Abstract:Learning multi-agent system dynamics has been extensively studied for various real-world applications, such as molecular dynamics in biology. Most of the existing models are built to learn single system dynamics from observed historical data and predict the future trajectory. In practice, however, we might observe multiple systems that are generated across different environments, which differ in latent exogenous factors such as temperature and gravity. One simple solution is to learn multiple environment-specific models, but it fails to exploit the potential commonalities among the dynamics across environments and offers poor prediction results where per-environment data is sparse or limited. Here, we present GG-ODE (Generalized Graph Ordinary Differential Equations), a machine learning framework for learning continuous multi-agent system dynamics across environments. Our model learns system dynamics using neural ordinary differential equations (ODE) parameterized by Graph Neural Networks (GNNs) to capture the continuous interaction among agents. We achieve the model generalization by assuming the dynamics across different environments are governed by common physics laws that can be captured via learning a shared ODE function. The distinct latent exogenous factors learned for each environment are incorporated into the ODE function to account for their differences. To improve model performance, we additionally design two regularization losses to (1) enforce the orthogonality between the learned initial states and exogenous factors via mutual information minimization; and (2) reduce the temporal variance of learned exogenous factors within the same system via contrastive learning. Experiments over various physical simulations show that our model can accurately predict system dynamics, especially in the long range, and can generalize well to new systems with few observations.

What problem does this paper attempt to address?

The paper aims to address how to learn and generalize continuous dynamics of multi-agent systems in different environments. Existing models are mostly capable of learning single system dynamics, learning from historical observation data to predict future trajectories. However, in practice, there may be multiple systems generated in different environments, which have variations in potential external factors such as temperature and gravity. A simple solution is to learn a specific model for each environment, but this fails to leverage possible commonalities between different environments and to provide accurate predictions when data is sparse or limited. To this end, the paper proposes the GG-ODE (Generalized Graph Ordinary Differential Equation) framework, a machine learning framework parameterized by neural networks and graph neural networks, used to capture the continuous interactions between agents. The model assumes that the dynamics in different environments are controlled by common physical laws and captures these laws by learning a shared ODE function, while incorporating the unique latent external factors of each environment into the ODE function to reflect their differences. To improve model performance, the paper also designs two regularization loss functions, one to enforce orthogonality between initial states and external factors by minimizing mutual information, and another to reduce the temporal variation of external factors learned within the same system through contrastive learning. Experiments show that the GG-ODE model is able to accurately predict system dynamics, especially in long-term prediction tasks, and can generalize well on new systems, even with only a small amount of observational data.

Generalizing Graph ODE for Learning Complex System Dynamics across Environments

Learning Continuous System Dynamics from Irregularly-Sampled Partial Observations

PGODE: Towards High-quality System Dynamics Modeling

TANGO: Time-Reversal Latent GraphODE for Multi-Agent Dynamical Systems

Neural Dynamics on Complex Networks

Coupled Graph ODE for Learning Interacting System Dynamics

Stochastic Neural Simulator for Generalizing Dynamical Systems Across Environments

Towards Complex Dynamic Physics System Simulation with Graph Neural ODEs

Environment-Aware Dynamic Graph Learning for Out-of-Distribution Generalization

Causal Graph ODE: Continuous Treatment Effect Modeling in Multi-agent Dynamical Systems

Autoregressive GNN-ODE GRU Model for Network Dynamics

Prometheus: Out-of-distribution Fluid Dynamics Modeling with Disentangled Graph ODE

Relaxing Continuous Constraints of Equivariant Graph Neural Networks for Broad Physical Dynamics Learning

Signed Graph Neural Ordinary Differential Equation for Modeling Continuous-time Dynamics

Learning System Dynamics without Forgetting

LEADS: Learning Dynamical Systems that Generalize Across Environments

Relaxing Continuous Constraints of Equivariant Graph Neural Networks for Physical Dynamics Learning

HOPE: High-order Graph ODE for Modeling Interacting Dynamics.

Learning Governing Equations of Unobserved States in Dynamical Systems

OnsagerNet: Learning Stable and Interpretable Dynamics Using a Generalized Onsager Principle

Learning Coarse-Grained Dynamics on Graph