Pith. sign in

REVIEW 4 major objections 8 minor 37 references

Online Robot Motion Planning Methodology Guided by Group Social Proxemics Feature

T0 review · 4 major / 8 minor · reviewed 2026-08-08 · deepseek-v4-flash

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

Pith's one-line read A robot can cruise among human groups by navigating a vector field assembled from each person's social space.

desk verdict A real-robot social navigation pipeline with a genuinely new proxemics representation, but the OOP computation and model constants need validation before the 'optimal' and accuracy claims hold up. read the letter →

arxiv 2502.04837 v1 pith:EF5FU2YU submitted 2025-02-07 cs.RO cs.SYeess.SY

Xuan Mu , Xiaorui Liu , Shuai Guo , Wenzheng Chi , Wei Wang , Shuzhi Sam Ge This is my paper · ORCID
classification cs.ROcs.SYeess.SY
keywords socialroboticsproxemicsgroupclusteringmagneticdipolemodeloptimalobservationpositionsampling-basedpathplanninghuman-robotinteractionnavigation
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper claims that a service robot can navigate among human groups in a socially aware way by representing each person's personal space as a two-dimensional vector field shaped like a magnetic dipole and adding these fields together for groups and the whole scene. It introduces a graph-based clustering step that groups people using both social relevance and a spatial-confidence score, then defines the best place to stand near each group, the optimal observation position (OOP), as the point where the group's total field is exactly cancelled by a social damping vector. The planner, called VMD-RRT*, steers the robot with a cost that rewards moving along the field while keeping out of the groups' social gaze space. In simulations and on a physical service robot with volunteers, the robot re-clusters people as the scene changes and visits each group's OOP at a socially comfortable distance. The paper's point is that proxemics can be a planning input for interaction, not just an avoidance constraint.

What carries the argument

The load-bearing object is the magnetic-dipole social proxemics field: a two-dimensional vector field $\mathcal{M}$ defined in Eqs. (7)-(8) with coefficient $\alpha=-615$, rotated by the person's heading to form the individual's social space $\boldsymbol{S}$. Group and scenario maps are built by linear superposition ($\boldsymbol{V}_{g_k} = \sum_j \boldsymbol{S}_j$, global field $\boldsymbol{V}_{gC} = \sum_k \boldsymbol{V}_{g_k}$). The optimal observation position is defined by the zero-damping condition $\boldsymbol{V}_{g_k}(O_k) + \boldsymbol{\xi}_k = 0$, where $\boldsymbol{\xi}_k$ has amplitude $\beta e^{-n}/n$ and direction given by the mean angle from the group plus $\pi$. This single field structure does double duty: it produces the OOP grid used as goals, and it supplies the direction and magnitude costs $\mathcal{F}_{dir}$ and $\mathcal{F}_{mag}$ inside VMD-RRT*, so the same representation both selects where to stand and shapes the path to get there.

What would settle it

A direct test would be a perception experiment: for groups of two to four people at varied spacings and headings, compare the predicted optimal observation position from Eq. (13) with the positions human observers actually choose for a robot to approach, or with questionnaire ratings of approach comfort; if the predicted point consistently falls outside the preferred interaction region, or if small heading changes (around 10 degrees) move the predicted OOP by more than about a meter, the dipole superposition is not capturing group proxemics.

Watch

Extended reading notes

Core claim

The central discovery is that group proxemics can be reduced to a single continuous two-dimensional vector field whose structure tells a robot both where to stand and how to get there. Each person contributes a magnetic-dipole field $\mathcal{M}$ (Eqs. (7)-(8)) rotated by their heading, the group field is the linear superposition $\boldsymbol{V}_{g_k} = \sum_j \boldsymbol{S}_j$, and the robot's target, the OOP, is the point where $\boldsymbol{V}_{g_k}(O_k) + \boldsymbol{\xi}_k = 0$ with damping amplitude $\beta e^{-n}/n$ set by group size. The same field, summed over all groups, is fed into an RRT*-style planner whose edge cost penalizes path length, field-direction mismatch, and field magnitude, so the robot cruises between OOPs without entering social gaze spaces. The authors report that this framework, tested in simulation and with a real service robot, recognizes groups, re-plans when groups merge or move, and keeps a distance above the 0.4 m threshold while visiting all groups.

Load-bearing premise

The load-bearing premise is that a person's social space really has the shape of the two-dimensional magnetic dipole field with the fixed constant $\alpha=-615$, and that linearly adding those fields gives a meaningful group map whose zero-damping point is where a robot should stand; no calibration against human comfort or perception data is provided.

Editorial extensions

If this is right

  • A robot using this framework can treat people as interaction targets rather than obstacles, so the same proxemics field that keeps it at a respectful distance also tells it where to stop and initiate contact.
  • Because the clustering and field are recomputed online, a change in group composition (people joining, leaving, or moving) propagates directly into a new OOP set and a new cruise path without switching to a separate avoidance mode.
  • The reported comparison against RRT and RRT* on the physical platform indicates that field-guided sampling can reach the same observation goals with shorter path length and fewer explored nodes (about 24 m and 6,500 nodes for VMD-RRT* versus 26 m and 20,000 nodes for RRT*).
  • The design space has a documented trade-off: enlarging the robot motion space radius speeds planning but lowers success in dynamic scenes (from 10/10 at $r_{rob}=1.9$ m to 4/10 at $r_{rob}=2.7$ m with two moving individuals), so the radius must be tuned to the expected scene dynamics.

Reading between the lines

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

  • Going beyond the paper, the same zero-damping condition could be read as a general social-equilibrium viewpoint generator, applicable to multi-party human-robot interaction or camera viewpoint selection for group photography, not only robot cruising.
  • A testable extension the paper does not perform is calibration: fitting $\alpha$, $\beta$, and the Weibull parameters $a,b,c$ to measured human comfort ratings or perceptual grouping judgments would turn the model from an analogy into a predictive proxemics theory; the current constants are fixed without such data.
  • Because the framework assumes full pose information from a motion capture system, a practical next step is to replace that input with onboard person detection and tracking; the sensitivity of OOP positions to heading noise would then be the key question.
  • The TSP-plus-sampling decomposition suggests a natural online variation: predict group motion and re-solve the visiting sequence over a short horizon, which would reduce redundant observations like the duplicate visit to $\mathcal{G}_2'$ seen in one dynamic experiment.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 8 minor

Summary. The paper proposes an online robot motion planning framework that blends social proxemics of human groups into navigation. It introduces a graph-based clustering method combining social relevance and spatial confidence (Section II-A), a magnetic-dipole vector-field model of individual and group proxemics (Section II-B), a method to compute optimal observation positions (OOPs) via a vector balance equation (Eq. (13)), and a sampling-based planner called VMD-RRT* that uses the proxemics field in its cost function (Section II-C). The authors validate the approach with simulations in a 14 m x 14 m map with static and moving pedestrians, and with real-robot experiments using a Vicon motion capture system and the Xiaopang robot in static and dynamic social scenarios. They report high group recognition, successful OOP arrival, and improved time/path/node statistics compared with RRT and RRT*.

Significance. If the proxemics model and the OOP computation are sound, the paper offers a useful way to make robot navigation interaction-aware rather than obstacle-avoidance-only, and the inclusion of real-robot experiments with dynamic group reconfiguration is a strength. The paper also provides a reproducible simulation setup and a direct planner comparison (Table III) that gives independent evidence for the value of the hierarchical planning wrapper. However, the paper's central claim depends on the magnetic-dipole proxemics model and the OOP equation being meaningful and well-posed, and that is not currently established: the model constants are asserted rather than calibrated, the implicit OOP equation is not shown to have a unique solution, and the main success metric (Eq. (25)) is partly self-referential. These gaps make the contribution conditional rather than fully convincing.

major comments (4)
  1. [II-B, Eq. (13)-(16)] The OOP is defined as a solution of the implicit vector equation V_gk(O_k) + xi_k = 0, but the amplitude of xi_k in Eq. (14) is negative for the reported beta = -32, making the sign convention ambiguous. Moreover, Arg(xi_k) in Eq. (15) depends on O_k through Eq. (16), so Eq. (13) is a nonlinear implicit equation in O_k. The paper provides no existence or uniqueness argument, no description of the numerical solver, and no residuals for the OOPs displayed in Figs. 8 and 14. As written, the 'optimal observation position' is not actually shown to exist, be unique, or be computable for the reported configurations. Please provide a solver, a sign convention (e.g., define Amp(xi_k) with absolute value or change beta), and a numerical check of existence/uniqueness over the tested group configurations.
  2. [II-B, Table I] The constants alpha = -615, beta = -32, a = 5.102, b = 0.748, c = 0.087, r = 0.05, and the cost weights in Eq. (20) are asserted without calibration against human proxemics data or any perceptual ground truth. Since the social cost in VMD-RRT* is computed from V_gC, and the OOP is defined as a zero of a field with these constants, the reported paths and OOPs are not yet shown to reflect human social preferences. At minimum, provide a sensitivity analysis over alpha and beta, and ideally compare the resulting OOP distances and approach orientations with established proxemics ranges (e.g., Hall's proxemics or the social interaction field model of Ref. [12]).
  3. [IV-B, Eq. (25)] The Arrive Rate metric A_r measures error against the OOP and the damping orientation Arg(xi_k) that the model itself generates, so high A_r largely demonstrates self-consistency of the optimization, not external validity of the proxemics model. To support the claim that the robot reaches socially appropriate observation positions, please add an independent evaluation, for example human-ratings of the robot's approach, comparison with a random or heuristic OOP baseline, or a measure based on the group gaze center gamma_k defined in Eq. (19).
  4. [III-B, Table III] The comparison with RRT and RRT* in Table III shows that VMD-RRT* reduces path length and nodes, but the practical framework also uses a TSP-based visiting sequence, informed sampling in an ellipse, and hierarchical sub-goal replanning (Section II-C3). These factors, rather than the proxemics field itself, may drive the improvement. Please include an ablation that separates the effect of the vector-field cost from the effect of the hierarchical/TSP wrapper, so the reader can see how much the social proxemics field contributes.
minor comments (8)
  1. [Abstract and Section III] The abstract claims 'promising performance on group recognition accuracy', but no quantitative clustering accuracy metric is reported; the dynamic scenario in Fig. 14 is shown only qualitatively. Please add a clustering evaluation (e.g., precision/recall against labeled groupings).
  2. [II-A, Eq. (1)] Equation (1) is difficult to parse: the placement of the norm in the denominator and the layered product are unclear. Please rewrite with explicit parentheses and define all terms before use.
  3. [II-A, Eq. (2)-(3)] The function psi(cos theta) and the angles theta_A and theta_B are defined only in prose. Please give precise definitions of the angle arguments in the equations, and note that psi is not differentiable at 0 if that matters for the clustering.
  4. [II-C, Eq. (24)] In Eq. (24), 'ln V_gC(sigma)' appears to take the logarithm of a vector, while the text says 'ln is logarithmic function with base e'. If the intended quantity is the log-magnitude, please write it explicitly as ln ||V_gC(...)||_2.
  5. [Algorithm 1] Algorithm 1 has duplicate line numbers 17, and the 'CollisionFree(s_new, s_min)' check only tests two states rather than the edge between them. Also, the variable 'near' in Line 9 is introduced without definition; it should be s_nearest or s_near.
  6. [IV-B, Eq. (25)] Equation (25) contains a likely typo: the second term is written '(x_y - O_y)^2' instead of '(y_r - O_y)^2'. Please correct.
  7. [Table II] The column header says 'Success Rate' but the entries are fractions such as '10/10'. This is understandable, but please state explicitly that the denominator is the number of repeated trials; currently that is only implied in the text.
  8. [II-B, Eq. (17)-(19)] The gaze-point constant r = 0.05 in Table I seems very small compared with typical interpersonal distances; please clarify its units and justify the value, or state that it is a normalized quantity.

Circularity Check

2 steps flagged · score 4.0 of 10

OOP is defined by a self-consistency equation and the Arrive Rate is keyed to the same model outputs, so the proxemics evaluation is partly self-referential; no fitted parameter is renamed as a prediction and the RRT/RRT* benchmark is independent.

  1. self definitional [Section II-B2, Eqs. (13)-(16)]
    "Based on equation (13), the OOP can be obtained based on group proxemics fields. ... V_gk(O_k)+ξ_k=0 (13); Amp(ξ_k)=β e^{-n}/n (14); Arg(ξ_k)=(Σθ_j)/n+π (15); θ_j=tan^{-1}((O_ky-y_pj)/(O_kx-x_pj)) (16)."

    The OOP is presented as the output of a calculation, but Eq. (13) is not a derived optimality condition; it is a defining self-consistency equation whose unknown O_k appears both in the field term V_gk(O_k) and in the damping vector ξ_k, since ξ_k's argument is built from the bearings from O_k to the group members (Eqs. (15)-(16)). Thus the 'optimal observation position' is, by construction, a root of an equation defined around that same position. There is no independent objective function, calibration against human proxemics data, or external target against which optimality is tested. Subsequent use of the OOPs as navigation goals and as evidence of successful interaction restates the model's own construction rather than verifying an external prediction.

  2. self definitional [Section IV-B, Eq. (25)]
    "To evaluate whether the robot arrives at OOP and maintains a socially compliant orientation, Arrive Rate A_r that considers both position and heading is introduced for evaluation, A_r is defined as Eq. (25). A_r = e^{-((x_r-O_x)^2+(x_y-O_y)^2+sin^2(θ_r-Arg(ξ_k)-π))}."

    The success metric is defined against the very quantities the model produces: A_r is maximal when the robot is at the OOP coordinate and has heading Arg(ξ_k)-π, where OOP and ξ_k are outputs of Eqs. (13)-(16). Since the planner's cost is also built from the same vector field and the OOPs are the planned goals, a high A_r mainly demonstrates that the controller can reach the model-generated target. It is a self-consistency or tracking metric, not an external validation that these positions and orientations are socially appropriate. The reported A_r greater than 90% therefore does not independently confirm the proxemics model.

full rationale

The formal construction is mostly self-contained: the clustering method (Eqs. 1-6), the magnetic-dipole proxemics ansatz (Eqs. 7-12), and the VMD-RRT* cost (Eqs. 20-24) are explicit definitions and are not fitted predictions. No load-bearing self-citation chain or imported uniqueness theorem is present; self-citations [11] and [16] are peripheral. The first circularity is the OOP itself: Eq. (13) defines the OOP as a root of an equation whose damping term is defined from the candidate OOP, so the label 'optimal' is attached without any external objective or calibration. The second is the Arrive Rate: Eq. (25) scores success by closeness to that same model-generated OOP and heading, making the reported success partly self-referential. The RRT/RRT* comparison in Table III is independent evidence, but it validates the planner wrapper and not the proxemics model. The uncalibrated constants α=-615 and β=-32 and the possible non-uniqueness of roots of Eq. (13) are correctness risks rather than circularity. Overall, partial circularity in the OOP/Arrive-Rate loop warrants a score of 4, not higher, because the path-planning comparison remains independent and no fitted parameter is renamed as a prediction.

Assumptions & free parameters 6 free parameters · 6 assumptions · 2 invented entities

Everything downstream rests on the dipole analogy and the hand-set constants in Table I. The framework is internally consistent, but it is not calibrated to human proxemics, so the OOPs and path costs are products of the model's own assumptions rather than independently validated targets.

free parameters (6)
  • alpha (magnetic dipole coefficient) = -615
    Sets the amplitude of the social vector field in Eqs. (7)-(8); no calibration protocol or sensitivity analysis is given.
  • beta (social damping coefficient) = -32
    Controls OOP distance through Eq. (14); chosen by hand and not justified from human data.
  • Weibull confidence constants a, b, c = a=5.102, b=0.748, c=0.087
    Shape the spatial confidence in Eq. (5); no fitting procedure or confidence intervals are reported.
  • gaze distance r = 0.05
    Used in Eqs. (17)-(18) to define gaze points; arbitrary small constant.
  • cost weights delta_dis, delta_dir, delta_mag and w/l = 1, 5, 7, w/l = 5/4
    Define the path cost in Eqs. (20)-(23); hand-tuned and no sensitivity analysis is shown.
  • robot motion space radius r_rob = 1.9 m
    Chosen as basically satisfactory from Table II; planning success depends strongly on this value.
assumptions (6)
  • domain assumption Motion states of all people, robot, and obstacles are available exactly from a motion capture system.
    Stated at the start of Section II: position and heading are assumed known; this removes perception error from the evaluation.
  • ad hoc to paper Human proxemics is representable as a static 2D magnetic-dipole vector field with constant coefficient alpha.
    Introduced in Section II-B Eqs. (7)-(8); no human-subject calibration supports the dipole analogy.
  • ad hoc to paper Group proxemics is the linear superposition of member dipole fields.
    Eq. (12) sums member fields; linearity is assumed and not tested against measured group behavior.
  • ad hoc to paper The OOP is the point where group field plus social damping vanishes, with fixed damping magnitude and direction.
    Eqs. (13)-(16) define optimality implicitly; there is no objective function or comparison with human-preferred observation positions.
  • domain assumption RRT* with a vector-field cost yields near-optimal paths, and social gaze space can be treated as an obstacle region.
    Section II-C; standard RRT* convergence is assumed and social gaze space is excluded from free space by definition.
  • domain assumption In simulation, the robot follows the planned path perfectly and kinematic constraints are ignored.
    Assumptions 1-3 in Section III-A.
invented entities (2)
  • Magnetic-dipole social proxemics field
    purpose: Encodes each individual's social space as a vector field and, by superposition, the group proxemics field used for OOP and path cost.
    No external data or perceptual study validates the dipole shape; only the robot's ability to reach computed OOPs is shown.
  • Social damping vector xi_k
    purpose: Introduced in Eq. (13) to balance the group field and thereby define each group's OOP.
    Amplitude beta e^{-n}/n and direction average bearing plus pi are ad hoc model components without independent measurement.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Online Robot Motion Planning Methodology Guided by Group Social Proxemics Feature." pith.science (2026). https://pith.science/paper/EF5FU2YU

@misc{pith2026250204837,
  author       = {Pith},
  title        = {Pith review of: Online Robot Motion Planning Methodology Guided by Group Social Proxemics Feature},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/EF5FU2YU}},
  note         = {Machine review of arXiv:2502.04837}
}
read the original abstract

Nowadays robot is supposed to demonstrate human-like perception, reasoning and behavior pattern in social or service application. However, most of the existing motion planning methods are incompatible with above requirement. A potential reason is that the existing navigation algorithms usually intend to treat people as another kind of obstacle, and hardly take the social principle or awareness into consideration. In this paper, we attempt to model the proxemics of group and blend it into the scenario perception and navigation of robot. For this purpose, a group clustering method considering both social relevance and spatial confidence is introduced. It can enable robot to identify individuals and divide them into groups. Next, we propose defining the individual proxemics within magnetic dipole model, and further established the group proxemics and scenario map through vector-field superposition. On the basis of the group clustering and proxemics modeling, we present the method to obtain the optimal observation positions (OOPs) of group. Once the OOPs grid and scenario map are established, a heuristic path is employed to generate path that guide robot cruising among the groups for interactive purpose. A series of experiments are conducted to validate the proposed methodology on the practical robot, the results have demonstrated that our methodology has achieved promising performance on group recognition accuracy and path-generation efficiency. This concludes that the group awareness evolved as an important module to make robot socially behave in the practical scenario.

Figures

Figures reproduced from arXiv: 2502.04837 by the authors.

Figure 1
Figure 1. are portions of the previously generated quasi-optimal path. When sub-goal reaches the determination state, the robot would determine which group is more deserving to approach, green lines and blue lines indicate the paths to different groups. According to the obtained cruising path, robot will try to visit all group in the proxemics-OOPs system. When reaching an OOP, robot can observe this group, then decides wheth… view at source ↗
Figure 3
Figure 3. The process of clustering based on graph connectivity [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

37 extracted references · 37 canonical work pages

  1. [12]

    A social interaction field model accurately identifies static and dynamic social groupings,

    C. Zhou, M. Han, Q. Liang, Y.-F. Hu, and S.-G. Kuai, “A social interaction field model accurately identifies static and dynamic social groupings,” Nat. Hum. Behav., vol. 3, no. 8, pp. 847–855, Jun. 2019

    work page 2019

  2. [1]

    Data -driven Communicative Behaviour Generation: a survey,

    N. Oralbayeva, A. Aly, A. Sandygulova, and T. Belpaeme, “Data -driven Communicative Behaviour Generation: a survey,” ACM Trans. Hum. - Robot Interact., vol. 13, no. 1, pp. 1–39, Jan. 2024

    work page 2024

  3. [2]

    Stable learning establishes some common ground between causal inference and machine learning,

    P. Cui and S. Athey, “Stable learning establishes some common ground between causal inference and machine learning,” Nat. Mach. Intell., vol. 4, no. 2, pp. 110–115, Feb. 2022

    work page 2022

  4. [3]

    In situ bidirectional human -robot value alignment,

    L. Yuan, X. Gao, Z. Zheng, M. Edmonds, Y. N. Wu, F. Rossano, H. Lu, Y. Zhu, and S.-C. Zhu, “In situ bidirectional human -robot value alignment,” Sci. Robot., vol. 7, no. 68, Jul. 2022

    work page 2022

  5. [4]

    Human-behaviour-based social locomotion model improves the humanization of social robots,

    C. Zhou, M.-C. Miao, X.-R. Chen, Y.-F. Hu, Q. Chang, M.-Y. Yan, and S.- G. Kuai, “Human-behaviour-based social locomotion model improves the humanization of social robots,” Nat. Mach. Intell., vol. 4, no. 11, pp. 1040– 1052, Nov. 2022

    work page 2022

  6. [5]

    Individual differe nces in human gaze behavior generalize from faces to objects,

    M. D. Broda and B. De Haas, “Individual differe nces in human gaze behavior generalize from faces to objects,” Proc. Natl. Acad. Sci., vol. 121, no. 12, Mar. 2024

    work page 2024

  7. [6]

    A kinematic model generates non-circular human proxemics zones,

    F. Camara and C. Fox, “A kinematic model generates non-circular human proxemics zones,” Adv. Robot., vol. 37, no. 24, pp. 1566–1575, Oct. 2023

    work page 2023

  8. [7]

    Human-Aware path planning with improved virtual Doppler method in highly dynamic environments,

    K. Cai, W. Chen, C. Wang, S. Song, and M. Q. -h. Meng, “Human-Aware path planning with improved virtual Doppler method in highly dynamic environments,” IEEE Trans. Autom. Sci. Eng. , vol. 20, no. 2, pp. 1304 – 1321, Apr. 2023

    work page 2023

Show all 37 references
  1. [8]

    A Gaussian mixture model based fast motion planning method through online environmental feature learning,

    Y. Yuan, J. Liu, W. Chi, G. Chen, and L. Sun, “A Gaussian mixture model based fast motion planning method through online environmental feature learning,” IEEE Trans. Ind. Electron., vol. 70, no. 4, pp. 3955–3965, Apr. 2023

    2023

  2. [9]

    The grand challenges of Science Robotics,

    G.-Z. Yang, J. Bellingham, P. E. Dupont et al. “The grand challenges of Science Robotics,” Sci. Robot., vol. 3, no. 14, Jan. 2018

    2018

  3. [10]

    Relational visual representations underlie human social interaction recognition,

    M. Malik and L. Isik, “Relational visual representations underlie human social interaction recognition,” Nat. Commun., vol. 14, no. 1, Nov. 2023

    2023

  4. [11]

    Siamese Transformer for group Re -Identification via multiscale feature transform and joint learning,

    X. Lu, X. Li, C. Hu, F. Dunkin, H. Li, and S. S. Ge, “Siamese Transformer for group Re -Identification via multiscale feature transform and joint learning,” IEEE Trans. Instrum. Meas., vol. 73, pp. 1–10, Jan. 2024

    2024

  5. [13]

    A force -based model for adaptively controlling the spatial configuration of pedestrian subgroups at non-extreme densities,

    W. Wu, W. Yi, X. Wang, and X. Zheng, “A force -based model for adaptively controlling the spatial configuration of pedestrian subgroups at non-extreme densities,” Transp. Res. Pt. C -Emerg. Technol., vol. 152, p. 104154, Jul. 2023

    2023

  6. [14]

    Persuasive Polite Robots in Free -Standing Conversat ional Groups,

    S. Zojaji, A. B. Latupeirissa, I. Leite, R. Bresin and C. Peters, "Persuasive Polite Robots in Free -Standing Conversat ional Groups," in Proc. IEEE/RSJ Int. Conf. Intell. Robots Syst. (IROS), Detroit, USA, Dec. 2023, pp. 4006-4013

    2023

  7. [15]

    Core Challenges of social Robot Navigation: a surve y,

    C. Mavrogiannis, F. Baldini, A. Wang, D. Zhao, P. Trautman, A. Steinfeld, and J. Oh, “Core Challenges of social Robot Navigation: a surve y,” ACM Trans. Hum.-Robot Interact., vol. 12, no. 3, pp. 1–39, Apr. 2023

    2023

  8. [16]

    Saturation -Tolerant prescribed control for nonlinear systems with unknown control directions and external disturbances,

    R. Ji, S. S. Ge, and D. Li, “Saturation -Tolerant prescribed control for nonlinear systems with unknown control directions and external disturbances,” IEEE T. Cybern., vol. 54, no. 2, pp. 877–889, Feb. 2024

    2024

  9. [17]

    Human-Aware Navigation in Crowded Environments Using Adaptive Proxemic Area and Group Detection,

    C. Medina -Sá nchez, S. Janzon, M. Zella, J. Capitá n, and P. J. Marró n, “Human-Aware Navigation in Crowded Environments Using Adaptive Proxemic Area and Group Detection,” in Proc. IEEE/RSJ Int. Conf. Intell. Robots Syst. (IROS), Detroit, USA, Dec. 2023, pp: 6741-6748

    2023

  10. [18]

    Reliable, Robust, accurate and Real-Time 2D LIDAR human Tracking in cluttered Environment: a social Dynamic Filtering approach,

    H. Bozorgi, X. T. Truong, and T. D. Ngo, “Reliable, Robust, accurate and Real-Time 2D LIDAR human Tracking in cluttered Environment: a social Dynamic Filtering approach,” IEEE Robot. Autom. Lett., vol. 7, no. 4, pp. 11689–11696, Oct. 2022

    2022

  11. [19]

    Dynamic warning zone and a short -distance goal for autonomous robot navigation using deep reinforcement learning,

    E. E. Montero, H. Mutahira, N. Pico, and M. S. Muhammad, “Dynamic warning zone and a short -distance goal for autonomous robot navigation using deep reinforcement learning,” COMPLEX INTELL. SYST., vol. 10, no. 1, pp. 1149–1166, Aug. 2023

    2023

  12. [20]

    Space Invaders: pedestrian proxemic utility functions and trust zones for autonomous vehicle interactions,

    F. Camara and C. Fox, “Space Invaders: pedestrian proxemic utility functions and trust zones for autonomous vehicle interactions,” Int. J. Soc. Robot., vol. 13, no. 8, pp. 1929–1949, Dec. 2020

    1929

  13. [21]

    A New Approach for Including Social Conventions into Social Robots Navigation by Using Polygonal Triangulation and Group Asymmetric Gaussian Functions,

    R. M. De Sousa, D. Barrios -Aranibar, J. Diaz -Amado, R. E. Patiñ o - Escarcina, and R. M. P. Trindade, “A New Approach for Including Social Conventions into Social Robots Navigation by Using Polygonal Triangulation and Group Asymmetric Gaussian Functions,” Sensors, vol. 22, n...

    2022

  14. [22]

    Feature - and Distribution-Based LIDAR SLAM with generalized feature representation and heuristic nonlinear optimization,

    J. Wang, M. Xu, G. Zhao, and Z. Chen, “Feature - and Distribution-Based LIDAR SLAM with generalized feature representation and heuristic nonlinear optimization,” IEEE Trans. Instrum. Meas. , vol. 72, pp. 1 –15, Jan. 2023

    2023

  15. [23]

    Socially compliant robot navigation in crowded environment by human behavior resemblance using deep reinforcement learning,

    S. S. Samsani and M. S. Muhammad, “Socially compliant robot navigation in crowded environment by human behavior resemblance using deep reinforcement learning,” IEEE Robot. Autom. Lett., vol. 6, no. 3, pp. 5223– 5230, Jul. 2021

    2021

  16. [24]

    Guiding vector fields for following occluded paths,

    W. Yao, B. Lin, B. D. O. Anderson, and M. Cao, “Guiding vector fields for following occluded paths,” IEEE Trans. Autom. Control., vol. 67, no. 8, pp. 4091–4106, Aug. 2022

    2022

  17. [25]

    The domain of attraction of the desired path in Vector -Field-Guided path following,

    W. Yao, B. Lin, B. D. O. Anderson, and M. Cao, “The domain of attraction of the desired path in Vector -Field-Guided path following,” IEEE Trans. Autom. Control., vol. 68, no. 11, pp. 6812–6819, Nov. 2023

    2023

  18. [26]

    Efficient robot motion planning using Bidirectional-Unidirectional RRT Extend function,

    J. Wang, W. Chi, C. Li, and M. Q. -h. Meng, “Efficient robot motion planning using Bidirectional-Unidirectional RRT Extend function,” IEEE Trans. Autom. Sci. Eng., vol. 19, no. 3, pp. 1859–1868, Jul. 2022

    2022

  19. [27]

    Robust State -Estimator-Based control of uncertain Semi-Markovian jump systems subject to actuator failures and Time-Varying delay,

    J. Yu, Z. Liu, and P. Shi, “Robust State -Estimator-Based control of uncertain Semi-Markovian jump systems subject to actuator failures and Time-Varying delay,” IEEE Trans. Autom. Control ., vol. 69, no. 1, pp . 487–494, Jan. 2024

    2024

  20. [28]

    Multipartite RRTs for rapid replanning in dynamic environments,

    M. Zucker, J. Kuffner, and M. Branicky, “Multipartite RRTs for rapid replanning in dynamic environments,” in Proc. IEEE Int. Conf. Robot. Automat. (ICRA), Roma, Italy, Apr. 2007, pp. 1603–1609

    2007

  21. [29]

    Kinodynamic motion planning on Gaussian mixture fields,

    L. Palmieri, T. P. Kucner, M. Magnusson, A. J. Lilienthal, and K. O. Arras, “Kinodynamic motion planning on Gaussian mixture fields,” in Proc. IEEE Int. Conf. Robot. Automat. (ICRA), Singapore, May. 2017, pp. 6176–6181. 14 > REPLACE THIS LINE WITH YOUR MANUSCRIPT ID NUMBER (DO...

    2017

  22. [30]

    Distributionally robust sampling -based motion pl anning under uncertainty,

    T. Summers, “Distributionally robust sampling -based motion pl anning under uncertainty,” in Proc. IEEE/RSJ Int. Conf. Intell. Robots Syst. (IROS), Madrid, Spain, Oct. 2018, pp. 6518–6523

    2018

  23. [31]

    Safe Artificial Potential Field - novel local path planning algorithm maintaining safe distance from obstacles,

    R. Szczepanski, “Safe Artificial Potential Field - novel local path planning algorithm maintaining safe distance from obstacles,” IEEE Robot. Autom. Lett., vol. 8, no. 8, pp. 4823–4830, Aug. 2023

    2023

  24. [32]

    TrIPField: a 3D potential field model and its applications to local path planning of autonomous vehicles,

    Y. Ji, L. Ni, C. Zhao, C. Lei, Y. Du, and W. Wang, “TrIPField: a 3D potential field model and its applications to local path planning of autonomous vehicles,” IEEE Trans. Intell. Transp. Syst., vol. 24, no. 3, pp. 3541–3554, Mar. 2023

    2023

  25. [33]

    The ‘Social Gaze Space’: A Taxonomy for Gaze -Based Communication in Triadic Interactions,

    M. Jording, A. Hartz, G. Bente, M. Schulte-Rüther, and K. Vogeley, “The ‘Social Gaze Space’: A Taxonomy for Gaze -Based Communication in Triadic Interactions,” Front. Psychol., vol. 9, pp. 1-18, Feb. 2018

    2018

  26. [34]

    Robot gaze during autonomous navigation and its effect on social presence,

    K. He, W. P. Chan, A. Cosgun, A. Joy, and E. A. Croft, “Robot gaze during autonomous navigation and its effect on social presence,” Int. J. Soc. Robot., vol. 16, no. 5, pp. 879–897, Jun. 2023

    2023

  27. [35]

    Magne tic field diagnostics in the solar upper atmosphere,

    J. T. Bueno and T. Del Pino Alemán, “Magne tic field diagnostics in the solar upper atmosphere,” Annu. Rev. Astron. Astrophys., vol. 60, no. 1, pp. 415–453, Aug. 2022

    2022

  28. [36]

    On Hierarchical Multi -UAV Dubins Traveling Salesman problem paths in a complex obst acle environment,

    J. Fu, G. Sun, J. Liu, W. Yao, and L. Wu, “On Hierarchical Multi -UAV Dubins Traveling Salesman problem paths in a complex obst acle environment,” IEEE T. Cybern., vol. 54, no. 1, pp. 123–135, Jan. 2024

    2024

  29. [37]

    Proxemics [and comments and replies],

    E. T. Hall et al., “Proxemics [and comments and replies],” Current Anthropol., vol. 9, nos. 2–3, pp. 83–108, 1968. Xuan Mu received the B.S. in automation from the Qingdao University, Qingdao, China, in 202 5. He is currently pursuing the master’s degree in Qingdao University,...

    1968

Pith tools

Reviewed August 8, 2026 · model on record in the stance chip above.