TSCA-Net: Temporal-Spatial Clique Attention for Interpretable Multimodal Pedestrian Trajectory Prediction

2026-07-11

AI Summary

This work addresses the challenges of motion multimodality and complex dynamic interactions in pedestrian trajectory prediction within dense scenes by proposing a unified framework featuring three key innovations. First, a spatiotemporal clique attention (TSCA) mechanism with learnable temporal gating enables dynamic interaction modeling between historical and target trajectories. Second, an asymmetric clique potential model (CPCP) built upon a dynamic social graph captures time-varying interpersonal relationships. Third, an entropy-driven adaptive KAN grid refinement (AKGR) strategy enhances the LSTM decoder with a Kolmogorov–Arnold Network, dynamically adjusting output resolution based on prediction uncertainty. The method achieves state-of-the-art performance on the ETH/UCY and Stanford Drone Dataset, yielding errors of 0.13/0.20 meters and 6.95/10.43 pixels, respectively, while significantly improving model generalization and interpretability.

Abstract

Accurate pedestrian trajectory prediction in crowded environments remains challenging due to the multimodal uncertainty of human motion and the variable complexity of motion dynamics across different scene contexts. Existing goal-conditioned models rely on static displacement structures that assign equal weight to all historical time steps, standard graph attention mechanisms, and fixed-capacity motion decoders that cannot adapt to local prediction complexity. To address these limitations, we propose TSCA-Net, a trajectory prediction framework built upon three complementary modules. The Temporal-Spatial Clique Attention (TSCA) module introduces learnable temporal gating into clique-based goal-history interaction, enabling time-aware modulation of historical observations relative to each candidate goal. The Cross-Pedestrian Clique Potential (CPCP) module models asymmetric pairwise agent relationships through a dynamic clique potential framework with a time-varying social graph. The Adaptive KAN Grid Refinement (AKGR) mechanism dynamically adjusts the B-spline grid resolution of a Kolmogorov-Arnold Network-augmented LSTM decoder based on per-agent goal distribution entropy, balancing model expressiveness against overfitting across varying motion complexities. Extensive experiments on the ETH/UCY and Stanford Drone Dataset benchmarks demonstrate that TSCA-Net achieves state-of-the-art performance, with average ADE/FDE of 0.13/0.20 m on ETH/UCY and 6.95/10.43 pixels on SDD. Comprehensive ablation studies confirm the complementary contributions of all three proposed modules.

Problem

• pedestrian trajectory prediction
• multimodal uncertainty
• motion dynamics
• crowded environments
• goal-conditioned prediction

Innovation

• Temporal-Spatial Clique Attention
• Cross-Pedestrian Clique Potential
• Adaptive KAN Grid Refinement
• Multimodal Trajectory Prediction
• Interpretable Pedestrian Motion Modeling