SPIRE: Scalable Processing of RFID Event Streams

1
SPIRE Scalable Processing of RFID Event Streams

• Mian Ahmad Zeb
• ID 200955425
• Department of Computer Engineering

2
Contents

• Introduction
• Architecture
• Technical Components
• A. Filtering and Smoothing
• B. Data Compression
• C. Event Processing
• D. Global Processing
• RFID Supply Chain Simulator
• Status and Future Work

3
General System Deployment of RFID Application
This paper focus on transforming the raw data
4
Recent Technical Trends in RFID

• Todays trend that physical objects are tagged.
  
• Work on RFID tagging sensing to combine with
• ubiquitous n/working to make information
• infrastructure.

5
Four (4) Fundamental Problems with RFID Technology

• Data-information mismatch for monitoring
• 2. Incomplete, insufficient data for
  track-and-trace
• 3. Scalability.
• 4. Latency

6
Proposed System

• Radio frequency identification (RFID) is an
  automatic
• identification method, relying on storing and
  remotely retriev- ing data using devices called
  RFID tags.
• SPIRE (Scalable Processing for RFID Event Stream)
  is to manage
• RFID data and fast information transformation.

7
Proposed Solutions (1/3)

• Techniques of filtering and smoothing are used to
  improve quality of data.
• The techniques have further processes
• Data Cleaning
• Data Compression
• Event Processing
• Distributed Event Processing

8
Proposed Solutions (2/2)

• Data Cleaning-
• A layer is introduced directly to the readers,
  to filter
• abnormal and corrupted reading, remove
  duplicate readings, and smooth readings to
  assist in recreating missing data.
• Data Compression -
• Compression technique are used to reduce
  overall data volume originated from the readers
  and support precise location tracking and event
  detection at same time.
• Event Processing -
• This will extract meaningful information from
  stream of
• data. User define query would monitor the
  incoming compressed tag readings to detect
  patterns and disappeared tag.
• Distributed Event Processing -
• Scalability is achieved from compression
  techniques with
• a distributed architecture. Compressing reduce
  the data
• volume to a scale that become manageable for
  central repository

9
Architecture (1/2)
10
Architecture (2/2)

• The internal architecture is divided into three
  Layers/Components.
• Physical Layer
• Cleaning, Compression and Association Layer
• Complex Event Processor, Event Database

Layer
Physical Device
Complex Event Generation
Cleaning, Compression Association
Queries
Filtering Smoothing
RFID Device
Data Compression
Tag
Event Processing
Antenna
Global Processing
11
Physical Device Layer

• The physical device layer consists of RFID
  readers, antennas,
• and tags.
• RFID readers scan their surrounding areas in
  regular
• intervals and return a reading for each tag
  detected in the form of (Tag ID, Reader ID,
  Timestamp).

Tag ID Reader ID, Timestamp
RFID Tag with Antenna
RFID Reader
12
Cleaning, Compression, and Association Layer

• The middle layer serves three important
  functions.
• (1) First, it manages with idiosyncrasies of
  readers and performs data cleaning, such as
  filtering and smoothing.
• (2) Second, it uses two compression techniques
  to effectively reduce
• data volume between the readers and the event
  processor.
• (3) Third, it uses attributes such as product
  name, expiration date, and saleable state to
  create events.

Read Data from Reader
Filtering Smoothing
compression
Event Generation
13
Complex Event Processor

• SPIRE uses SASE as query language and its
  functions are below
• Detection of missing items.
• Removal of duplicate data through queries.
• Integrate stream processing and database
  lookup.

Complex Event Processor
Query over Stream
Query over History
14
Technical Components

• This section focus on 2nd layer and its
  components and their functionalities.
• FILTERING AND SMOOTHING
• Filtering out invalid tag readings.
• Smoothing tag readings to fill missing gaps in
  the data.

Filtering Smoothing
Tag Data
Temporal Smoothing
Time Conversion
De Duplication
Anomaly Filtering
decides about the object presence
Removal of duplicates
Removal of spurious readings
Appended timestamp
15
Data Compression

• There are two varieties of compression
• LOCATION COMPRESSION
• CONTAINMENT COMPRESSION

16
LOCATION COMPRESSION

• (1) A new event is generated to indicate tags
  arrival.
• (2) No new event is generated except
  confirmation if at same
• location.
• (3) A cache is used to avoid unnecessary
  access to the event
• database.
• (4) If a tag is missed at a reader x times, an
  event is
• generated to indicate that the tag has
  left the reader.

17
CONTAINMENT COMPRESSION

• This compression method is about tagged objects
  move together in groups.
• Supply chain has case of products may be
  packaged.
• Containment relation ship between the case and
  its products.
• Use of the case tag to represent the location for
  both the case and Products.
• To create the containment relationships, two
  options can be used
• Specific readers physically configured
• Containment relationships will have to be
  manually entered.
• Caching mechanisms to reduce the amount of direct
  queries handled by the event database.

18
Trade Off

• Some tradeoffs to consider when utilizing
  containment compression.
• First, creation of containment relationships
• depends on the capacity to configure a
  special reader
• setup.
• (2) Second, it is not always possible to
  directly verify a containment relationship.
• (3) Finally, similar to location compression,
  the number of consecutive readings before a
  relationship is marked as stale needs to be
  carefully chosen.

19
Event Processing

• The Event Processing of SPIRE has two stages.
• (1) EVENT GENERATION
• (2) COMPLEX EVENT PROCESSING

20
EVENT GENERATION

• StartLocation(TagA, Location B, Timestamp),
  EndLocation.
• Generated event used to start and end
  containment relationship.
• Stored directly in event database and transformed
  into desire schema.
• Direct query to events for a specific tag ID
  allow for trace-and-track.
• An important step in event generation is to
  obtain additional attributes (product name,
  expiration date, etc)

21
COMPLEX EVENT PROCESSING

• It is used to allow a user to specify custom
• continuous queries over both the incoming event
• stream and historical data. Queries specified in
• the SASE language.
• FROM ltstream namegt
• EVENT ltevent patterngt
• WHERE ltqualificationgt
• WITHIN ltwindowgt
• RETURN ltreturn event patterngt

EVENT SEQ(PACKAGING_READING x, !(EXIT_READING
y)) WHERE x.TagId y.TagId WITHIN 12
hours RETURN x.TagId, x.ProductName, x.TimeStampc
22
Global Processing

• Additionally allowing for location tracking and
  automated query
• processing.
• For web service, a centralized query system
  provide for easy
• accessibility to real time and historical data
  across an entire
• distributed network of locations.
• This is an area of ongoing research.

23
Assumption on Simulation

• Assumptions made in simulation
• Three types of RFID tagged objects are used
• Pallets, Cases, and Products.
• Cases removed from Pallet and kept in Shelve of
  warehouse, wait
• for some user specified amount of time.
• Ready to be sent to a packaging area and grouped
  into new pallets.
• These new pallets are recorded and sent to
  another warehouse.

24
Conclusion

• SPIRE described main problems face in large
  scale RFID systems.
• Techniques used for data cleaning, compression,
  and event processing
• RFID data event stream is more reliable,
  manageable, and informative
• than raw readings produced from numerous RFID
  readers.
• SPIRE will provide a scalable solution to user
  defined continuous
• queries.