Moti Abu & Roee Ben Halevi
Supervisors: Prof. Mark Last, Mr. Hanan Friedman
SMART DATA SELECTOR
12 December 2010
Telemetry
Velocities Vibrations Pressures ……………..
SNR (dB)
Bit Errors
𝑝𝑒=1
√𝜋 ∫√𝑆𝑁𝑅
∞
𝑒−𝑡2
𝑑𝑡
Bit Errors
Solutions• Best Source Selector
Selects the best source based on received signallevel.
• Best Data SelectorSelects the best data based on SNR and receivedsignal level.
• Correlated Source SelectorSelects the best data bits based on best data fit.
• Smart Source SelectorCombination of all the above.
Solutions
Solutions
Survey12 December 2010
RequirementsSymbol Requirement Description Priority[R1] Quality of output BER in output should be reduced
by at least 90% in all simulationsCritical
[R3] Halt and resume The system should support halting and resuming master generation
Critical
[R5] User interface The system should include an ergonomic GUI and command line API for configuration. GUI should be in English
Critical
[R8] Late integration of sources
The system should be able to integrate more inputs to the master output without re-reading of early sources
High
[R9] real-time progress notifications
The system should include real-time progress notifications on GUI and command line interface
High
[R10] Built-in data-mining algorithms
The system should offer built-in data-mining algorithms for optimizing the quality of output
Medium
12 December 2010
RequirementsSymbol Requirement Description Priority[R11] Statistical reports The system should provide a
user-defined statistical reports regarding the quality of output and sources
Medium
[R12] Configuration files The system should allow saving and loading current configuration status to and from configuration files
Medium
[R13] User override The system should allow user to override algorithm decisions
Medium
[R15] API for user-defined plugins and extensions
The system should support user-defined plugins and extensions by providing a user-friendly API
Medium
[R16] Wizard The system should allow users to create new jobs by a user-friendly wizard
Medium
[R20] Integrated probability models
The system should include integrated probability models for bit-error prediction
Low
12 December 2010
RequirementsSymbol Requirement Description Priority[R22] Speed Master output should be
generated at a rate of 1Mb/sec for 5 sources with BER of 2%
Critical
[R23] Reliability The system cannot crash at any circumstances
Critical
[R24] Usability GUI should be user-friendly and simple to manage for the common user. Not overwhelming the user with redundant messages. Every option should be up to 3 clicks away
High
[R25] Modularity The system should enable replacing the provided data selection algorithms and formats by user-defined algorithms and formats
Medium
[R26] Portability The system should work perfectly on Windows environment. Linux version is optional
Low
12 December 2010
RequirementsSymbol Requirement Description Priority[R30] Testing The system should be tested
with artificial benchmarksCritical
[R31] Demo First prototype of the system will include only core components
Critical
[R32] Quality measurements
Quality of output will be measured with pre-defined quality measurements like similarity value and bias.
Critical
12 December 2010
The quest for the best similarity value…
Use Cases12 December 2010
Use CasesUse Case 2 [UC2]: generate masterPrimary Actor: Telemetry technician (or common user).Interests: The technician wants to generate a master output from source inputs.Pre-conditions: User opened a new job, loaded sources and defined a configuration.Post-conditions: The master is generated and saved in the job directory with metadata information.Main success scenario: 1. User clicks on "play" icon. 2. An estimate for the execution time is displayed for the user and he is asked to confirm. 3. After confirmation, master output and metadata information are generated according to user configuration and saved in job directory.Main fail scenario: 1. User clicks on "play" icon. 2. An estimate for the execution time is displayed for the user and he is asked to confirm. 3. User cancels master generation.
12 December 2010
Use CasesUse Case 3 [UC3]: Integrate inputPrimary Actor: Telemetry technician.Interests: The technician wants to integrate input source to master output.Pre-conditions: Configuration file, master output and metadata information are present in the job directory. User loaded inputs.Post-conditions: The new master is generated and saved in the job directory with new metadata information.Main success scenario: 1. User clicks on "integrate sources" icon. 2. An estimate for the execution time is displayed for the user and he is asked to confirm. 3. After confirmation, new master output and metadata information are generated according to user configuration and saved in job directory.Main fail scenario: 1. User clicks on "integrate sources" icon. 2. An estimate for the execution time is displayed for the user and he is asked to confirm. 3. User cancels integration
12 December 2010
Proposed Solution12 December 2010
Dataflow in SolutionExperimental Aircraft
Experimental aircraft transmits data to ground receivers
Ground ReceiversSeveral ground receivers record the data
Preprocess unit (XXX-Telemetry records creator)
The raw data that was recorded in each receiver is formatted and metadata files are created
Smart Data Selector (SDS)
Integrates all inputs to one master record that is as close as possible to the original data
Analysis tools and QuickView Master record is used for analysis of flight and plotting
12 December 2010
Software Context12 December 2010
Minor
Minor
Major
Minor
Minor
Minor
Minor
Minor
Minor
Minor
Minor
Minor
Minor
Minor
Minor
Major
Minor
Minor
Minor
Minor
Minor
Minor
Minor
Minor
Minor
Minor
Minor
Minor
Major
Minor
Minor
Minor
Minor
Minor
Minor
Minor
Minor
Minor
Minor
Survey12 December 2010
Software Context12 December 2010
Master UnitBest
record Pack
Slave UnitSlave Unit
Slave Unit
Best raw dataBest metadata
Master ReadersSchedulerMaster Builder
Output to GUI
Slave ReaderSelecting Algorithm
12 December 2010
Master Reader
Master Unit – Part A
Meta Data
RawData
Top Rated Record
Master Reader
1
Master Reader
N/2]]
Master Reader
N
Part I
Part[N/2]
Part N
Read
Read
Read
MetaData
Threads
Master Record
Builder
12 December 2010
MasterReaders
GoodMinorsQueue
Minor
RawMinor
MetaMinor
Have Sync?Yes
Have CRC?Yes
Pack as GoodMinor!
BadMinorsQueue
RawMinor
MetaMinor
Have Bad Sync?OR
Have Bad CRC?
Pack as BadMinor!
Bad Minor
12 December 2010
GoodMinorsQueue
BadMinorsEntry
QueueBad signature
Scheduler
TaskQueue Task
Queue TaskQueue Task
QueueSlaveSlave
SlaveSlave
Packet
Packet
PacketPacket
ApplyAlgorithm Apply
AlgorithmGoodMinorGood
Minor
SlaveMinor
Signature
Slave ReaderRaw Data
Records
MetaData
RecordsFind Good Minor In Others Records?Yes!
Send Location to Builder
All Records have bad minor?? Apply algorithm .
AlgorithmMachine
Algorithms can be added and/or changed by user (Support Reflection mechanism )
Algorithm examples:
- BitVoting.
-Pattern recognition .
-User Algorithm.
1101010
1001010
1001000
1001010
1001010
Origin data
Station A
Station B
Station C
BitVote
Master Record
Builder
GoodMinorsQueueGood
Minor
GoodMinor
To GUI
Block Diagram12 December 2010
QUESTIONS…
12 December 2010