I’m interested in understanding how a cognitve map in hippocampus is learned and readout to support high-level brain functions, such as reward-based learning, path planning or flexible navigation. I study this important open question through theoretical modeling and computer simulations. The commonly used modeling tools include continuous attractor neural network (CANN), differential equations, graph theory, stochastic process, etc. The simulation of my models is typically performed on a high-fidelity virtual rat in the MuJoCo physics simulator to nearly reproduce behavior observed in actual rats. One highlight of my recent research is a computational model that explains how layout-conforming replay of place cells supports flexible navigation of a rat in a complex and dynamically changing maze.