Abstract:Trajectory inference seeks to recover the temporal dynamics of a population from snapshots of its (uncoupled) temporal marginals, i.e. where observed particles are not tracked over time. Lavenant et al. <a class="link-https" data-arxiv-id="2102.09204" href="https://arxiv.org/abs/2102.09204">arXiv:2102.09204</a> addressed this challenging problem under a stochastic differential equation (SDE) model with a gradient-driven drift in the observed space, introducing a minimum entropy estimator relative to the Wiener measure. Chizat et al. <a class="link-https" data-arxiv-id="2205.07146" href="https://arxiv.org/abs/2205.07146">arXiv:2205.07146</a> then provided a practical grid-free mean-field Langevin (MFL) algorithm using Schrödinger bridges. Motivated by the overwhelming success of observable state space models in the traditional paired trajectory inference problem (e.g. target tracking), we extend the above framework to a class of latent SDEs in the form of observable state space models. In this setting, we use partial observations to infer trajectories in the latent space under a specified dynamics model (e.g. the constant velocity/acceleration models from target tracking). We introduce PO-MFL to solve this latent trajectory inference problem and provide theoretical guarantees by extending the results of <a class="link-https" data-arxiv-id="2102.09204" href="https://arxiv.org/abs/2102.09204">arXiv:2102.09204</a> to the partially observed setting. We leverage the MFL framework of <a class="link-https" data-arxiv-id="2205.07146" href="https://arxiv.org/abs/2205.07146">arXiv:2205.07146</a>, yielding an algorithm based on entropic OT between dynamics-adjusted adjacent time marginals. Experiments validate the robustness of our method and the exponential convergence of the MFL dynamics, and demonstrate significant outperformance over the latent-free method of <a class="link-https" data-arxiv-id="2205.07146" href="https://arxiv.org/abs/2205.07146">arXiv:2205.07146</a> in key scenarios.

Trajectory Inference of Unknown Linear Systems Based on Partial States Measurements

Adaptive backstepping control of uncertain linear systems under unknown actuator delay

CLINS: Continuous-Time Trajectory Estimation for LiDAR-Inertial System

Event-Triggered Control for Safety-Critical Systems With Unknown Dynamics

Observability-Aware Trajectory Optimization: Theory, Viability, and State of the Art

Active Information Acquisition under Arbitrary Unknown Disturbances

Abstraction-based Control of Unknown Continuous-Space Models with Just Two Trajectories

Partially Observed Trajectory Inference using Optimal Transport and a Dynamics Prior

From a Single Trajectory to Safety Controller Synthesis of Discrete-Time Nonlinear Polynomial Systems

Guaranteed Trajectory Tracking under Learned Dynamics with Contraction Metrics and Disturbance Estimation

An Indirect Data-Driven Method for Trajectory Tracking Control of a Class of Nonlinear Discrete-Time Systems

Adaptive intrusion tolerant control for a class of uncertain nonlinear cyber–physical systems with full-state constraints

Online learning-based trajectory tracking for underactuated vehicles with uncertain dynamics

Adaptive physics-informed trajectory reconstruction exploiting driver behavior and car dynamics

Robust trajectory tracking for incrementally passive nonlinear systems.

Event-triggered adaptive secure tracking control for nonlinear cyber–physical systems against unknown deception attacks

Embedding the State Trajectories of Nonlinear Systems via Multimodel Linear Descriptions: A Data-Driven-Based Algorithm

Novel Trajectory Optimization Algorithm of Vehicle-borne LiDAR Mobile Measurement System

Traj-LIO: A Resilient Multi-LiDAR Multi-IMU State Estimator Through Sparse Gaussian Process

Guaranteed Reachability on Riemannian Manifolds for Unknown Nonlinear Systems

Model-free tracking control of complex dynamical trajectories with machine learning