用多步哈密顿高斯过程学习物理一致的连续动力学,无需隐状态。
Learning Passive Continuous-Time Dynamics with Multistep Port-Hamiltonian Gaussian Processes
- 在哈密顿面上设高斯过程先验,通过线性函数约束实现闭式推断。
- 在弹簧质量、范德波尔等系统上提升向量场恢复精度与哈密顿不确定性校准。
- 适合需物理一致性建模的连续时间系统,如机器人运动预测。
我们提出多步端口-哈密顿高斯过程(MS-PHS GP),用于从噪声且不规则采样的轨迹中学习物理一致的连续时间动力学,并获得哈密顿量的后验分布。通过对哈密顿量表面 $H$ 设定高斯过程先验,并将可变步长多步积分器约束表示为有限线性泛函,该方法可在无隐状态的情况下实现向量场与哈密顿量表面的闭式条件推断,且设计上保证能量守恒与无源性。我们给出了一个有限样本下的向量场误差界,将估计误差与可变步长离散化项分离。最后,在质量-弹簧、范德波尔及杜芬系统基准测试中,展示了更优的向量场恢复效果和校准良好的哈密顿量不确定性。
原文摘要 · Abstract (English)
We propose the multistep port-Hamiltonian Gaussian process (MS-PHS GP) to learn physically consistent continuous-time dynamics and a posterior over the Hamiltonian from noisy, irregularly-sampled trajectories. By placing a GP prior on the Hamiltonian surface $H$ and encoding variable-step multistep integrator constraints as finite linear functionals, MS-PHS GP enables closed-form conditioning of both the vector field and the Hamiltonian surface without latent states, while enforcing energy balance and passivity by design. We state a finite-sample vector-field bound that separates the estimation and variable-step discretization terms. Lastly, we demonstrate improved vector-field recovery and well-calibrated Hamiltonian uncertainty on mass-spring, Van der Pol, and Duffing benchmarks.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品;中文卡片由大模型生成,请以原文为准。