The importance of automated guided vehicles in flexible manufacturing systems is increasing because they can flexibly respond to changes in facilities and factory layouts. Herein, we propose an autonomous conveyance system for automated guided vehicles based on the operation of a taxi transportation system to solve indefinite and accidental problems. The system focuses on the application of traffic engineering knowledge to a flexible taxi transportation system. A taxi is a transport unit in a traffic system, with high adaptability to traveling routes and arrival/departure points. We also propose a task linear density based on transport density, which is an indicator of the scale of transport, such as that of rail transport. We evaluated the efficiency of an automated guided vehicle (AGV) transportation system that introduced waiting and cruising operations, which determine the taxi behavior, using the task linear density. The results indicate that the average matching time and task linear density were negatively correlated and that the transport efficiency improved when the P/Ds were frequently transported to a centralized location.