We consider the problem faced by a carrier who needs to manage a fleet of homogeneous vehicles over space and time to serve a number of transportation requests. Each transportation request is defined by an origin–destination pair. The service for the request must start at a specific instant in time or be lost. The carrier may provide customers a portfolio of service levels that vary in travel time (24 h, 48 h or 3 days). For example, when customers place orders from E-commercial companies (such as amazon.com, 360buy.com, Taobao.com), they can choose delivery service level ranged from 3 days, 2 days and 1 day with different rates. Carriers, given orders with various service level and demand, need to allocate their fleets to fulfill the tasks. If customer demands a service level with shorter travel time, the carrier may need to pay additional transportation cost and ask for a higher rate. Customers select one service level when they place the order. This problem belongs to a class of dynamic fleet management problems (DFMPs), which can be viewed as a type of spatial, dynamic inventory management problem with reusable vehicles. At any point in space and time, a vehicle may be assigned to satisfy a revenue generating activity. It may be repositioned empty to another point in space and time or held in inventory.