We consider the dynamic scheduling of a pickup and delivery system in which mobile robots are used to transport materials among a set of places. With the help of Internet of Thing, the system has real-time information for demands of transportation and status of robots. How to dynamically scheduling the robots to fulfill the transportation tasks is a key challenge. We propose a method for it in which an innovative and
adaptive mixed integer programming model (MIP) is developed. This MIP model can be called in a periodic or event-driven or hybrid way. Numerical results show it can solve real problems fast and significantly reduce transportation cost, comparing with a heuristic policy.