Modeling animal locomotion in complex terrain is challenging. In a previous study, our lab developed an energy landscape approach to study locomotor transitions in complex 3-D terrain (Othayoth, Thoms, Li, 2020, PNAS). However, the potential energy landscape was quasi-static and did not allow prediction of dynamics. To understand and simulate transition dynamics, we developed an energy landscape based dynamic model and tested it using a system of a self-propelled ellipsoid body traversing two beam obstacles. The model simplifies the animal moving through complex 3-D terrain as a self-propelled active particle moving on a potential energy landscape in 6-D space (3-D position and 3-D orientation), whose dynamics is described by a Langevin equation. Translational and rotational acceleration of the system results from the sum of all forces and their torques. The forces include conservative forces (weight and elastic forces) described by the potential energy landscape, plus a propulsive force, a viscous force, and a random force, which model the self-propulsion, damping, and stochasticity, respectively. Conservative forces were directly calculated from the potential energy landscape gradient. Because it is challenging to calculate torques of conservative forces directly from landscape gradient using Euler angles, we used virtual rotation to calculate them from the landscape, which was verified using the body-beam interaction system. Although our dynamic model neglects collisional dynamics and assumes simple non-conservative forces, it is useful for simulating, explaining, and predicting locomotion in complex terrain. This is not only useful for understanding dynamic transitions of animals but also useful for design, control, and motion planning of robots during locomotion in complex 3-D terrain.