Metro network works as a backbone of the public transport system for metropolitans. Feeder buses are designed to connect major metro stations and homes of travelers so as to solve the last mile problem. With the recent implement of the real-time bus timetable program in Shanghai, feeder buses could better coordinate with metro to further improve the public transport patronage. This paper proposed a linear program to optimize feeder bus schedule based on GPS and smart card data. We consider the feeder bus operation in a multi-mode transit system which contains two modes: bus and metro. Reliability of the feeder buses are specifically considered in setting the headway. In case study section, 5 representative feeder bus lines in Shanghai are selected to investigate their performance changes before and after the headway adjustment. The results demonstrate the applicability of the proposed optimization method. Headway adjustment together with the real-time time table program could significantly decrease travelers' travel time and smooth travelers departure time from home.