Online Replanning in Belief Space for Partially Observable Task and Motion Problems

A toolbox for trajectory optimization of dynamical systems

Synthesizing safe robot policies in joint physical-belief spaces with deep RL! - CoRL 2023

Experiments in dynamic treatment regimes using reinforcement learning.

Mapping religious and philosophical orientations as a low-dimensional belief space using survey data, dimensionality reduction, and clustering.

The robot navigates a discrete grid while interacting with an uncertain crowd characterized by distance, flow direction, and density, which are only partially observable. Social navigation behaviors—such as following crowd flow, yielding, overtaking, waiting, or searching for gaps—are explicitly modeled as actions.