Predicting where multiple agents will move next is fundamental to autonomous driving, robotics, and any system that must share space with people safely. Modern stochastic predictors read off inter-agent interactions once from the observed past and keep that view fixed throughout the generative rollout, so the reciprocal geometric constraints that emerge as the joint future sharpens are not re-examined. We address this gap with SRA (Spatial Reasoning Adapter), a modular graph denoiser that is integrated into existing diffusion and flow-matching predictors without redesigning their generative cores. At every reverse step, SRA reads the host's current future estimate, builds a sparse Euclidean interaction graph on the predicted trajectories, and returns a residual correction to the host's decoding path. Our adapter consists of three core components: a sparse neighborhood retrieved with a learned query-key score and a geometric bias, a future-conditioned relational feature recomputed on the evolving prediction, and uncertainty-weighted message passing where confident agents act as anchors for uncertain ones. We instantiate SRA on three structurally different hosts (LED, MID, and MoFlow) across four multi-agent benchmarks: SDD, NBA, Soccer, and Football. SRA improves ADE for every host on every dataset, with average gains of 5.8% in ADE and 7.0% in FDE. A cumulative ablation shows that the three components act as one mechanism: only together do they improve every host on every sports dataset, indicating that spatial reasoning over the model's own evolving future, weighted by how far that future can be trusted, is broadly useful across denoising architectures. Code is available at https://github.com/Jaewoo97/SRA.
Figures & tables
Figure 2: Overview of SRA at one reverse step τ . From the host’s current future estimate Y^τ , node features hiτ , and uncertainty σˉ (bottom left), SRA (green, Sec. 3.3 ) scores every pair with a semantic and a geometric term and keeps the top- Knbr neighbors N(i) ; (yellow, Sec. 3.4 ) encodes the per-timestep relative positions and headings of each selected pair into edge features e^ij,tτ ; (blue, Sec. 3.5 ) pools them over time with uncertainty-biased weights, uses them in edge-conditioned attention, and gates each message by σˉi−σˉj ; and (right) decodes the aggregated message into a residual δτ that is added through the gate g to obtain the refined trajectory.
NBA
Soccer
Football
SDD
Method
ADE
FDE
ADE
FDE
ADE
FDE
ADE
FDE
MID [ 15 ]
1.037
1.472
0.451
0.783
0.274
0.459
8.85
15.92
MID + SRA
0.957 (+7.7%)
1.346 (+8.6%)
0.370 (+18.0%)
0.595 (+24.0%)
0.265 (+3.3%)
0.438 (+4.6%)
8.53 (+3.6%)
16.61 ( − 4.3%)
LED [ 37 ]
0.839
1.204
0.372
0.821
0.289
0.580
8.14
13.95
LED + SRA
0.778 (+7.3%)
1.161 (+3.6%)
0.364 (+2.2%)
0.710 (+13.5%)
0.253 (+12.5%)
0.505 (+12.9%)
7.91 (+2.8%)
13.31 (+4.6%)
MoFlow [ 14 ]
0.703
0.884
0.406
0.593
0.251
0.421
5.44
8.30
Table 1: Main results. ADE / FDE (best of K=20 modes; lower is better) of three hosts with and without SRA on NBA, Soccer, Football, and SDD. Percentages are the relative improvement of SRA over its host; the only negative value (MID, FDE on SDD) is shown in black.
Figure 3
Dataset
Configuration
Enc.
Sel.
Unc.
MID
LED
MoFlow
NBA
Host baseline
1.037 / 1.472
0.839 / 1.204
0.703 / 0.884
+ encoding (all neighbors)
✓
1.005 / 1.417
0.776 / 1.176
0.698 / 0.876
+ sparse selection
✓
✓
0.963 / 1.367
0.886 / 1.238
0.705 / 0.881
+ uncertainty (full SRA)
✓
✓
✓
0.957 / 1.346
0.778 / 1.161
0.695 / 0.870
Soccer
Host baseline
0.451 / 0.783
0.372 / 0.821
0.406 / 0.593
+ encoding (all neighbors)
✓
0.438 / 0.648
0.384 / 0.748
0.386 / 0.520
Table 2: Cumulative component analysis (ADE / FDE). Starting from the unmodified host, we add one SRA component at a time: future relational encoding over all neighbors (Enc.), sparse top- Knbr neighbor selection (Sel.), and uncertainty-weighted message passing (Unc.). The last row of each block is the full SRA reported in Table 1 . Best per dataset and host in bold .
Figure 5: ADE of LED vs. LED+SRA stratified by predicted uncertainty σ on the NBA dataset (bars, left axis), and the ADE improvement of LED+SRA over LED (line, right axis). The baseline error grows with σ , and SRA’s gain appears in every bin and is largest in the highest- σ bin.
Table 6
Appendix figures & tables10 assets
Supplementary material from the paper’s appendix.
Appendix
Symbol
Meaning
N , Tp , Tf
number of agents; observed and predicted horizon lengths
X , xi,t
observed trajectories; position of agent i at past step t ( t=0 : last observation)
host’s current clean-future estimate at reverse step τ
Appendix
Table 5: Notation used in Sec. 3 of the main paper.
Model
Knbr
ADE
FDE
MID + SRA
3
0.960
1.346
5
0.957
1.346
10
0.968
1.356
LED + SRA
3
0.775
1.163
5
0.778
1.161
10
0.783
2.303
Appendix
Table 6: Ablation on the number of selected neighbors Knbr on NBA (ADE / FDE, best of 20). Knbr=5 is the default.
NBA
Soccer
Football
Method
JADE
JFDE
JADE
JFDE
JADE
JFDE
MID
1.977
3.706
0.789
1.577
0.581
1.184
MID + SRA
1.945
3.597
0.755
1.518
0.603
1.238
LED
1.609
2.851
0.601
1.121
0.550
1.071
LED + SRA
1.580
2.842
0.575
1.055
0.501
0.969
MoFlow
1.476
2.708
0.654
1.067
0.509
0.991
Appendix
Table 7: Joint metrics JADE / JFDE [ 62 ] on the sports benchmarks, which select one mode jointly for all agents in the scene. Better of each host pair in bold .
Host
Baseline
+ level- k [ 20 ]
+ SRA (ours)
MID (diffusion)
1.037 / 1.472
0.984 / 1.394
0.957 / 1.346
LED (leapfrog)
0.839 / 1.204
0.836 / 1.266
0.778 / 1.161
MoFlow (flow matching)
0.703 / 0.884
0.949 / 1.098
0.695 / 0.870
Module parameters
—
1.30 M
0.63 M
Appendix
Table 8: Adapter-level comparison on NBA (ADE / FDE). GameFormer’s level- k reasoning [ 20 ] is placed in the same slot as SRA, with the same input, injection point, and training protocol.
Category
Method
ADE ↓
FDE ↓
Standalone future-aware
GameFormer [ 20 ]
0.848
0.951
C2F [ 23 ]
0.904
1.063
Host + adapter
MID [ 15 ]
1.037
1.472
MID + SRA
0.957 (+7.7%)
1.346 (+8.6%)
LED [ 37 ]
0.839
1.204
LED + SRA
0.778 (+7.3%)
1.161 (+3.6%)
Appendix
Table 9: Standalone future-aware predictors vs. hosts with SRA on NBA (ADE / FDE). Percentages are the relative improvement of SRA over its host.
Host
NBA
Soccer
Football
Average
MID
74.7%
81.5%
71.0%
75.7%
LED
87.3%
82.6%
89.7%
86.5%
MoFlow
66.6%
74.0%
59.7%
66.8%
Appendix
Table 10: Percentage of test scenes in which SRA lowers the host’s ADE.
Method
Per-batch (ms)
Per-sample (ms)
Δ per-sample (ms)
Δ %
LED
233.2
14.58
–
–
LED + SRA
256.8
16.05
+1.47
+10.1%
MoFlow
86.0
5.37
–
–
MoFlow + SRA
222.8
13.93
+8.56
+159.4%
MID
803.7
50.23
–
–
MID + SRA
847.4
52.96
+2.73
+5.4%
Appendix
Table 11: Inference latency on NBA ( K=20 , batch 16, single RTX 3090). Per-sample time is the per-batch time divided by the batch size; Δ is the overhead added by SRA.
Figure 6: Additional qualitative comparisons on NBA, in the same format as Fig. 4 . Rows: hosts (LED, MID, MoFlow); left: host alone, right: host + SRA. Light, medium, and dark dots are the observed past, the ground-truth future, and the prediction (best of K=20 modes by ADE); red boxes mark where the predictions differ.
Figure 7: Additional qualitative comparisons on NBA. Rows (top to bottom): LED, MID, and MoFlow as hosts; left: host alone, right: host + SRA. Light, medium, and dark dots are the observed past, the ground-truth future, and the prediction (best of K=20 modes by ADE). Black boxes mark where the predictions differ and are enlarged in the insets.
Figure 8: Additional denoising examples without (left) and with (right) uncertainty, LED on NBA, in the same format as Fig. 4 . Rows above the full scene (bottom) show the boxed region from high noise (top) to the final step; blue: past, pink: ground-truth future, green: prediction shaded by predicted uncertainty. Agents in the crowded center receive high uncertainty and are predicted more accurately with uncertainty-weighted message passing.
Predicting traffic agent trajectories plays an important role in autonomous driving, traffic operations, transportation safety analysis, etc. Although many deep learning algorithms are devised to predict future agent trajectories, the trajectory prediction problem is still challenging due to the complexity of decision-making process, interactions with surrounding vehicles, and the existence of multiple possible intentions for the traveling agents even under similar scenarios. Most existing methods are limited by the requirement of graph structures (e.g., Graph Neural Network) or the requirement of manually labeled intentions. In this study, we propose a pure Transformer-based deep learning model for multi-modal trajectory prediction considering temporal dependencies and agent-agent spatial interactions. After encoding the historical trajectories, two parallel decoders are employed to generate trajectories and probabilities on separate decoder tracks. The model is evaluated on two real-world datasets, one highway dataset and the other pedestrian dataset with solid performance. One important insight is that following a ``propose-then-select'' strategy, the agent-agent spatial interactions are only considered for probability estimation instead of trajectory generation. In summary, the proposed model provides a potential direction to design more robust and effective multi-modal trajectory prediction models.
Diyi Liu, Zihan Niu, Tu Xu +2
College of Metropolitan Transportation, Beijing University of Technology, Beijing, China · Zhejiang Police College, Zhejiang, China
Trajectory prediction methods have demonstrated remarkable capabilities in capturing complex motion patterns. However, existing methods rely on global state assumptions, suffer from insufficient belief inference under partial observability, and lack cognitive behavioral constraints in prediction. These limitations severely compromise both deployment feasibility and physical plausibility in real-world settings. In this work, we propose FEP-Diff, an agent-centric trajectory prediction framework grounded in the Free Energy Principle, aimed at achieving cognitively plausible predictions under realistic constraints. Specifically, a dual-branch spatiotemporal encoder extracts ego-motion dynamics and social interaction cues from local observations. Building upon this, a goal-conditioned belief learner infers multimodal latent belief distributions optimized via a free-energy objective, with a social consistency constraint on the local neighborhood graph to promote cognitive alignment among neighboring agents. Finally, a residual diffusion trajectory generator is conditioned on the learned belief representations with token-level proxy conditioning, producing precise and diverse future predictions. Extensive experiments on five public benchmarks demonstrate that FEP-Diff consistently outperforms state-of-the-art methods under restricted observability. Code: https://anonymous.4open.science/r/FEP-Diff-8876.
Yanping Wu, Ji Zhang, Hao Chen +2
University of Glasgow Glasgow, UK · Southwest Jiaotong University Chengdu, China
Trajectory prediction has shifted toward structured formulations with explicit social modeling. However, existing methods inadequately distinguish the functional roles of social influence in trajectory planning. Observing that agents typically form motion plans by anticipating others' future behaviors before making local reactive adjustments, we identify social interactions as playing staged roles, namely planning precedes reaction. We propose INTraJ, a unified framework that decomposes social influence into two stages: a planning stage constructs reference trajectories using future social information, and a reaction stage recovers local adjustments from the residual between full-context prediction and the reference. INTraJ supports both multi-target and single-target paradigms. Extensive experiments on four standard benchmarks, including Argoverse 2, Argoverse 2-ped, ETH/UCY, and SDD, demonstrate consistent improvements, particularly in FDE and long-horizon consistency, with state-of-the-art performance achieved in several settings. INTraJ reframes trajectory prediction as a planning-driven two-stage process, validating that staged social modeling is critical for stable predictions. The code is publicly available at https://github.com/11isnotavailable/INTraJ.
Jiaheng Chen, Jiaxing Li, Tinghe Zhang +1
Software College, Northeastern University Shenyang, China