a clear statement why it was not reimplemented) would sharpen the novelty claim of in-the-loop joint prediction.
  4. [Figs. 4–6] Fig. 4 caption mentions violin plots and Tukey HSD; Fig. 6 is described as a box plot of the ablation. Ensure all significance markers and which pairs are tested are fully specified in captions so the figures stand alone.
  5. [Table I, Sec. I–II] Minor wording/typos: e.g., “DWB GSC” / “MPPI GSC” spacing is inconsistent in Table I; “DWA” appears as “DWB” in places; “people’s trajectories are often precomputed even in already proposed social MPC formulations” is awkward. A careful copy-edit pass would help.
  6. [Sec. III-C, IV-A] Prediction/control horizons (Tp = 2 s, Δts = 0.1 s) are given for NMPC, but Tc and block length B are introduced without the numerical values used in the experiments; state them explicitly.

Circularity Check

1 steps flagged · score 4.0 of 10

SW_step (primary social metric) is defined as average SFM force and is exactly the quantity minimized by the optimized J_work cost, so the headline social-compliance win is partly by construction; other metrics remain independent.

  1. self definitional [Sec. IV-B (metrics) + Eq. 9 (J_work) + Table I / Sec. V-A]
    "the social work per step (SW step) ... evaluated as the average of the social forces generated by the robot along its trajectory, according to the formulation of the SFM [12]. ... J_work = α_work ∑_{i=1}^{N_p} (∑_{k=1}^{N_A} ||F_robot→agent k||² ) ... SFM-NMPC ... best SW_step ... overall evaluation."

    J_work explicitly minimizes the squared SFM forces that SW_step later averages; the controller is therefore optimized for the exact quantity used as the primary social-compliance score. The reported improvement on SW_step is therefore expected by construction once the cost is active (confirmed by the ablation that isolates J_work). AMD and navigation metrics do not share this identity.

full rationale

The paper's engineering contribution (embedding SFM dynamics inside NMPC + custom social costs, real-time 20 Hz, high SR, smooth trajectories) is self-contained and does not reduce to a definitional loop. The only clear circularity is evaluative: SW_step is defined (Sec. IV-B) via the identical Helbing SFM forces that appear in both the dynamics constraints (Eqs. 3-5) and the soft cost J_work (Eq. 9). Consequently the optimizer is directly rewarded for lowering the quantity later reported as the decisive social metric (Table I, Abstract, Sec. V-A). Ablation (Fig. 6) confirms the direct causal link. AMD, proxemics, path length, TTG and visual trajectory quality are independent of that term, and the comparison against external baselines remains informative inside the simulator. HuNavSim is a self-citation with author overlap, but it is used uniformly for all methods and does not itself force the ranking. No fitted-parameter-as-prediction, uniqueness theorem, or ansatz-smuggling chain exists. Score 4 reflects partial circularity confined to one advertised metric while the central algorithmic claim retains independent content.

Assumptions & free parameters 3 free parameters · 3 assumptions · 1 invented entities

The method rests on standard kinematic and force models plus a handful of hand-tuned weights and horizons; the only invented pieces are the particular social cost terms. No new physical entities are postulated. The free parameters and the SFM-as-truth axiom are the main load-bearing choices that a reader must accept.

free parameters (3)
  • social and navigation cost weights (α_obs, α_ρ, α_work, α_sθ, α_cross, …) = repository defaults
    Hand-chosen scalars that balance the multi-objective NLS; values live only in the public code, not derived.
  • prediction horizon Tp = 2 s, Δts = 0.1 s, control horizon Tc, block length B = Tp=2 s, Δts=0.1 s
    Discretization and horizon lengths that trade computation against foresight; chosen for 20 Hz real-time.
  • SFM force magnitudes A_soc, characteristic distances d0, ds, Softplus scales
    Parameters of the repulsive and heading costs; inherited from Helbing-style SFM or tuned for the experiments.
assumptions (3)
  • domain assumption Human agents obey the Social Force Model (repulsive + driving forces) even when interacting with the robot over the prediction horizon.
    Enters Eqs. 3–5 and the dynamics constraints of Eq. 6; never validated on real humans.
  • standard math Robot kinematics are unicycle with Euler integration (Eq. 2).
    Standard non-holonomic model used for local planning.
  • domain assumption Local costmap and global A* path supply a reliable reference; no online global re-planning is required.
    Stated in Sec. IV-A experimental protocol.
invented entities (1)
  • heading social costs Jsθlong and Jsθcross (Eqs. 10–11)
    purpose: Penalize head-on alignment and wrong-side crossing turns so the optimizer prefers socially polite maneuvers.
    Custom soft constraints not present in classical SFM or standard NMPC; ablation shows they drive most of the social gain.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Navigating the Crowd: Non-linear MPC with Social Forces Dynamics for Human-Aware Robot Navigation." pith.science (2026). https://pith.science/paper/YKIJRFPA

@misc{pith2026260710374,
  author       = {Pith},
  title        = {Pith review of: Navigating the Crowd: Non-linear MPC with Social Forces Dynamics for Human-Aware Robot Navigation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YKIJRFPA}},
  note         = {Machine review of arXiv:2607.10374}
}
read the original abstract

Safe and socially compliant navigation remains a fundamental challenge for autonomous robots operating in human-populated environments. Beyond collision avoidance, robots must anticipate human motion and respect personal space to ensure human comfort. Model Predictive Control (MPC) offers a robust alternative to classical and data-driven methods, although its effectiveness strongly depends on accurate human motion prediction and efficient computation. This paper introduces SFM-NMPC, a Social Force Model-based Non-linear Model Predictive Control framework that embeds human motion prediction directly within the optimization loop. By incorporating the Social Force Model into the dynamic model of surrounding agents, the controller jointly predicts the trajectories of humans and robots over the prediction horizon, thereby enabling socially-aware planning. A tailored set of social cost functions guides the optimization toward human-compliant behaviors. Despite the increased model complexity, the proposed formulation runs in real time at 20 Hz. Extensive simulated testing in crowded environments demonstrates that SFM-NMPC outperforms state-of-the-art baselines in social compliance metrics while maintaining efficient and smooth navigation. Visual trajectory analysis and an ablation study further highlight the contribution of the embedded SFM dynamics and social cost terms, confirming the effectiveness of the proposed approach for real-world social navigation.

Figures

Figures reproduced from arXiv: 2607.10374 by the authors.

Figure 1
Figure 1. SFM-NMPC embeds the Social Force Model in the prediction model of a Non-linear MPC controller as well as in the custom social cost terms to enhance human comfort and safety. Recent research on human-aware navigation has shown that socially compliant behavior improves human comfort and trust [5]. However, notable challenges have been high￾lighted for real-world deployment in crowded spaces [6]. In social contexts, ro… view at source ↗
Figure 2
Figure 2. The diagram of the social forces considered by the SFM[12]. where Fsoc represents the repulsive interaction forces and Fdrive is a relaxation force representing the agent’s resis￾tance to deviate from its current state. Crucially, Fsoc(p R i , p Ak i ) depends on the robot’s pose p R i and the agent pose p Ak i at the current predicted time step (see [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Maps used for the tests and trajectories of robots and humans across the different scenarios. Each algorithm was tested 30 times per scenario to ensure repeatability and consistent results. The experiments were performed on a workstation equipped with a Intel i9-12900K CPU and 64 GB of RAM. A. Baselines A wide set of relevant state-of-the-art algorithms for human-aware local planning has been selected as competitive… view at source ↗
Figures from the paper (3 more)
Figure 5
Figure 5. Figure 5: shows the proxemics scores: the percentage of time spent in each of Hall’s proxemic zones around humans [31], accumulated over the open space map, the mixed map, and overall. The SFM-NMPC demonstrates safe navigation and acceptable intrusion level, improving upon [PIT…
Figure 6
Figure 6. Figure 6: Box plot of the results on the open space crowded environment. SFM-NMPC exhibits the best social behavior while maintaining navigation efficiency. (*) indicate the p-values from the pairwise Tukey HSD test. C. Ablation study An ablation study is presented to investigat…
Figure 7
Figure 7. Figure 7: Comparison of trajectories of robot (red) and humans (blue) on the open space crowded map. Color shade indicates the temporal evolution of the experiment. Nonetheless, the major boost in social metrics, measured by the median value and the robustness (as indicated by t…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

31 extracted references · 1 linked inside Pith

  1. [1]

    Marvin: An innovative omni-directional robotic assistant for domestic environments,

    A. Eirale, M. Martini, L. Tagliavini, D. Gandini, M. Chiaberge, and G. Quaglia, “Marvin: An innovative omni-directional robotic assistant for domestic environments,”Sensors, vol. 22, no. 14, p. 5261, 2022

    2022

  2. [2]

    Service robots in the healthcare sector,

    J. Holland, L. Kingston, C. McCarthy, E. Armstrong, P. O’Dwyer, F. Merz, and M. McConnell, “Service robots in the healthcare sector,” Robotics, vol. 10, no. 1, p. 47, 2021

    2021

  3. [3]

    Social robotic wheelchair centered on passenger and pedestrian comfort,

    Y . Morales, T. Miyashita, and N. Hagita, “Social robotic wheelchair centered on passenger and pedestrian comfort,”Robotics and Au- tonomous Systems, vol. 87, pp. 355–362, 2017

    2017

  4. [4]

    Autonomous cars: Research results, issues, and future challenges,

    R. Hussain and S. Zeadally, “Autonomous cars: Research results, issues, and future challenges,”IEEE Communications Surveys & Tutorials, vol. 21, no. 2, pp. 1275–1313, 2018

    2018

  5. [5]

    Evolution of socially-aware robot navigation,

    S. Guill ´en-Ruiz, J. P. Bandera, A. Hidalgo-Paniagua, and A. Bandera, “Evolution of socially-aware robot navigation,”Electronics, vol. 12, no. 7, p. 1570, 2023

    2023

  6. [6]

    A survey on socially aware robot navigation: Taxonomy and future challenges,

    P. T. Singamaneni, P. Bachiller-Burgos, L. J. Manso, A. Garrell, A. Sanfeliu, A. Spalanzani, and R. Alami, “A survey on socially aware robot navigation: Taxonomy and future challenges,”The International Journal of Robotics Research, vol. 43, no. 10, pp. 1533–1572, 2024

    2024

  7. [7]

    The dynamic window approach to collision avoidance,

    D. Fox, W. Burgard, and S. Thrun, “The dynamic window approach to collision avoidance,”Robotics & Automation Magazine, IEEE, vol. 4, pp. 23 – 33, 04 1997

    1997

  8. [8]

    Sarl: Deep reinforcement learning based human-aware navigation for mobile robot in indoor environments,

    K. Li, Y . Xu, J. Wang, and M. Q.-H. Meng, “Sarl: Deep reinforcement learning based human-aware navigation for mobile robot in indoor environments,” in2019 IEEE International Conference on Robotics and Biomimetics (ROBIO), 2019, pp. 688–694

    2019

Show all 31 references
  1. [9]

    Social lstm: Human trajectory prediction in crowded spaces,

    A. Alahi, K. Goel, V . Ramanathan, A. Robicquet, L. Fei-Fei, and S. Savarese, “Social lstm: Human trajectory prediction in crowded spaces,” inProceedings of the IEEE conference on computer vision and pattern recognition, 2016, pp. 961–971

    2016

  2. [10]

    Predictive active steering control for autonomous vehicle systems,

    P. Falcone, F. Borrelli, J. Asgari, H. E. Tseng, and D. Hrovat, “Predictive active steering control for autonomous vehicle systems,” IEEE Transactions on control systems technology, vol. 15, no. 3, pp. 566–580, 2007

    2007

  3. [11]

    Motion planning among dynamic, decision-making agents with deep reinforcement learning,

    M. Everett, Y . F. Chen, and J. P. How, “Motion planning among dynamic, decision-making agents with deep reinforcement learning,” inIROS, 2018

    2018

  4. [12]

    Social force model for pedestrian dynamics

    Helbing and Moln ´ar, “Social force model for pedestrian dynamics.” Physical review. E, Statistical physics, plasmas, fluids, and related interdisciplinary topics, vol. 51 5, pp. 4282–4286, 1995

    1995

  5. [13]

    Modelling pedestrian trajectory patterns with gaussian processes,

    D. Ellis, E. Sommerlade, and I. Reid, “Modelling pedestrian trajectory patterns with gaussian processes,” in2009 IEEE 12th international conference on computer vision workshops, ICCV workshops. IEEE, 2009, pp. 1229–1234

    2009

  6. [14]

    Planning- based prediction for pedestrians,

    B. D. Ziebart, N. Ratliff, G. Gallagher, C. Mertz, K. Peterson, J. A. Bagnell, M. Hebert, A. K. Dey, and S. Srinivasa, “Planning- based prediction for pedestrians,” in2009 IEEE/RSJ International Conference on Intelligent Robots and Systems. IEEE, 2009, pp. 3931– 3936

    2009

  7. [15]

    Reciprocal velocity obsta- cles for real-time multi-agent navigation,

    J. Van den Berg, M. Lin, and D. Manocha, “Reciprocal velocity obsta- cles for real-time multi-agent navigation,” in2008 IEEE international conference on robotics and automation. Ieee, 2008, pp. 1928–1935

    2008

  8. [16]

    Sicnav: Safe and interactive crowd navigation using model predictive control and bilevel optimization,

    S. Samavi, J. R. Han, F. Shkurti, and A. P. Schoellig, “Sicnav: Safe and interactive crowd navigation using model predictive control and bilevel optimization,”IEEE Transactions on Robotics, vol. 41, no. 3, pp. 801–818, 2024

    2024

  9. [17]

    Glmp- realtime pedestrian path prediction using global and local movement patterns,

    A. Bera, S. Kim, T. Randhavane, S. Pratapa, and D. Manocha, “Glmp- realtime pedestrian path prediction using global and local movement patterns,” in2016 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2016, pp. 5528–5535

    2016

  10. [18]

    Human motion trajectory prediction: A survey,

    A. Rudenko, L. Palmieri, M. Herman, K. M. Kitani, D. M. Gavrila, and K. O. Arras, “Human motion trajectory prediction: A survey,” The International Journal of Robotics Research, vol. 39, no. 8, pp. 895–935, 2020

    2020

  11. [19]

    Evaluation of socially-aware robot navi- gation,

    Y . Gao and C.-M. Huang, “Evaluation of socially-aware robot navi- gation,”Frontiers in Robotics and AI, vol. 8, p. 721317, 2022

    2022

  12. [20]

    Adaptive social force window planner with rein- forcement learning,

    M. Martini, N. P ´erez-Higueras, A. Ostuni, M. Chiaberge, F. Caballero, and L. Merino, “Adaptive social force window planner with rein- forcement learning,” in2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, 2024, pp. 4816–4822

    2024

  13. [21]

    Constrained model predictive control: Stability and optimality,

    D. Q. Mayne, J. B. Rawlings, C. V . Rao, and P. O. M. Scokaert, “Constrained model predictive control: Stability and optimality,”Au- tomatica, 2000

    2000

  14. [22]

    Topology-driven model predictive control for safe navigation in dynamic environments,

    H. De Groot, J. Smith, and W. Li, “Topology-driven model predictive control for safe navigation in dynamic environments,” inIEEE Inter- national Conference on Robotics and Automation (ICRA), 2024, pp. 1234–1240

    2024

  15. [23]

    Shine: Social homology identification for navigation in crowded environments,

    D. Martinez-Baselga, O. de Groot, L. Knoedler, L. Riazuelo, J. Alonso-Mora, and L. Montano, “Shine: Social homology identification for navigation in crowded environments,”The International Journal of Robotics Research, vol. 45, no. 1, pp. 60–79, 2026. [Online]. Available: http...

    2026

  16. [24]

    Dr-mpc: Deep residual model predictive control for real-world social navigation,

    J. R. Han, H. Thomas, J. Zhang, N. Rhinehart, and T. D. Barfoot, “Dr-mpc: Deep residual model predictive control for real-world social navigation,”IEEE Robotics and Automation Letters, 2025

    2025

  17. [25]

    Towards autonomous robot navigation in human populated environments using an universal sfm and parametrized mpc,

    E. Fiasch ´e, P. Martinet, and E. Malis, “Towards autonomous robot navigation in human populated environments using an universal sfm and parametrized mpc,” in2023 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, 2023, pp. 7422– 7427

    2023

  18. [26]

    Ceres solver,

    S. Agarwal, K. Mierle, and T. C. S. Team, “Ceres solver,” 10 2023. [Online]. Available: https://github.com/ceres-solver/ceres-solver

    2023

  19. [27]

    Hunavsim: A ros 2 human navigation simulator for benchmarking human-aware robot navigation,

    N. P ´erez-Higueras, R. Otero, F. Caballero, and L. Merino, “Hunavsim: A ros 2 human navigation simulator for benchmarking human-aware robot navigation,”IEEE robotics and automation letters, vol. 8, no. 11, pp. 7130–7137, 2023

    2023

  20. [28]

    Model predictive path integral control using covariance variable importance sampling,

    G. Williams, A. Aldrich, and E. A. Theodorou, “Model predictive path integral control using covariance variable importance sampling,” CoRR, vol. abs/1509.01149, 2015. [Online]. Available: http://arxiv. org/abs/1509.01149

    2015 arXiv

  21. [29]

    The marathon 2: A navigation system,

    S. Macenski, F. Mart ´ın, R. White, and J. G. Clavero, “The marathon 2: A navigation system,” in2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, 2020, pp. 2718–2725

    2020

  22. [30]

    NA V2 social costmap plugin ros2,

    “NA V2 social costmap plugin ros2,” https://github.com/robotics-upo/ nav2 social costmap plugin, accessed: 2026-03-02

    2026

  23. [31]

    E. T. Hall,The Hidden Dimension, 1st ed. Garden City, NY , USA: Doubleday & Company, Inc., 1966

    1966

Pith tools

Reviewed July 14, 2026 · model on record in the stance chip above.