Development of a Sensor Database System for Communication Robots
Hideyuki Kawashima · 2004
To gratify researchers of human-robot interactions, a database system should solve three problems: (1) Freshness and persistence of sensor data, because they not only monitor the current condition of robots during experiments(e.g. checking motor error), but also analyze the stored sensor data after the exper-iments(e.g. calculating similarity and correlation). (2) Continual monitoring of sensor data, because they require to monitor conditions of robots continually during experiments. (3) Similar sequence retrieval methods for stored sensor data, because they require to understand a series of contexts from experimental results during or after the experiments. To solve all of the three problems, I propose a new sensor database sys-tem, Kernel of Real-time Active & Fresh Time-series data manager(KRAFT). To solve problem (1), KRAFT applies an enhanced remote logging method[1] for WAL. To solve problem (2), KRAFT provides a service to make functions that periodically monitors incoming sensor data. To solve problem (3), KRAFT provides for clients with a new abstract data type that is permitted to retrieve