Abstract: Aiming at the modeling of movement states and path optimization for the folk activity “Bench Dragon”, this study proposes a recursive modeling method based on planar geometric analysis and kinematic constraints: First, establish recursive models for handle positions and velocities, then develop a U-turn path length optimization model incorporating collision constraints. After solving motion parameters of handles at each time step through the recursive model, a collision time estimation model is constructed via geometric relationship analysis and coordinate system transformation. The constraints of the model are determined based on path non-intersection, collision avoidance, and non-reverse movement of the dragon head. A multi-search algorithm is designed to achieve precise determination of collision timings. Under the given helix pitch and U-turn space constraints, the approximate shortest U-turn path for the “bench dragon” calculated using a genetic algorithm is 11.7915 meters. Computed results demonstrate that when the optimized path requires the speed of each handle to not exceed 2 m/s, the dragon head speed must be constrained to no more than 0.9670 m/s. This outcome provides a quantitative basis for safety control measures.