To solve the problems in planning and design of automobile mixed-model assembly line, this paper puts forward the improved genetic algorithm-based equilibrium optimization algorithm for the automobile mixed assembly line and establishes corresponding theoretical model. The convergence and feasibility of the model are analysed, and the optimization model presented in this paper is verified by the assembling situation of the actual assembly line of an automobile door. The research conclusions are as follows. The optimized scheduling mathematical model under multiple constrains of the automobile assembly line was established and improvements were made to the traditional genetic algorithm. Self-adaptive genetic operator was added to the original model. The performance verification indicated that the time consumption of CPU in the proposed improved algorithm is much less, and its maximum load is larger, so it has better convergence compared with traditional genetic algorithm. The improved optimal algorithm of automobile mixed ASSEMBLY LINE was verified taking into consideration such constraint conditions as the proportion of a single product put into assembly line, staffing, and balance of the left door and right door. It is found that the overall balance efficiency is about 92 %, reaching the standard for leaving factory. When the proportion of a single product that was put into production gradually rises, the overall time-consumption of the whole assembly line becomes shorter and shorter and the balance efficiency of the mixed assembly line presents a "U-shape" variation trend, first decreasing and then increasing. The growth of workers doesn't have an obvious impact on the assembling time consumption.