Determining physical location of indoor objects is one of the key issues in development of context-aware applications in ubiquitous computing environments. This is mainly because context information obtained from sensor networks is meaningful only when the physical location of the context information source is determined. Recently, several indoor location information systems, such as Active Bat and Cricket, have been developed for precise indoor object localization. However, to provide accurate physical location tracking in large-scale space, those systems requires a lot of manual configuration for all the ultrasonic sensor nodes. To reduce configuration costs, we developed a new indoor positioning system called DOLPHIN. The DOLPHIN system consists of distributed wireless sensor nodes which are capable of sending and receiving RF and ultrasonic signals. These nodes are attached to various indoor objects. And using a novel distributed positioning algorithm in the nodes, DOLPHIN enables autonomous positioning of the objects with minimal manual configuration. This paper describes the design and implementation of the DOLPHIN system, and evaluates basic performance through several experiments in an indoor environment.