Spatiotemporal Forecasting as Planning: A Model-Based Reinforcement Learning Approach with Generative World Models
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , , , , , , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2025
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866909833664397312 |
|---|---|
| author | Wu, Hao Gao, Yuan Shi, Xingjian Li, Shuaipeng Xu, Fan Zhang, Fan Zhu, Zhihong Wang, Weiyan Luo, Xiao Wang, Kun Wu, Xian Huang, Xiaomeng |
| author_facet | Wu, Hao Gao, Yuan Shi, Xingjian Li, Shuaipeng Xu, Fan Zhang, Fan Zhu, Zhihong Wang, Weiyan Luo, Xiao Wang, Kun Wu, Xian Huang, Xiaomeng |
| contents | To address the dual challenges of inherent stochasticity and non-differentiable metrics in physical spatiotemporal forecasting, we propose Spatiotemporal Forecasting as Planning (SFP), a new paradigm grounded in Model-Based Reinforcement Learning. SFP constructs a novel Generative World Model to simulate diverse, high-fidelity future states, enabling an "imagination-based" environmental simulation. Within this framework, a base forecasting model acts as an agent, guided by a beam search-based planning algorithm that leverages non-differentiable domain metrics as reward signals to explore high-return future sequences. These identified high-reward candidates then serve as pseudo-labels to continuously optimize the agent's policy through iterative self-training, significantly reducing prediction error and demonstrating exceptional performance on critical domain metrics like capturing extreme events. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2510_04020 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Spatiotemporal Forecasting as Planning: A Model-Based Reinforcement Learning Approach with Generative World Models Wu, Hao Gao, Yuan Shi, Xingjian Li, Shuaipeng Xu, Fan Zhang, Fan Zhu, Zhihong Wang, Weiyan Luo, Xiao Wang, Kun Wu, Xian Huang, Xiaomeng Machine Learning Artificial Intelligence To address the dual challenges of inherent stochasticity and non-differentiable metrics in physical spatiotemporal forecasting, we propose Spatiotemporal Forecasting as Planning (SFP), a new paradigm grounded in Model-Based Reinforcement Learning. SFP constructs a novel Generative World Model to simulate diverse, high-fidelity future states, enabling an "imagination-based" environmental simulation. Within this framework, a base forecasting model acts as an agent, guided by a beam search-based planning algorithm that leverages non-differentiable domain metrics as reward signals to explore high-return future sequences. These identified high-reward candidates then serve as pseudo-labels to continuously optimize the agent's policy through iterative self-training, significantly reducing prediction error and demonstrating exceptional performance on critical domain metrics like capturing extreme events. |
| title | Spatiotemporal Forecasting as Planning: A Model-Based Reinforcement Learning Approach with Generative World Models |
| topic | Machine Learning Artificial Intelligence |
| url | https://arxiv.org/abs/2510.04020 |