This paper considers the closed-loop navigation of a hypothetical ocean-sampling vehicle in the presence of an idealized ocean-eddy pair. This problem embodies many of the challenges of spatiotemporal ocean sampling using minimally actuated Lagrangian sensors. We extend our existing guidance strategy known as the Boundary-Touring Algorithm (BTA) to steer a self-propelled vehicle to a unique streamline in a two-vortex flow. The Gaussian Mixture Kalman Filter (GMKF) provides non-Gaussian state estimation of the vortex parameters based on linear observations of Lagrangian sensor position. Taken together, BTA and GMKF constitute a novel guidance framework for adaptive, Lagrangian data assimilation. Results from numerical experiments are presented for a drifting vehicle, a controlled vehicle that has knowledge of the flow field, and a controlled vehicle that navigates based on its own flow field estimate using the BTA/GMKF framework.