Agents hired by on-demand platforms to serve users have freedom to choose orders they will serve, and they may strategically idle themselves to avoid some orders. It is necessary to design reasonable mechanism to regulate agents' behavior. To begin, we model the agents' equilibrium strategy of accepting or rejecting an order, and identify the trade-offs of agents' strategic idling. Then, we proposed two mechanisms to effectively eliminate the strategic idling behavior based on the concept of punishment and subsidy, respectively. Under the punishment mechanism, the platform reduces the frequency of assigning a more beneficial order to the agents if they rejected to serve less beneficial orders, thereby decreasing the unit expected profit of the agents who strategically idling themselves. Under the subsidy mechanism, the platform subsidizes agents if they are serving a less beneficial order to increase the unit expected profit of the agents who do not reject any order.