Development of a Sensor Database System for Communication

Development of a Sensor Database System for Communication Robots Hideyuki KAWASHIMA Graduate School of Keio University Yokohama, JAPAN

Background: Human-Robot Communication • Sensors – – Video Microphone Touch sensor Ultrasonic sensor

Required properties • Monitoring sensor data during experiments • Persisting precious sensor data immediately • Retrieving similar sensor sequences

Problem formulation 1. Continual query 2. Freshness of data 3. Similar sequence retrieval

Problem 1: Continual query Monitoring Client Backend Sensor DB 1 Data 23 IPC waste & period lag DBMS

Problem 2: Freshness of Data Sensor Process Stale data because of Tardy disk access Database backend Log file Buffer Pool

Problem 3: Similar Sequence Retrieval • Metrics – Euclid: Σ (Di-Qi) 2 = 2 – Dynamic time warping: MIN{ Σ Wk } =√ 2/5 K Wk = A distance on a path 5 16 9 4 3 4 1 2 1 1 4 1 1 0 0 1 2 3 4 4 2 1 3 2 4 3 5 4

My approach: Extending RDBMS for problems 1. Continual query 2. Freshness of data 3. Similar sequence retrieval

Continual query Monitoring Client Sensor Register CQ Transfer Backend DB 1 Data Monitoring Data 23 Backend DBMS Efficient IPC & Exactly periodic monitoring

Freshness of data Sensor Process Database backend Buffer Pool Network Remote log server & Check pointer

Similar sequence retrieval • Select sonar from robovie where simseq external [1, 2, 3] with dtw < 10 Sliding window: 1 id 1 2 Experiment_date 01/05/2004 02/05/2004 2 3 sonar

Design of KRAFT Trans. Mgr Index Mgr Parser Mem. Mgr Lock Mgr Executor Recovery Mgr Conn. Mgr TCP Client Storage Mgr Local Log Mgr Mem. Mgr Remote Log Mgr TCP Over 17000 lines by C

Future directions • Enhancing analytical methods for sensor data • Scalability for continual monitors • Accelerating query processing – Index for similarity retrieval – Parallel execution – Device conscious

Summary • Motivation – Management of data for human-robot interactions • Problems – Continual query – Freshness of data – Similar sequence retrieval • My approach – Extending RDBMS for problems • Future directions