Search papers, labs, and topics across Lattice.
This paper introduces FEP-Diff, a novel agent-centric trajectory prediction framework based on the Free Energy Principle, designed to improve prediction accuracy and plausibility under partial observability. FEP-Diff employs a dual-branch encoder to capture ego-motion and social cues, a goal-conditioned belief learner optimized with a free-energy objective, and a diffusion-based trajectory generator. Experiments on five benchmarks show FEP-Diff outperforms existing methods, demonstrating improved performance in scenarios with limited information.
Achieve more realistic and accurate trajectory predictions by modeling agents as cognitive entities that infer beliefs and align with social norms, even with limited information.
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.
CMU ML