Pith. sign in

REVIEW 4 major objections 5 minor 1 cited by

Hierarchical Learning-Enhanced MPC for Safe Crowd Navigation with Heterogeneous Constraints

T0 review · 4 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read A hybrid planner pairing a reinforcement-learning goal selector with a model-predictive-control safety backend reports 99.4% success in crowded benchmark simulations and 84% across 20 hardware trials, trained end-to-end in a low-fidelity…

desk verdict Solid hybrid DRL-MPC crowd navigation with a genuinely useful training scheme; the safety claim overreaches because the MPC constraints trust predicted trajectories as ground truth. read the letter →

arxiv 2506.09859 v2 pith:QWMQNSCA submitted 2025-06-11 cs.RO

Huajian Liu , Yixuan Feng , Wei Dong , Kunpeng Fan , Chao Wang , Yongzhuo Gao This is my paper
classification cs.RO
keywords crowdnavigationreinforcementlearningmodelpredictivecontrolgraphneuralnetworkmotionplanningspatio-temporalsearchactionmaskingheterogeneousconstraints
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 safe crowd navigation for a nonholonomic robot is best built as a hierarchy: a graph neural network trained by reinforcement learning estimates the cost-to-go by recommending a local goal, and a spatio-temporal path search plus a model-predictive-control (MPC) backend turns that goal into an explicitly constraint-satisfying trajectory. In a benchmark corridor with walls, polygonal and circular obstacles, and crossing pedestrians, the system reports a 99.4% success rate with a 0.5% collision rate, outperforming all eight baselines, and 84% success across 20 hardware trials that the paper itself cautions are not a strictly rigorous comparison because pedestrian behavior is not reproducible. The claim that would matter if true is that this learning-optimization hybrid does not need a high-fidelity training simulator: two mechanisms, an incremental action-masking scheme that prunes candidate goals that make the backend infeasible and a privileged learning scheme that scores entire planned trajectories, make the full chain trainable end-to-end in a simple 2D environment, cutting training cost and easing sim-to-real transfer. Against the closest hybrid predecessors, the paper argues it is the first to handle structural constraints, dynamic obstacles, and the robot's kinematic limits simultaneously.

What carries the argument

The load-bearing object is the closed local-goal loop that connects three modules, plus two training mechanisms that make the loop learnable. The frontend is a graph neural network (GNN) actor-critic whose input is a heterogeneous graph with four node classes — ego-robot, pedestrian, circular obstacle, line obstacle — each carrying hand-built velocity-obstacle features, including an artificial collision-risk variable $\zeta = 1/(\xi+0.5)$ built from expected collision time; its output is a categorical distribution over $D=81$ discretized local goals. The spatio-temporal search module builds a directed acyclic graph on an augmented $(x,y,\psi,t)$ grid in which predicted futures $\hat{Y}$ of dynamic agents are projected as time-varying occupancy, expands nodes with seven pulse-width-modulation-inspired motion primitives (maximum velocity, zero velocity, or an in-place wait over a fixed step), and runs A*, the classic best-first graph search, with a cost that penalizes path length, reversing, turning, and straying off the straight line to the goal. The MPC backend solves the non-convex program of Eq. (8), using fourth-order Runge-Kutta discretized differential-drive dynamics, distance constraints $\|p_k - \hat{p}_k^i\| \geq \tilde{\rho}_i$ against predicted dynamic agents, and smooth hyperplane-separation constraints for convex polygons obtained by the strong-duality reformulation of signed distance, with dual and slack variables. The incremental action mask writes backend-infeasible candidate goals into the policy's logits as large negative numbers, which the paper argues produces a valid policy gradient, and privileged learning replaces the single-step reward with $R(\tilde{S}) = r_g + r_\theta + r_t + \sum_{k=1}^{M} \lambda_p^{k-1} R_s^k$, computed by forward-simulating the environment $M$ steps along the planned trajectory and backtracking. These two mechanisms are what let a non-convex optimizer sit inside an end-to-end reinforcement-learning loop; the paper's central engineering discovery is that feasibility feedback plus trajectory-level credit assignment cures the sparse-reward timeout problem that otherwise blocks hybrid training.

What would settle it

Inject known errors into the prediction module and watch the collision rate: run the full planner in the benchmark corridor while adding zero-mean Gaussian noise of growing variance to the pedestrians' predicted positions at each replanning step. If collisions rise steeply with prediction variance even though the executed controls satisfy every nominal constraint of Eq. (8), the safety guarantee is shown to be inherited from the predictor rather than from the MPC constraints themselves. A sharper event-level version is a pedestrian who suddenly reverses direction mid-horizon, a prediction violation invisible to the uncertainty-blind backend, in a situation where all nominal constraints were satisfied at planning time.

Watch

Extended reading notes

Core claim

The paper's central claim, in its own words, is that the proposed method "effectively addresses local planning in complex dynamic environments, achieving state-of-the-art performance" by pairing implicit learning with explicit optimization. Concretely, the planner (labeled HALO in the paper's figures) decomposes the navigation problem through dynamic programming: a GNN policy picks a discrete local goal from 81 candidates around the robot, a spatio-temporal A* search on an $(x,y,\psi,t)$ grid lifts that point into a reference trajectory using explicit predictions of pedestrian futures, and a non-convex MPC program enforces differential-drive kinematics, distance margins around pedestrians and circular obstacles, and signed-distance hyperplane constraints around polygonal obstacles. The paper reports 99.4% success and 0.5% collision on 500 unseen simulation cases, against all eight baselines, and 84% success in 20 physical trials where the strongest learning baselines reach 72% and 39%. Its closing section concedes two limitations that bound the claim: the optimization backend never models the uncertainty of dynamic agents, so collisions remain possible under high perception noise, and the frontend exhibits modal collapse, habitually favoring one side of the corridor.

Load-bearing premise

The load-bearing premise is that the predicted future trajectories of pedestrians are accurate enough that the spatio-temporal occupancy grid and the MPC's distance constraints describe what will actually happen; the paper's own concluding section states that the optimization backend does not account for the uncertainty of dynamic agents, so under high perception noise the predicted trajectories can be wrong and collisions may occur.

Editorial extensions

If this is right

  • The explicit MPC backend gives the planner a defined safety behavior in distribution-shifted corner cases, which the paper identifies as the reason it beats end-to-end learning baselines on the physical robot even where margins in simulation are small.
  • Because the inputs are agent-level graph features rather than raw sensors, the same policy trains in the simple 2D corridor and transfers to hardware, so hybrid planners no longer depend on high-fidelity simulation assets.
  • Restricting spatio-temporal search to a local goal inside the MPC horizon makes joint space-time search over open areas computationally feasible at a 20 Hz replanning rate, rather than being limited to sparse road scenarios.
  • The framework covers constraint types that the closest hybrids each miss: structural walls and polygons (outside GO-MPC's scope), dynamic pedestrians (outside AC-MPC's scope), and the robot's nonholonomic kinematics.
  • Safety and social compliance cost time: at 16.57 seconds mean navigation time the planner is about 20% slower than the fastest baseline, a trade the paper presents as the price of its gains in success and reduced intrusions into pedestrian space.

Reading between the lines

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

  • The masking mechanism should generalize well beyond this planner: any learning system whose outputs feed a feasibility-checked optimizer could learn faster by treating infeasibility feedback as supervised pruning of its action space, since the ablations show masking, not reward shaping, is what makes convergence possible.
  • The admitted blind spot points to the natural next upgrade: propagate uncertainty from the trajectory predictor into both the occupancy grid and the MPC constraints, replacing the deterministic $\hat{Y}$ with confidence regions so the safety layer can hedge against prediction error.
  • The paper's reported modal collapse, combined with its 20% speed penalty, suggests the RL optimizer rather than the MPC backend now sets the performance ceiling; an off-policy discrete-action algorithm, which the authors name as future work, could plausibly recover alternate routing modes and speed without touching the safety layer.
  • A testable prediction of the hierarchy's design: real-world margins will track perception quality directly, because every safety constraint inherits the tracker's errors; the 84% hardware success rate should degrade gracefully as detection noise grows and recover when the external perception stack is improved.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper proposes a hierarchical local planning framework for crowd navigation in environments with heterogeneous constraints. A GNN-based RL frontend outputs discrete local goal candidates; a spatio-temporal A* module (Algorithm 1) generates a reference path using explicit multi-agent trajectory predictions Ŷ; and an MPC backend (Eq. (8)) refines this path under kinematic, static-obstacle, and dynamic-agent distance constraints. Two training mechanisms are introduced: an incremental action-masking scheme that masks spatially unreachable or MPC-infeasible goals, and a privileged learning scheme (Algorithm 2) that evaluates the planned trajectory over M future steps to improve temporal consistency. The method is trained in a low-fidelity 2D simulator and evaluated against eight baselines in simulation, plus two baselines in 20 hardware trials. The paper reports 99.4% success rate and 0.5% collision rate in simulation, 84% success on hardware, and claims state-of-the-art performance with explicit constraint guarantees.

Significance. If the claims hold, the framework is a useful practical contribution: it demonstrates that a hybrid learning-optimization planner can be trained end-to-end in a low-fidelity simulator, achieves very high success rates and low collision rates in benchmark simulation, and transfers to a physical robot. The incremental action masking and privileged reward evaluation are plausible and potentially transferable techniques for other hierarchical RL-MPC systems. The open-source commitment and hardware validation are strengths. However, the unqualified 'state-of-the-art' claim and the 'explicit constraint guarantees' wording are not fully supported: the reported navigation time is about 20% slower than RL-RVO, and the safety properties depend on the accuracy of point predictions that are not modeled with uncertainty.

major comments (4)
  1. [Abstract and Section IV-B1, Table II] The abstract states that the method achieves 'state-of-the-art (SOTA) performance' without qualification, but Table II shows that the proposed method's average navigation time is 16.57 s versus 13.33 s for RL-RVO, a regression of about 20%, and it is also slower than several other baselines. Since navigation time is listed as a key evaluation metric, the unqualified SOTA claim is not supported by the reported results. The authors should either qualify the claim to specify the metrics on which the method is state of the art (e.g., safety and social compliance), or provide a defined aggregate objective and statistical evidence that justifies an overall ranking.
  2. [Section VI and Eq. (8)] The claimed 'explicit constraint guarantees' (Contributions item 1; also reiterated in Section VI) are conditional on the accuracy of the predicted pedestrian trajectories Ŷ from Section III-D. Equation (8) imposes ||p_k − p̂_i_k|| ≥ ρ̃_i at every prediction step using point predictions with no uncertainty model, and Algorithm 1 constructs its spatio-temporal occupancy grid from the same Ŷ. The paper's own conclusion (Section VI, first limitation) admits that 'the optimization-based backend does not explicitly account for the uncertainty of dynamic agents' and that 'unexpected collisions under high perception noise' can occur. The real-world collision rate of 16% (Table V) versus 0.5% in simulation is consistent with this gap. The safety guarantee should either be restated as a property conditional on sufficiently accurate predictions, or the backend should be augmented with prediction-uncertainty handling (e.g., chance constraints or enlarged robust margins) to support the strong claim made in the abstract.
  3. [Table IV] The ablation table is garbled and ambiguous in its current form. The checkmark columns do not align clearly with the three components (Action Mask, S-T Joint Search, Privileged), and there are duplicate or unlabeled rows, e.g., '✓ ✓' rows are not identifiable as to which component is removed, and 'MPC only ✓' is followed by an unlabeled row. Since the ablation study is used to support the contributions of the three components, each row must be clearly labeled with which component is present or removed and the corresponding SR/CR/NT/DN values must be attributed unambiguously. Without this, the ablation claims cannot be verified.
  4. [Section III-F, Eq. (9)] The reward term r_θ in Eq. (9) is never defined. Equation (5) defines r_c, r_g, r_t, r_ψ, r_v, and r_s, but no r_θ is introduced anywhere in the paper. Since Eq. (9) defines the privileged return used to train the network, the training objective is underspecified. The authors should define r_θ explicitly, or correct the notation if it is a typo for r_ψ.
minor comments (5)
  1. [Table II] There are formatting errors with duplicated numbers and missing separators, e.g., '0.994/0.002 0.994/0.0020.994/0.002' for the proposed method's success rate and '13.33/0.35 13.33/0.3513.33/0.35' for RL-RVO's navigation time. These should be cleaned for readability.
  2. [Section IV-C, Fig. 5] The proposed method is referred to as 'HALO' in the qualitative evaluation and Appendix without ever being introduced by that name in the methodology. Introduce the name consistently or rename the method to avoid confusion.
  3. [Eq. (7) and Algorithm 1] The cost terms g_r and g_ψ are described only as piecewise constant functions but their explicit forms are not given. Additionally, the heuristic h(n_st) includes a time-based term h_t(t) that is said to 'prioritize nodes with larger t values'; if this term is not an admissible lower bound on the remaining cost, A* cannot be guaranteed to return the 'optimal path P⋆' claimed in Algorithm 1. The authors should either specify the terms and prove admissibility or soften the optimality claim.
  4. [Section III-C] The sentence 'mask(·) can be viewed as a state-dependent function differentiable to its parameters θ' is misleading, since the mask is applied to the logits based on environment-dependent invalidity and does not introduce a learnable dependence on θ. The correct statement is that the masking operation zeroes the gradient of invalid action logits; please rephrase.
  5. [Section VI, second paragraph] The sentence 'MPC ensures collision avoidance even if the robot fails to find a feasible trajectory' is too strong given the limitation stated two paragraphs earlier that the backend does not account for prediction uncertainty. Please qualify this claim, e.g., 'under sufficiently accurate predictions.'

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity; visible self-citations to SAGE, PMM-Net, and [44] are component-level borrowings rather than load-bearing reductions, and the admitted prediction-uncertainty gap is a robustness limitation rather than a circular step.

full rationale

The central claim—99.4% success / 0.5% collision in simulation and 84% success in real-world trials—is established by benchmarking against eight external baselines in a shared simulator (Table II) and by repeated hardware trials (Table V), not by algebraically transforming a fitted constant. The derivation chain connects three distinct artifacts: a GNN/RL frontend that outputs local goals, an A*-based spatio-temporal search over predicted trajectories, and an MPC backend (Eq. 8) that enforces kinematic and obstacle constraints. None of these steps defines its output in terms of the claimed result. The paper's self-citations are component-level: Section III-B adopts SAGE [39] for scene classification, features, and reward design; Section III-D relies on the authors' PMM-Net [42] for trajectory prediction; Section III-E delegates the signed-distance proof to [43],[44]. These cited items are externally checkable publications and are not invoked as uniqueness theorems or as the sole proof of the headline numbers, so they do not make the argument circular. The genuine weakness is the assumption that predicted trajectories \hat{Y} are accurate: Eq. (8) imposes \|p_k - \hat{p}^i_k\| \ge \tilde{\rho}_i using point predictions, and Section VI concedes that 'the optimization-based backend does not explicitly account for the uncertainty of dynamic agents' and admits 'unexpected collisions under high perception noise.' That is a correctness and robustness limitation of the safety claim, not a circular derivation, because the paper openly derives the constraints from predicted states rather than assuming the collision-free conclusion. A minor presentation gap is that the method is called 'HALO' in figure captions without a formal definition; this does not affect circularity. Overall, no load-bearing reduction of the headline result to its own inputs was found.

Assumptions & free parameters 10 free parameters · 6 assumptions · 0 invented entities

The framework rests on standard optimal control decomposition, a hand-designed reward, learned trajectory prediction from the authors' prior PMM-Net, and a differentiable masking heuristic. The main non-standard assumptions are the reliability of predicted pedestrian trajectories and the sufficiency of the low-fidelity simulation for sim-to-real transfer. No new physical entities are introduced.

free parameters (10)
  • Reward weights kappa1..kappa7 = 25.0, -25.0, 0.3, 1.0, -3.0, -1.5, 50.0
    Hand-tuned coefficients in reward Eq. (5); directly shape the learned local-goal policy and are central to the planner's behavior.
  • Trajectory evaluation discount lambda_p = 0.9
    Discount factor for future simulation rewards in the privileged reward Eq. (9); chosen by hand and not justified by a sensitivity study.
  • Social safety margin mu_s = 0.2 m
    Threshold in the social reward r_s; determines how close to pedestrians is penalized.
  • Collision-risk threshold xi_c = 3.0 s
    Determines which velocity-obstacle risks contribute to the reward term r_v.
  • S-T heuristic coefficient kappa_c = 0.01
    Balances the angle cost C in the heuristic h(n_st) in Eq. (7); hand-chosen and never varied.
  • Action space dimension D = 81
    Discretization of the robot neighborhood into candidate local goals; selected from ablation results in Table III and affects the expressiveness of the policy.
  • MPC horizon N and timestep Delta t = 10, 0.25 s
    Prediction and control horizon in Eq. (8); influences feasibility, computation time, and the length of the space-time search.
  • Privileged evaluation steps M = 4
    Number of forward simulation steps used for trajectory-level reward evaluation in Algorithm 2.
  • Curriculum episode allocation = 4k, 8k, 8k, 30k
    Manual phase schedule in Section IV-A1 that gradually increases the number of agents; not derived from any principle.
  • PPO hyperparameters = lr 2.5e-4, rollout buffer 2048, entropy coef 0.001
    Standard PPO settings with no sensitivity analysis; the discount factor gamma is described as tunable but its value is not reported.
assumptions (6)
  • standard math Principle of optimality: the global optimal control problem (1) can be decomposed into local waypoint selection (Problem 1) and local MPC (Problem 2).
    Invoked in Section III-A1; standard dynamic programming decomposition, but it requires the truncated trajectory to be optimal and the cost-to-go J* to exist.
  • domain assumption Environmental dynamics are Markovian and the hand-designed reward (5) captures navigation performance including safety and social compliance.
    The MDP formulation in Section III-B assumes S_t is a sufficient state and that reward weights kappa1..kappa7 produce the desired behavior. These weights are manually tuned.
  • ad hoc to paper The predicted future trajectories \hat{Y} from PMM-Net [42] are accurate enough for S-T occupancy mapping and MPC constraints.
    Used in Algorithm 1 and Eq. (8). The paper does not provide error analysis, and its own conclusion admits that the backend ignores dynamic-agent uncertainty, so inaccurate predictions can cause collisions.
  • standard math The signed-distance dual-cone reformulation of polygon obstacle constraints in Eq. (8) is valid as cited from [43] and [44].
    Relies on strong duality and convex analysis; the paper cites Boyd & Vandenberghe and its own prior planner for the proof without reproducing it.
  • standard math The incremental action mask update yields a valid policy gradient for the masked policy pi'_theta.
    Stated in Section III-C; assumes that masking by a large negative logit is differentiable and zeroes invalid-action gradients. No formal proof is given.
  • domain assumption The low-fidelity 2D simulation environment (Section IV-A1) is representative enough that policies trained there transfer to the real world.
    The sim-to-real claim rests on this premise. Real-world experiments with external perception only partially validate it, with 20 trials and the authors' own caveat that results are not strictly rigorous.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Hierarchical Learning-Enhanced MPC for Safe Crowd Navigation with Heterogeneous Constraints." pith.science (2026). https://pith.science/paper/QWMQNSCA

@misc{pith2026250609859,
  author       = {Pith},
  title        = {Pith review of: Hierarchical Learning-Enhanced MPC for Safe Crowd Navigation with Heterogeneous Constraints},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/QWMQNSCA}},
  note         = {Machine review of arXiv:2506.09859}
}
read the original abstract

In this paper, we propose a novel hierarchical framework for robot navigation in dynamic environments with heterogeneous constraints. Our approach leverages a graph neural network trained via reinforcement learning (RL) to efficiently estimate the robot's cost-to-go, formulated as local goal recommendations. A spatio-temporal path-searching module, which accounts for kinematic constraints, is then employed to generate a reference trajectory to facilitate solving the non-convex optimization problem used for explicit constraint enforcement. More importantly, we introduce an incremental action-masking mechanism and a privileged learning strategy, enabling end-to-end training of the proposed planner. Both simulation and real-world experiments demonstrate that the proposed method effectively addresses local planning in complex dynamic environments, achieving state-of-the-art (SOTA) performance. Compared with existing learning-optimization hybrid methods, our approach eliminates the dependency on high-fidelity simulation environments, offering significant advantages in computational efficiency and training scalability. The code will be released as open-source upon acceptance of the paper.

Figures

Figures reproduced from arXiv: 2506.09859 by the authors.

Figure 1
Figure 1. (a) Transformation from a crowd scenario to a heterogeneous graph; [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. An illustration of the architecture of the proposed GNN-based policy [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Illustration of the proposed hierarchical planner, where [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Comparative experimental results of the proposed method against [PITH_FULL_IMAGE:figures/full_fig_p010_4.png]
Figure 5
Figure 5. Figure 5: Qualitative comparison of the navigation process in representative simulation scenario. The ego-robot is depicted in black hollow circle with a red [PITH_FULL_IMAGE:figures/full_fig_p011_5.png]
Figure 6
Figure 6. Figure 6: Temporal visualizations of typical experimental scenarios, including synchronized perception and planning results. Thin lines with colors corresponding [PITH_FULL_IMAGE:figures/full_fig_p013_6.png]
Figure 7
Figure 7. Figure 7: Visualized results of long-distance multi-point navigation experiments. The lettered labels in the figure indicate representative frames. [PITH_FULL_IMAGE:figures/full_fig_p016_7.png]
Figure 8
Figure 8. Figure 8: Qualitative comparison of the ablation models in representative simulation scenario. As an extension of the legend in Fig. [PITH_FULL_IMAGE:figures/full_fig_p017_8.png]

Discussion (0). Sign in to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. RAVEN: Reinforcement-Adaptive Visibility-Graph Planning for Robust Humanoid Navigation with Collision-Free MPC

    cs.RO 2026-07 conditional novelty 5.0 of 10

    Reinforcement learning that adjusts obstacle-inflation radii in a visibility-graph planner improves humanoid navigation robustness under delay and noise, beating a static MPC baseline and an end-to-end RL policy in a ...

Reference graph

Works this paper leans on

52 extracted references · 26 canonical work pages · cited by 1 Pith paper

  1. [1]

    The dynamic window approach to collision avoidance,

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

    1997

  2. [2]

    Dynablox: Real-time detection of diverse dynamic objects in complex environments,

    L. Schmid, O. Andersson, A. Sulser, P. Pfreundschuh, and R. Siegwart, “Dynablox: Real-time detection of diverse dynamic objects in complex environments,”IEEE Robotics and Automation Letters, vol. 8, no. 10, pp. 6259–6266, 2023

    2023

  3. [3]

    Lv-dot: Lidar-visual dynamic obstacle detection and tracking for autonomous robot navigation,

    Z. Xu, H. Shen, X. Han, H. Jin, K. Ye, and K. Shimada, “Lv-dot: Lidar-visual dynamic obstacle detection and tracking for autonomous robot navigation,”arXiv preprint arXiv:2502.20607, 2025

    arXiv 2025

  4. [4]

    Ram-nas: Resource- aware multiobjective neural architecture search method for robot vision tasks,

    S. Mao, M. Qin, W. Dong, H. Liu, and Y . Gao, “Ram-nas: Resource- aware multiobjective neural architecture search method for robot vision tasks,” in2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, 2024, pp. 2712–2719

    work page 2024

  5. [5]

    Social force model for pedestrian dynamics,

    D. Helbing and P. Molnar, “Social force model for pedestrian dynamics,” Physical review E, vol. 51, no. 5, p. 4282, 1995

    1995

  6. [6]

    Reciprocal n- body collision avoidance,

    J. Van Den Berg, S. J. Guy, M. Lin, and D. Manocha, “Reciprocal n- body collision avoidance,” inRobotics Research: The 14th International Symposium ISRR. Springer, 2011, pp. 3–19

    2011

  7. [7]

    Congested traffic states in empirical observations and microscopic simulations,

    M. Treiber, A. Hennecke, and D. Helbing, “Congested traffic states in empirical observations and microscopic simulations,”Physical review E, vol. 62, no. 2, p. 1805, 2000

    2000

  8. [8]

    Frozone: Freezing-free, pedestrian-friendly navigation in human crowds,

    A. J. Sathyamoorthy, U. Patel, T. Guan, and D. Manocha, “Frozone: Freezing-free, pedestrian-friendly navigation in human crowds,”IEEE Robotics and Automation Letters, vol. 5, no. 3, pp. 4352–4359, 2020. JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2021 13 (a) (b) (c) action mask & local goal plotted in blue static obstacles dynamic obstacles goal ...

    work page 2020

Show all 52 references
  1. [9]

    Model predictive contouring control for collision avoidance in unstructured dynamic environments,

    B. Brito, B. Floor, L. Ferranti, and J. Alonso-Mora, “Model predictive contouring control for collision avoidance in unstructured dynamic environments,”IEEE Robotics and Automation Letters, vol. 4, no. 4, pp. 4459–4466, 2019

    2019

  2. [10]

    Autonomous flights in dynamic environments with onboard vision,

    Y . Wang, J. Ji, Q. Wang, C. Xu, and F. Gao, “Autonomous flights in dynamic environments with onboard vision,” in2021 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, 2021, pp. 1966–1973

    2021

  3. [11]

    Dpmpc-planner: A real- time uav trajectory planning framework for complex static environments with dynamic obstacles,

    Z. Xu, D. Deng, Y . Dong, and K. Shimada, “Dpmpc-planner: A real- time uav trajectory planning framework for complex static environments with dynamic obstacles,” in2022 International Conference on Robotics and Automation (ICRA). IEEE, 2022, pp. 250–256

    2022

  4. [12]

    Fapp: Fast and adaptive percep- tion and planning for uavs in dynamic cluttered environments,

    M. Lu, X. Fan, H. Chen, and P. Lu, “Fapp: Fast and adaptive percep- tion and planning for uavs in dynamic cluttered environments,”IEEE Transactions on Robotics, 2024

    2024

  5. [13]

    Dynamic control barrier function-based model predictive control to safety-critical obstacle-avoidance of mobile robot,

    Z. Jian, Z. Yan, X. Lei, Z. Lu, B. Lan, X. Wang, and B. Liang, “Dynamic control barrier function-based model predictive control to safety-critical obstacle-avoidance of mobile robot,” in2023 IEEE International Confer- ence on Robotics and Automation (ICRA). IEEE, 2023, pp. 3679–3685

    2023

  6. [14]

    An efficient spatial-temporal trajectory planner for autonomous vehicles in unstructured environments,

    Z. Han, Y . Wu, T. Li, L. Zhang, L. Pei, L. Xu, C. Li, C. Ma, C. Xu, S. Shenet al., “An efficient spatial-temporal trajectory planner for autonomous vehicles in unstructured environments,”IEEE Transactions on Intelligent Transportation Systems, vol. 25, no. 2, pp. 1797–1814, 2023

    2023

  7. [15]

    Distributed multi-robot collision avoidance via deep reinforcement learning for navigation in complex scenarios,

    T. Fan, P. Long, W. Liu, and J. Pan, “Distributed multi-robot collision avoidance via deep reinforcement learning for navigation in complex scenarios,”The International Journal of Robotics Research, vol. 39, no. 7, pp. 856–892, 2020

    2020

  8. [16]

    Towards multi- modal perception-based navigation: A deep reinforcement learning method,

    X. Huang, H. Deng, W. Zhang, R. Song, and Y . Li, “Towards multi- modal perception-based navigation: A deep reinforcement learning method,”IEEE Robotics and Automation Letters, vol. 6, no. 3, pp. 4986– 4993, 2021

    2021

  9. [17]

    Learning perception- aware agile flight in cluttered environments,

    Y . Song, K. Shi, R. Penicka, and D. Scaramuzza, “Learning perception- aware agile flight in cluttered environments,” in2023 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2023, pp. 1989–1995

    2023

  10. [18]

    Socially aware motion planning with deep reinforcement learning,

    Y . F. Chen, M. Everett, M. Liu, and J. P. How, “Socially aware motion planning with deep reinforcement learning,” in2017 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, 2017, pp. 1343–1350

    2017

  11. [19]

    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,” in2018 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, 2018, pp. 3052–3059

    2018

  12. [20]

    Crowd-robot interaction: Crowd-aware robot navigation with attention-based deep reinforcement learning,

    C. Chen, Y . Liu, S. Kreiss, and A. Alahi, “Crowd-robot interaction: Crowd-aware robot navigation with attention-based deep reinforcement learning,” in2019 international conference on robotics and automation (ICRA). IEEE, 2019, pp. 6015–6022

    2019

  13. [21]

    Relational graph learning for crowd navigation,

    C. Chen, S. Hu, P. Nikdel, G. Mori, and M. Savva, “Relational graph learning for crowd navigation,” in2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, 2020, pp. 10 007–10 013

    2020

  14. [22]

    Decentralized structural-rnn for robot crowd navigation with deep reinforcement learning,

    S. Liu, P. Chang, W. Liang, N. Chakraborty, and K. Driggs-Campbell, “Decentralized structural-rnn for robot crowd navigation with deep reinforcement learning,” in2021 IEEE international conference on robotics and automation (ICRA). IEEE, 2021, pp. 3517–3524

    2021

  15. [23]

    Intention aware robot crowd navigation with attention-based interaction graph,

    S. Liu, P. Chang, Z. Huang, N. Chakraborty, K. Hong, W. Liang, D. L. McPherson, J. Geng, and K. Driggs-Campbell, “Intention aware robot crowd navigation with attention-based interaction graph,” in2023 IEEE international conference on robotics and automation (ICRA). IEEE, 2023,...

    2023

  16. [24]

    Drl-vo: Learning to navigate through crowded dy- namic scenes using velocity obstacles,

    Z. Xie and P. Dames, “Drl-vo: Learning to navigate through crowded dy- namic scenes using velocity obstacles,”IEEE Transactions on Robotics, vol. 39, no. 4, pp. 2700–2719, 2023

    2023

  17. [25]

    Height: Heterogeneous interaction graph transformer for robot navigation in crowded and constrained environments,

    S. Liu, H. Xia, F. C. Pouria, K. Hong, N. Chakraborty, and K. Driggs- Campbell, “Height: Heterogeneous interaction graph transformer for robot navigation in crowded and constrained environments,”arXiv preprint arXiv:2411.12150, 2024

    2024

  18. [26]

    A graph neural network to model disruption in human- aware robot navigation,

    P. Bachiller, D. Rodriguez-Criado, R. R. Jorvekar, P. Bustos, D. R. Faria, and L. J. Manso, “A graph neural network to model disruption in human- aware robot navigation,”Multimedia tools and applications, vol. 81, no. 3, pp. 3277–3295, 2022

    2022

  19. [27]

    Reinforcement learned distributed multi-robot navigation with recipro- cal velocity obstacle shaped rewards,

    R. Han, S. Chen, S. Wang, Z. Zhang, R. Gao, Q. Hao, and J. Pan, “Reinforcement learned distributed multi-robot navigation with recipro- cal velocity obstacle shaped rewards,”IEEE Robotics and Automation Letters, vol. 7, no. 3, pp. 5896–5903, 2022

    2022

  20. [28]

    Navigating robots in dynamic environment with deep reinforcement learning,

    Z. Zhou, Z. Zeng, L. Lang, W. Yao, H. Lu, Z. Zheng, and Z. Zhou, “Navigating robots in dynamic environment with deep reinforcement learning,”IEEE Transactions on Intelligent Transportation Systems, vol. 23, no. 12, pp. 25 201–25 211, 2022

    2022

  21. [29]

    Bevformer: learning bird’s-eye-view representation from lidar-camera via spatiotemporal transformers,

    Z. Li, W. Wang, H. Li, E. Xie, C. Sima, T. Lu, Q. Yu, and J. Dai, “Bevformer: learning bird’s-eye-view representation from lidar-camera via spatiotemporal transformers,”IEEE Transactions on Pattern Analysis and Machine Intelligence, 2024

    2024

  22. [30]

    Learning- based model predictive control: Toward safe learning in control,

    L. Hewing, K. P. Wabersich, M. Menner, and M. N. Zeilinger, “Learning- based model predictive control: Toward safe learning in control,”Annual Review of Control, Robotics, and Autonomous Systems, vol. 3, no. 1, pp. 269–296, 2020

    2020

  23. [31]

    A predictive safety filter for learning-based control of constrained nonlinear dynamical systems,

    K. P. Wabersich and M. N. Zeilinger, “A predictive safety filter for learning-based control of constrained nonlinear dynamical systems,” Automatica, vol. 129, p. 109597, 2021

    2021

  24. [32]

    Where to go next: Learning a subgoal recommendation policy for navigation in dynamic environments,

    B. Brito, M. Everett, J. P. How, and J. Alonso-Mora, “Where to go next: Learning a subgoal recommendation policy for navigation in dynamic environments,”IEEE Robotics and Automation Letters, vol. 6, no. 3, pp. 4616–4623, 2021

    2021

  25. [33]

    Learning terminal state of the trajectory planner: Application for collision scenarios of autonomous vehicles,

    J. Lim, K. Lee, J. Shin, and D. Kum, “Learning terminal state of the trajectory planner: Application for collision scenarios of autonomous vehicles,” in2024 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2024, pp. 7576–7582

    2024

  26. [34]

    Differen- tiable mpc for end-to-end planning and control,

    B. Amos, I. Jimenez, J. Sacks, B. Boots, and J. Z. Kolter, “Differen- tiable mpc for end-to-end planning and control,”Advances in neural information processing systems, vol. 31, 2018

    2018

  27. [35]

    Actor-critic model predictive control,

    A. Romero, Y . Song, and D. Scaramuzza, “Actor-critic model predictive control,” in2024 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2024, pp. 14 777–14 784

    2024

  28. [36]

    iplanner: Imperative path planning,

    F. Yang, C. Wang, C. Cadena, and M. Hutter, “iplanner: Imperative path planning,”Proceedings of Robotics: Science and System XIX, p. 064, 2023

    2023

  29. [37]

    Viplanner: Visual semantic imperative learning for local navigation,

    P. Roth, J. Nubert, F. Yang, M. Mittal, and M. Hutter, “Viplanner: Visual semantic imperative learning for local navigation,” in2024 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2024, pp. 5243–5249

    2024

  30. [38]

    Neupan: Direct point robot navigation with end- to-end model-based learning,

    R. Han, S. Wang, S. Wang, Z. Zhang, J. Chen, S. Lin, C. Li, C. Xu, Y . C. Eldar, Q. Haoet al., “Neupan: Direct point robot navigation with end- to-end model-based learning,”IEEE Transactions on Robotics, 2025

    2025

  31. [39]

    Sample- efficient learning-based dynamic environment navigation with transfer- ring experience from optimization-based planner,

    L. Huajian, D. Wei, M. Shouren, W. Chao, and G. Yongzhuo, “Sample- efficient learning-based dynamic environment navigation with transfer- ring experience from optimization-based planner,”IEEE Robotics and Automation Letters, 2024

    2024

  32. [40]

    Starcraft ii: A new challenge for reinforcement learning,

    O. Vinyals, T. Ewalds, S. Bartunov, P. Georgiev, A. S. Vezhnevets, M. Yeo, A. Makhzani, H. K ¨uttler, J. Agapiou, J. Schrittwieseret al., “Starcraft ii: A new challenge for reinforcement learning,”arXiv preprint arXiv:1708.04782, 2017

    2017 arXiv

  33. [41]

    Pathrl: An end-to-end path generation method for collision avoidance via deep reinforcement learning,

    W. Yu, J. Peng, Q. Qiu, H. Wang, L. Zhang, and J. Ji, “Pathrl: An end-to-end path generation method for collision avoidance via deep reinforcement learning,” in2024 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2024, pp. 9278–9284

    2024

  34. [42]

    Pmm-net: Single- stage multi-agent trajectory prediction with patching-based embedding and explicit modal modulation,

    H. Liu, W. Dong, K. Fan, C. Wang, and Y . Gao, “Pmm-net: Single- stage multi-agent trajectory prediction with patching-based embedding and explicit modal modulation,”arXiv preprint arXiv:2410.19544, 2024

    2024 arXiv

  35. [43]

    S. P. Boyd and L. Vandenberghe,Convex optimization. Cambridge university press, 2004

    2004

  36. [44]

    Optimization- based local planner for a nonholonomic autonomous mobile robot in semi-structured environments,

    H. Liu, W. Dong, Z. Zhang, C. Wang, R. Li, and Y . Gao, “Optimization- based local planner for a nonholonomic autonomous mobile robot in semi-structured environments,”Robotics and Autonomous Systems, vol. 171, p. 104565, 2024

    2024

  37. [45]

    Prox- imal policy optimization algorithms,

    J. Schulman, F. Wolski, P. Dhariwal, A. Radford, and O. Klimov, “Prox- imal policy optimization algorithms,”arXiv preprint arXiv:1707.06347, 2017

    2017 arXiv

  38. [46]

    Optimal reciprocal collision avoidance for multiple non- holonomic robots,

    J. Alonso-Mora, A. Breitenmoser, M. Rufli, P. Beardsley, and R. Sieg- wart, “Optimal reciprocal collision avoidance for multiple non- holonomic robots,” inDistributed autonomous robotic systems: The 10th international symposium. Springer, 2013, pp. 203–216

    2013

  39. [47]

    Integrated online trajec- tory planning and optimization in distinctive topologies,

    C. R ¨osmann, F. Hoffmann, and T. Bertram, “Integrated online trajec- tory planning and optimization in distinctive topologies,”Robotics and Autonomous Systems, vol. 88, pp. 142–153, 2017

    2017

  40. [48]

    Robot navigation in a crowd by integrating deep reinforcement learning and online planning,

    Z. Zhou, P. Zhu, Z. Zeng, J. Xiao, H. Lu, and Z. Zhou, “Robot navigation in a crowd by integrating deep reinforcement learning and online planning,”Applied Intelligence, vol. 52, no. 13, pp. 15 600– 15 616, 2022

    2022

  41. [49]

    Dynamic interval restrictions on action spaces in deep reinforcement learning for obstacle avoidance,

    T. Grams, “Dynamic interval restrictions on action spaces in deep reinforcement learning for obstacle avoidance,”arXiv preprint arXiv:2306.08008, 2023. JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2021 15

    2023 arXiv

  42. [50]

    Lio-sam: Tightly-coupled lidar inertial odometry via smoothing and mapping,

    T. Shan, B. Englot, D. Meyers, W. Wang, C. Ratti, and D. Rus, “Lio-sam: Tightly-coupled lidar inertial odometry via smoothing and mapping,” in2020 IEEE/RSJ international conference on intelligent robots and systems (IROS). IEEE, 2020, pp. 5135–5142

    2020

  43. [51]

    Unsupervised pre-training for person re-identification,

    D. Fu, D. Chen, J. Bao, H. Yang, L. Yuan, L. Zhang, H. Li, and D. Chen, “Unsupervised pre-training for person re-identification,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), June 2021, pp. 14 750–14 759

    2021

  44. [52]

    Soft actor-critic for discrete action settings,

    P. Christodoulou, “Soft actor-critic for discrete action settings,”arXiv preprint arXiv:1910.07207, 2019. APPENDIX ADDITIONALQUALITATIVERESULTS To validate the feasibility of the proposed method in a realistic deployment setting using only onboard sensors and without any prior...

    1910 arXiv

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.