Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper introduces MetaKoopman, a Bayesian meta-learning framework that utilizes linear latent representations to model nonlinear dynamics under distribution shifts. By employing a Matrix Normal-Inverse Wishart (MNIW) prior over the Koopman operator, MetaKoopman enables closed-form Bayesian updates and provides a posterior predictive distribution that captures both epistemic and aleatoric uncertainty. Evaluated on an autonomous truck and trailer system in challenging winter conditions, MetaKoopman outperforms existing methods in multi-step prediction accuracy and robustness to distributional shifts, demonstrating its practical utility in dynamic motion planning tasks.
MetaKoopman achieves superior prediction accuracy and uncertainty calibration in nonlinear dynamics, even under extreme environmental conditions.
Modeling and forecasting nonlinear dynamics under distribution shifts is essential for robust decision-making in real-world systems. In this work, we propose MetaKoopman, a Bayesian meta-learning framework for modeling nonlinear dynamics through linear latent representations. MetaKoopman learns a Matrix Normal-Inverse Wishart (MNIW) prior over the Koopman operator, enabling closed-form Bayesian updates conditioned on recent trajectory segments. Moreover, it provides a closed-form posterior predictive distribution over future state trajectories, capturing both epistemic and aleatoric uncertainty in the learned dynamics. We evaluate MetaKoopman on a full-scale autonomous truck and trailer system across a wide range of adverse winter scenarios, including snow, ice, and mixed-friction conditions, as well as in simulated control tasks with diverse distribution shifts. MetaKoopman consistently outperforms prior approaches in multi-step prediction accuracy, uncertainty calibration, and robustness to distributional shifts. Field experiments further demonstrate its effectiveness in dynamically feasible motion planning, particularly during evasive maneuvers and operation at the limits of traction. Project website: https://mahmoud-selim.github.io/MetaKoopman/