<p>Virtual coupling technology with an independent train as the transportation unit can solve the problem that it is difficult to match the passenger flow and capacity, and fully exploit the advantages of virtual coupling technology. In urban rail transit, trains need frequent traction and braking, and the minimum tracking distance between trains constantly change with the train status. However, current research on tracking control technology for station approach scenarios primarily focuses on fixed tracking distances. Few in-depth studies have been conducted on adaptive tracking distance techniques, and passenger comfort has not been thoroughly considered. In this paper, a tracking distance calculation model incorporating the traction transmission system is established under various train operational conditions, and a nonlinear model predictive control (NMPC) tracking algorithm with variable-weight jerk constraints is designed. The optimization objective integrates the adaptive tracking distance and a variable penalty factor based on urban rail standard, aiming to achieve smooth distance reduction during station approach and precise docking at the platform, while enhancing passenger comfort. The results demonstrate that the proposed algorithm can smoothly reduce the tracking distance, achieve precise stopping for train units during station approach, ensure compliance with jerk limits, and provide a technical reference for the efficient cooperative control of virtual coupling trains in station scenarios.</p>