Abstract: Aiming at the multi-objective optimization problem of urban rail transit (hereinafter referred to as urban rail) trains that must meet comprehensive performance indicators including safety, punctuality, energy efficiency, ride comfort and stopping accuracy in operation, this paper proposed an improved genetic algorithm. The algorithm optimized the train’s energy efficiency, ride comfort and stopping accuracy, and meanwhile improved its convergence speed and convergence accuracy. This paper established a single-mass-point dynamic model of urban rail trains and constructed a multi-objective optimization function matching the train performance indicators, designed a novel adaptive crossover operator, and further optimized the crossover operation of the genetic algorithm. On this basis, this paper conducted applied research on multi-objective performance optimization of urban rail trains on a simulated line via simulation software. Simulation results demonstrate that compared with traditional algorithms, the proposed adaptive crossover strategy delivers faster convergence speed and higher convergence accuracy and achieves better optimization performance in the multi-objective optimization of urban rail trains. The comprehensive evaluation index reaches 0.56695, which is superior to the comparison schemes. The results verify the feasibility and applicability of the improved genetic algorithm for multi-objective optimization of urban rail train operation.