Abstract: Robot dynamics links classical mechanics with complex engineering systems. However, when students progress from single rigid-body problems to high-degree-of-freedom robotic systems, they often encounter gaps in their dynamics modeling capabilities. Based on teaching practice in Introduction to Robotics at Fudan University, this study takes dynamics modeling competency development as the central objective and establishes a teaching framework featuring “multiple modeling perspectives + progressively complex systems.” The Lagrange, Newton–Euler, and Kane methods are used to develop complementary modeling perspectives, while the teaching cases progress from a two-degree-of-freedom (two-DOF) revolute-prismatic (RP) manipulator to a six-DOF collaborative manipulator, enabling a transition from complete analytical derivation to program implementation and model validation. Teaching practices concerning reduced formula expansion, coordination between manual derivation and modern computational tools, and process-oriented assessment are further summarized. The proposed approach provides a reference for teaching reform in robotics and related mechanics courses.