EnvSocial-Diff: A Diffusion-Based Crowd Simulation Model with Environmental Conditioning and Individual-Group Interaction

Bingxue Zhao; Qi Zhang; Hui Huang

cs.CVMar 25, 2026

EnvSocial-Diff: A Diffusion-Based Crowd Simulation Model with Environmental Conditioning and Individual-Group Interaction

Authors: Bingxue Zhao, Qi Zhang, Hui Huang

Organizations: VCC, College of Computer Science and Software Engineering, Shenzhen University

Abstract

Modeling realistic pedestrian trajectories requires accounting for both social interactions and environmental context, yet most existing approaches largely emphasize social dynamics. We propose EnvSocial-Diff: a diffusion-based crowd simulation model informed by social physics and augmented with environmental conditioning and individual--group interaction. Our structured environmental conditioning module explicitly encodes obstacles, objects of interest, and lighting levels, providing interpretable signals that capture scene constraints and attractors. In parallel, the individual--group interaction module goes beyond individual-level modeling by capturing both fine-grained interpersonal relations and group-level conformity through a graph-based design. Experiments on multiple benchmark datasets demonstrate that EnvSocial-Diff outperforms the latest state-of-the-art methods, underscoring the importance of explicit environmental conditioning and multi-level social interaction for realistic crowd simulation. Code is here: https://github.com/zqyq/EnvSocial-Diff.

Figures & tables

Figure 1: Environmental factors are important in crowd simulation. The target pedestrian is influenced by nearby neighbors, obstacles, objects of interest (OOI), and lighting conditions. The scene image is divided into grids to calculate lighting information.

Symbol	Meaning	Symbol	Meaning
$i,j$	Pedestrian indices	$t$	Time step index
$S_{i}^{t}$	State of $i$ at $t$ , $[\vec{p}_{i}^{t},\vec{v}_{i}^{t},\vec{a}_{i}^{t}]$	$H$	Prediction horizon (number of future steps)
$\vec{p}_{i}^{t}$	Position of $i$ at time $t$	$\vec{v}_{i}^{t}$	Velocity of $i$ at time $t$
$\vec{a}_{i}^{t}$	Acceleration of $i$ at time $t$	$f_{\theta}$	Denoising network in acceleration space
$\vec{F}_{i}^{\,\text{dest}}$	Destination-driving force of $i$	$c_{i}^{t}$	Conditioning signals $[\vec{F}_{i}^{\,\text{env}}\!\oplus\!\vec{F}_{i}^{\,\text{social}}\!\oplus\!\vec{F}_{i}^{\,\text{hist}}]$
$\vec{F}_{i}^{\,\text{env}}$	Environment-induced force on $i$	$\vec{F}_{i}^{\,\text{social}}$	Social interaction force on $i$

Table 1: Summary of main notations used in EnvSocial-Diff.

Figure 2: EnvSocial-Diff pipeline. Pedestrian motion is modeled as in the Social Force Model (SFM), where the destination force $\vec{F}_{i}^{\,\text{dest}}$ is applied outside the diffusion process to preserve long-term intent. The conditioning signals $c_{i}^{t}=[\vec{F}_{i}^{\,\text{env}}\oplus\vec{F}_{i}^{\,\text{social}}\oplus\vec{F}_{i}^{\,\text{hist}}]$ aggregate three interactive components: (1) Environmental Conditioning — obstacle and OOI features are encoded via cross-attention with pedestrians, while lighting features are extracted from grid-based scene brightness; (2) Individual–Group Interactions — GNNs encode individual-level (sim ${}^{1}_{ij}$ , sim ${}^{2}_{ij}$ ), group-level (sim ${}^{3}_{i}$ ), and relative motion ( $r_{ij}$ ) to produce the social force $\vec{F}_{i}^{\,\text{social}}$ ; and (3) Historical Trajectories — short-term motion trends are encoded from recent states using an LSTM. Given $c_{i}^{t}$ and Gaussian noise $\bm{\epsilon}\!\sim\!\mathcal{N}(0,1)$ , the denoiser $f_{\theta}$ performs reverse diffusion to recover clean accelerations $\hat{\mathbf{y}}^{\,t}_{i,0}$ , which are then combined with the destination force to yield the final prediction $\hat{\vec{a}}^{\,t}_{i}$ .

Group	Models	GC						UCY
Group	Models	MAE $\downarrow$	OT $\downarrow$	FDE $\downarrow$	MMD $\downarrow$	DTW $\downarrow$	Col $\downarrow$	MAE $\downarrow$	OT $\downarrow$	FDE $\downarrow$	MMD $\downarrow$	DTW $\downarrow$	Col $\downarrow$
Physics-based	CA	2.7080	5.4990	-	0.0620	-	1492	8.3360	79.4200	-	2.0220	-	4504
Physics-based	SFM	1.2590	2.1140	-	0.0150	-	622	2.5390	6.5710	-	0.1290	-	434
Data-driven	STGCNN	8.1608	15.8372	-	0.5296	5.1438	2076	7.5121	18.7721	-	0.5149	5.1695	1348
	PECNet	2.0669	4.3054	-	0.0397	0.7431	1142	3.9674	16.1412	-	0.1504	2.0986	1348
	MID	8.4257	35.1797	-	0.3737	4.2773	1620	8.2915	47.8711	-	0.4384	4.7109	1076

Table 2: Quantitative comparison on GC and UCY datasets. Results for baselines are directly reported from SPDiff ( Chen et al., 2024 ) , except for E-

V^{2}

-SC and Ours , which are reproduced under the same experimental settings. Here, Ours corresponds to our proposed EnvSocial-Diff model.

Figure 3: Comparison with baselines on UCY. (A) Predicted trajectories: our method (cyan) follows the ground truth (blue) more closely than SFM (magenta) and SPDiff (orange). (B) Error curves over time: our method consistently achieves lower MAE and OT, especially at longer horizons.

Model	Environment			GC						UCY
Model	Obs	OOI	Light	MAE $\downarrow$	OT $\downarrow$	FDE $\downarrow$	MMD $\downarrow$	DTW $\downarrow$	Col $\downarrow$	MAE $\downarrow$	OT $\downarrow$	FDE $\downarrow$	MMD $\downarrow$	DTW $\downarrow$	Col $\downarrow$
SPDiff	✓	✗	✗	0.9718	1.5450	0.9538	0.0100	0.3418	942	1.8853	4.2221	1.9000	0.0699	0.7496	634
SPDiff	✓	✗	✓	0.9359	1.4345	0.9404	0.0099	0.3395	958	1.8395	3.8602	1.9463	0.0669	0.7357	622
Ours	✗	✗	✗	0.9127	1.3909	0.9246	0.0087	0.3261	946	1.8597	3.8945	1.9948	0.0626	0.7353	738
	✓	✗	✗	0.8990	1.3727	0.9162	0.0087	0.3253	1000	1.8337	3.8550	1.9987	0.0604	0.7259	730
	✓	✓	✗	0.8873	1.3455	0.9037	0.0087	0.3279	910	1.8271	3.8541	1.9671	0.0586	0.7212	648

Table 3: Ablation study on structured environmental factors. The first two rows report results of the baseline SPDiff ( Chen et al., 2024 ) and its variant extended with an explicit lighting module (SPDiff+Lighting). The lower block corresponds to our proposed EnvSocial-Diff (Ours), where obstacles, objects of interest (OOI), and lighting are progressively added. Results are reported on GC and UCY datasets across six metrics, with checkmarks indicating the included factors.

Model Variant				GC						UCY
$r_{ij}$	$\text{sim}^{1}_{ij}$	$\text{sim}^{2}_{ij}$	$\text{sim}^{3}_{i}$	MAE $\downarrow$	OT $\downarrow$	FDE $\downarrow$	MMD $\downarrow$	DTW $\downarrow$	Col $\downarrow$	MAE $\downarrow$	OT $\downarrow$	FDE $\downarrow$	MMD $\downarrow$	DTW $\downarrow$	Col $\downarrow$
✓	✗	✗	✗	0.9066	1.3897	0.9192	0.0093	0.3272	990	1.9055	4.0101	2.0525	0.0628	0.7730	580
✓	✓	✗	✗	0.8946	1.3570	0.9194	0.0086	0.3303	1000	1.8846	3.8502	2.0139	0.0588	0.7720	752
✓	✓	✓	✗	0.9208	1.4137	0.9242	0.0087	0.3404	1024	1.8725	3.7937	2.0262	0.0575	0.7761	614
✓	✓	✓	✓	0.8861	1.3339	0.8997	0.0087	0.3269	906	1.8182	3.7292	1.8656	0.0598	0.7249	522

Table 4: Ablation on the IGI module. Starting from relative motion

r_{ij}

, we incrementally add individual-level similarities (

\text{sim}^{1}_{ij}

: approach tendency;

\text{sim}^{2}_{ij}

: motion alignment) and the group-level similarity (

\text{sim}^{3}_{i}

: group conformity). Results on GC and UCY show consistent gains on most metrics, with the full configuration yielding the strongest overall performance.

Appendix figures & tables16 assets

Supplementary material from the paper’s appendix.

Appendix

Figure 4: Comparison between the original GC subregion and the full GC scene. The left image highlights the cropped subarea (blue box) used in prior work, which limits spatial and interaction diversity. The right image shows the complete GC scene, covering a broader area with higher pedestrian density and environmental complexity, used in our extended evaluation.

Dataset: Full GC
Method	Social	Env	MAE $\downarrow$	OT $\downarrow$	FDE $\downarrow$	MMD $\downarrow$	DTW $\downarrow$	COL $\downarrow$
SPDiff			2.4624	4.6824	2.6707	0.0044	0.8873	1910
Ours	✓	✗	2.4478	4.4662	2.6683	0.0044	0.8836	1750
Ours	✓	✓	2.3527	4.1334	2.5891	0.0039	0.8473	1710

Appendix

Table 5: Comparison on the full GC dataset. SPDiff uses its own social interaction module and a basic environment treatment limited to obstacle repulsion. In contrast, our method replaces the social interaction module with the proposed IGI design (✓) and additionally integrates richer environmental cues (✓), including obstacles, objects of interest, and lighting. ✗ indicates the corresponding module is not used. In the table, Social and Env are marked with ‘*‘ for SPDiff’s built-in designs (social interaction and obstacle-only environment), ✓ when replaced by our modules, and ✗ when omitted.

Figure 5: Qualitative comparison between the GT, Ours, and SPDiff. (A) Near obstacles, SPDiff trajectories maybe to pass much closer to obstacles, whereas both GT and our method keep a more reasonable distance. (B) In crowded regions, SPDiff produces several near-collision interactions, while our predictions remain smoother and more socially consistent. (C)

Figure 6: Additional qualitative results for multiple target pedestrians. Each subplot visualizes one scene with a selected target pedestrian (green triangle). Ground-truth future trajectory (GT) is shown in blue, while predictions from SFM (magenta), SPDiff (orange), and our method (cyan) are overlaid. Our approach consistently aligns more closely with the GT across diverse scenarios.

Env		GC						UCY
Obs	OOI	MAE $\downarrow$	OT $\downarrow$	FDE $\downarrow$	MMD $\downarrow$	DTW $\downarrow$	Col $\downarrow$	MAE $\downarrow$	OT $\downarrow$	FDE $\downarrow$	MMD $\downarrow$	DTW $\downarrow$	Col $\downarrow$
✗	✗	0.9320	1.4425	0.9391	0.0094	0.3362	926	1.9320	3.9726	2.0367	0.0609	0.7706	550
✗	✓	0.9038	1.3645	0.9164	0.0090	0.3333	898	1.9119	3.7120	2.2160	0.0606	0.7563	688
✓	✗	0.8960	1.3562	0.9160	0.0086	0.3277	962	1.8346	4.0420	1.9951	0.0634	0.7042	710
✓	✓	0.8861	1.3339	0.8997	0.0087	0.3269	906	1.8182	3.7292	1.8626	0.0598	0.7249	522

Appendix

Table 6: Ablation on enhanced environmental features. We compare four settings: no enhancement (✗, ✗), OOI-only enhancement (✗, ✓), obstacle-only enhancement (✓, ✗), and joint enhancement with both (✓, ✓). Results on GC and UCY datasets show that the joint enhancement strategy achieves the most consistent improvements across metrics.

Dataset: UCY Students03
obstacle	lighting	OOI	MAE $\downarrow$	MMD $\downarrow$	OT $\downarrow$	COL $\downarrow$
✗	✗	✗	2.8943	7.9564	0.0954	308
✓	✗	✗	2.7822	7.3759	0.0899	266
✓	✓	✗	2.7324	7.2187	0.0803	216
✓	✓	✓	2.6742	6.7032	0.0708	216

Appendix

Table 7: Ablation study of adding environmental factors to the SFM method on UCY Students03 . ✓ / ✗ indicate whether the corresponding environmental module is enabled. Progressive addition of obstacle, lighting, and object-of-interest (OOI) cues leads to consistent improvements across all evaluation metrics.

Dataset: UCY Zara1
Obstacle	Lighting	OOI	MAE $\downarrow$	MMD $\downarrow$	OT $\downarrow$
✗	✗	✗	2.5954	9.2648	1.6676
✓	✗	✗	1.9585	5.2413	1.1437
✓	✓	✗	1.8981	5.1045	1.3224
✓	✓	✓	1.8282	4.7196	1.0968

Appendix

Table 8: Ablation study of adding environmental factors to the SFM method on UCY Zara1 . ✓ / ✗ indicate whether the corresponding environmental module is enabled. The results show that incorporating obstacle, lighting, and object-of-interest (OOI) cues progressively improve trajectory prediction performance in terms of MAE, MMD, and OT.

Figure 7: Qualitative comparisons on the GC dataset. Each subplot shows the predicted trajectory of a target pedestrian within the marked evaluation area (blue rectangle). Ground-truth (GT) future trajectories are depicted in black, while predictions from SPDiff and our method are shown in orange and cyan, respectively. Our approach produces more accurate and socially plausible predictions, particularly in scenarios involving sharp turns, long-range navigation, or subtle environmental conditioning.

	GC Dataset						UCY Dataset
$Top\_k$	MAE $\downarrow$	OT $\downarrow$	FDE $\downarrow$	MMD $\downarrow$	DTW $\downarrow$	Col $\downarrow$	MAE $\downarrow$	OT $\downarrow$	FDE $\downarrow$	MMD $\downarrow$	DTW $\downarrow$	Col $\downarrow$
2	0.8918	1.3409	0.9088	0.0086	0.3273	902	1.8149	3.6431	1.9209	0.0596	0.7251	588
4	0.8896	1.3321	0.8998	0.0087	0.3284	914	1.8142	3.7501	1.9542	0.0601	0.7007	552
6 (Ours)	0.8861	1.3339	0.8997	0.0087	0.3269	906	1.8182	3.7292	1.8656	0.0598	0.7249	522
8	0.8921	1.3599	0.9128	0.0089	0.3384	953	1.8154	3.7276	1.8999	0.0595	0.7159	476

Appendix

Table 9: Sensitivity analysis of the hyperparameter Top k on GC and UCY datasets. Top k controls how many nearest neighbors are used when computing individual-level similarities within the IGI module. All metrics follow the lower-is-better convention (

\downarrow

). We adopt

k=6

as it provides the most balanced overall performance.

	GC Dataset						UCY Dataset
Step	MAE $\downarrow$	OT $\downarrow$	FDE $\downarrow$	MMD $\downarrow$	DTW $\downarrow$	Col $\downarrow$	MAE $\downarrow$	OT $\downarrow$	FDE $\downarrow$	MMD $\downarrow$	DTW $\downarrow$	Col $\downarrow$
50	0.8870	1.3361	0.9019	0.0087	0.3269	984	1.8173	3.7486	1.8596	0.0616	0.7148	486
60	0.8860	1.3355	0.9006	0.0087	0.3273	912	1.8142	3.7358	1.8688	0.0603	0.7124	432
70(Ours)	0.8861	1.3339	0.8997	0.0087	0.3269	906	1.8182	3.7292	1.8656	0.0598	0.7249	522
80	0.8862	1.3449	0.8999	0.0088	0.3285	908	1.8214	3.8168	1.8705	0.0612	0.7118	582

Appendix

Table 10: Sensitivity analysis of the diffusion step count on GC and UCY datasets. The diffusion step controls the number of denoising iterations during sampling. All metrics follow the lower-is-better convention (

\downarrow

). We adopt 70 steps as the default, as it provides a stable and well-balanced performance across all evaluation metrics.

GC Dataset							UCY Dataset
Grid_Size	MAE $\downarrow$	OT $\downarrow$	FDE $\downarrow$	MMD $\downarrow$	DTW $\downarrow$	Col $\downarrow$	Grid_Size	MAE $\downarrow$	OT $\downarrow$ $\downarrow$	FDE $\downarrow$	MMD $\downarrow$	DTW $\downarrow$	Col $\downarrow$
220(Ours)	0.8861	1.3339	0.8997	0.0087	0.3269	906	90	1.8277	3.8858	1.8794	0.0608	0.7188	546
250	0.8842	1.3308	0.8999	0.0087	0.3275	914	100	1.8470	3.9977	1.9471	0.0629	0.7254	596
300	0.8848	1.3393	0.8986	0.0086	0.3276	916	110(Ours)	1.8182	3.7292	1.8656	0.0598	0.7249	522
400	0.8840	1.3396	0.9000	0.0086	0.3276	918	120	1.8314	3.8239	1.9613	0.0600	0.7196	561

Appendix

Table 11: Sensitivity analysis of the grid size used to construct the global scene representation on GC and UCY datasets. The grid size controls the spatial resolution of the environmental encoding, with lower-is-better metrics (

\downarrow

). We adopt a grid size of 220 for GC and 110 for UCY, as these settings yield the most balanced performance across the evaluation metrics for each dataset.

Dataset	MAE $\downarrow$	OT $\downarrow$	FDE $\downarrow$	MMD $\downarrow$	DTW $\downarrow$	COL $\downarrow$
GC (repulsive)	0.8970	1.3827	0.8974	0.0087	0.3370	894.0
GC (ours)	0.8861	1.3339	0.8997	0.0087	0.3269	906
UCY (repulsive)	1.8609	3.8766	1.9302	0.0646	0.7066	510.0
UCY (ours)	1.8182	3.7292	1.8656	0.0598	0.7249	522

Appendix

Table 12: Comparison between our method and the repulsive-force variant.

Dataset	MAE $\downarrow$	OT $\downarrow$	FDE $\downarrow$	MMD $\downarrow$	DTW $\downarrow$	COL $\downarrow$
GC (variant)	0.8862	1.3449	0.8999	0.0088	0.3284	908.0
GC (ours)	0.8861	1.3339	0.8997	0.0087	0.3269	906
UCY (variant)	2.0764	4.4494	2.1363	0.0730	0.7854	642.0
UCY (ours)	1.8182	3.7292	1.8656	0.0598	0.7249	522

Appendix

Table 13: Comparison between our full model and a simplified variant in which the environmental conditioning is removed from the diffusion inputs and applied only as forces outside the denoiser.

Dataset	MAE $\downarrow$	OT $\downarrow$	FDE $\downarrow$	MMD $\downarrow$	DTW $\downarrow$	COL $\downarrow$
GC (auto)	0.8919	1.3440	0.9059	0.0088	0.3365	1034.0
GC (ours)	0.8861	1.3339	0.8997	0.0087	0.3269	906
UCY (auto)	1.9083	3.9220	2.0423	0.0614	0.7466	644.0
UCY (ours)	1.8182	3.7292	1.8656	0.0598	0.7249	522

Appendix

Table 14: Performance comparison when replacing manually curated annotations with raw GroundSAM detections, which introduce missing and inaccurate boxes.

Method	MAE $\downarrow$	OT $\downarrow$	FDE $\downarrow$	MMD $\downarrow$	DTW $\downarrow$	COL $\downarrow$
SPDiff	0.4692	0.3153	1.7100	0.0886	0.2302	0.0
Ours	0.4083	0.2454	0.4639	0.0660	0.2162	0.0

Appendix

Table 15: Evaluation on the ETH dataset to assess cross-dataset generalization.

Dataset	MAE	OT	FDE	MMD	DTW	COL
GC (density)	0.9132	1.4298	0.9447	0.0092	0.3407	1140
GC (ours)	0.8861	1.3339	0.8997	0.0087	0.3269	906
UCY (density)	1.8542	3.9204	1.9490	0.0606	0.7223	690
UCY (ours)	1.8182	3.7292	1.8656	0.0598	0.7249	522

Appendix

Table 16: Replacing lighting conditioning with a global crowd-density representation constructed from a

K\times K

density grid.

Pedestrian Behavior Modeling

Pedestrian behavior modeling predicts and simulates how people move, interact, choose routes, and respond to vehicles, robots, hazards, and surroundings.

See all 40 papers

Aug 11, 2026

Social Graph Mamba: Forecasting Pedestrian Movements Based on Social Context

Forecasting pedestrian motion has always been fundamental for autonomous navigation in crowded environments. While attention-based methods achieve strong performance, they suffer from quadratic computational complexity in modeling social interactions, limiting scalability. Additionally, the existing methods often achieve high accuracy on prediction benchmarks at the individual level, but fail to fully capture the natural movement behaviors of crowds in real-world scenarios, particularly group structures. In this study, we propose Social Graph Mamba (SGM), a novel architecture that replaces attention-based social reasoning with Selective State Space Models (SSMs) operating on dynamically constructed interaction graphs. SGM introduces a dynamic interaction graph with social triplet factorization to decompose crowd interactions sequentially, and a community-aware module to effectively discover group structures via differentiable MinCut optimization and conditions both the embedding space and multi-modal decoder on group membership. Our experiments on standard benchmarks (ETH/UCY, SDD) demonstrate competitive performance with linear sequence complexity compared to quadratic attention-based methods. We further validate SGM in physical robot experiments by integrating predicted trajectories into a Social Force Model (SFM) for real-world implementation.

Hong-Son Nguyen, Yen-Chen Liu

Department of Mechanical Engineering, National Cheng Kung University, Tainan 70101, Taiwan

Pedestrian Trajectory Prediction Pedestrian Behavior Modeling Multimodal Trajectory Prediction Dynamic Graph Learning

Ask PDF

Oct 7, 2026

Controllable Crowd Generation through World-Model Planning

Crowd simulation plays a central role in robot navigation, autonomous driving, and urban planning. For these applications, realistic simulation requires crowds to adapt their behavior to environmental changes and user objectives. However, existing methods that rely on predefined control settings have limited flexibility in accommodating new user-specified objectives. To address this limitation, we propose Ctrl-CWM, a multi-agent Controllable Crowd World Model that integrates crowd generation and run-time control. Our key idea is to adapt the world-model principle of planning using imagined futures to crowd simulation. To this end, Ctrl-CWM consists of an encoder that learns a representation of human motion dynamics, an actor that proposes pedestrian displacements, a critic that evaluates imagined crowd trajectories, and a planner that selects actions. We first learn human motion dynamics through trajectory prediction on real-world pedestrian videos and then freeze the encoder to preserve them. Using this representation, the actor generates imagined crowd trajectories through repeated state updates, and the planner combines the critic's scores with user costs to select actions. Repeated planning advances the simulated crowd, while additional user costs introduce new control objectives without retraining. We extensively evaluate crowd generation under varied agent arrival conditions and run-time control across avoidance and attraction scenarios. Ctrl-CWM outperforms the state-of-the-art method on most crowd realism and collision metrics, and adapts crowd behaviors to user-specified objectives introduced during simulation. The project page is available at https://jungyu0413.github.io/Ctrl-CWM

JunGyu Lee, Jisu Shin, Seunghyun Shin +1

Yonsei University · GIST

Human Behavior Simulation Pedestrian Behavior Modeling World Model-Based Planning

Ask PDF

Jun 24, 2026

GROVE: Grounded Pedestrian Simulation via Natural Language for Interactive Social Robot Navigation

Pedestrian simulation is a critical component for training and deploying social robot navigation approaches, yet it remains a largely rigid system that repeatedly requires manual data generation to define even simple scenarios. We propose GROVE, a text-to-scenario pedestrian simulation framework that combines state-of-the-art approaches to produce realistic, socially challenging scenarios for social robot navigation. Our framework allows users to customize one of several common presets (emergency, queuing, normal) or even enter a fully independent prompt to generate a highly customizable pedestrian simulation. Multiple modules separately ensure the realism and soundness of long-horizon human behavior, medium-horizon pedestrian navigation, and short-horizon robot/social interactions. Each module is tuned by the prompt in a way that reflects the user intent across all aspects of pedestrian simulation. By dynamically selecting one of several state-of-the-art (SotA) approaches in our modules based on the scenario, we capture many situational nuances of pedestrian behavior in order to narrow the simulation-to-real (sim2real) gap. The human simulation is directly integrated into Isaac Sim, Gazebo, and RViz simulators for robot deployment in highly social environments. We validate our approach through qualitative comparison against existing pedestrian simulation baselines across scenarios of varying complexity in residential, hospital, and office environments. The result is a high-fidelity pedestrian simulation that challenges social robot navigation with complex, diverse, realistic human behaviors.

Duc Tai Nguyen, Volodymyr Shcherbyna, Anh Do Duc +3

1Singapore Management University · 2Technical University Berlin · 3National University of Singapore

Pedestrian Behavior Modeling Robotics Simulation Human Behavior Simulation Social Robot Navigation

Ask PDF

Abstract

Pedestrian Behavior Modeling

Explore similar work

Social Graph Mamba: Forecasting Pedestrian Movements Based on Social Context

Controllable Crowd Generation through World-Model Planning

GROVE: Grounded Pedestrian Simulation via Natural Language for Interactive Social Robot Navigation