Adaptive Cleaning for RFID Data Streams

RFID: Radio Frequency IDentification

RFID data is dirty A simple experiment: 2 RFID-enabled shelves 10 static tags 5 mobile tags

RFID Data Cleaning Time Raw readings Smoothed output RFID data has many dropped readings Typically, use a smoothing filter to interpolate SELECT distinct tag_id FROM RFID_stream [RANGE ‘5 sec’] GROUP BY tag_id SELECT distinct tag_id FROM RFID_stream [RANGE ‘5 sec’] GROUP BY tag_id Smoothing Filter

Smoothing filter Middleware Clean RFID Completeness Tag dynamics Read all tags in range

RFID Data Cleaning Time Raw readings Smoothed output RFID data has many dropped readings Typically, use a smoothing filter to interpolate SELECT distinct tag_id FROM RFID_stream [RANGE ‘5 sec’] GROUP BY tag_id SELECT distinct tag_id FROM RFID_stream [RANGE ‘5 sec’] GROUP BY tag_id But, how to set the size of the window? But, how to set the size of the window? Smoothing Filter

Window Size for RFID Smoothing Fido movingFido resting Small window Reality Raw readings Large window  Need to balance completeness vs. capturing tag movement

Truly Declarative Smoothing Problem: window size non-declarative  Application wants a clean stream of data  Window size is how to get it Solution: adapt the window size in response to data

RFID EpochTagIDReadRate Tag 1 Tag 2 Tag 3 Tag 4 Antenna & reader Tags E1E2E3E4E5E6E7E8E9E0 Read Cycle (Epoch) (For Alien readers) Tag List 1. Interrogation cycle 2. Epoch

Controlled condition real condition

SMURF Statistical Smoothing for Unreliable RFID Data Adapts window based on statistical properties Mechanisms for: Per-tag and multi-tag cleaning

Per-Tag Smoothing: Model and Background Epoch t, Tag population N t p i,t : Per epoch sampling prob. Response count of tag i per epoch (total interrogation cycle) EpochTagIDReadRate

Smoothing window size w i epoch Per epoch sampling prob: p i Number of successful observations of tag i  Binominal distribution B(wi,pi) Per-Tag Smoothing: Model and Background

Use a binomial sampling model Time (epochs) pipi 1 0 Smoothing Window w i Bernoulli trials p i avg SiSi (Read rate of tag i) E1E2E3E4E5E6E7E8E9E0 Set of epochs where tag i can be seen

We want to ensure that there are enough epochs in Wi such that tag i is observed (if it exists within the reader’s range)  Completeness Per-Tag Smoothing: Completeness

If the tag is there, read it with high probability  Want a large window pipi 1 0 Reading with a low p i Expand the window Time (epochs) E1E2E3E4E5E6E7E8E9E0

Per-Tag Smoothing: Completeness

Expected epochs needed to read With probability 1-  Desired window size for tag i

Per-Tag Smoothing: Transitions Detect transitions as statistically significant changes in the data pipi 1 0 Statistically significant difference Flag a transition and shrink the window The tag has likely left by this point Time (epochs) E1E2E3E4E5E6E7E8E9E0

Significant difference between mean observed sample size Si and expected size Find outlier (2  ) Number of successful epochs in a window SiSi Mean

Per-Tag Smoothing: Transitions # expected readings Is the difference “statistically significant”? # observed readings Statistically significantStatistically significant

Algorithm

SMURF in Action Fido movingFido resting SMURF  Experiments with real and simulated data show similar results

Normal sliding windowCompleteness Transition