Radio Frequency Management and Cognitive Engine Initial Results of the C-PMSE Project

Transcription

1 Radio Frequency Management and Cognitive Engine Initial Results of the C-PMSE Project Leonid Tomaschpolski Institute of Communications Technology Leibniz Universität Hannover December 7, 2011

2 C-PMSE System Architecture Cognitive Engine is the brain of the system 2

3 CEN: Architecture BS CEN CID DB GDB IA IM LQI LSPM RLM SCS List of Abbreviations Base Station Cognitive Engine Cochannel Interference Detection Data Base Geolocation based Data Base Information Acquisition Intermodulation Product Link Quality Indicator Local Spectrum Portfolio Management Radio Link Manager Scanning Controller Subsystem 3

4 CEN: Tasks Negotiation with Local Spectrum Portfolio Management (LSPM) Management of Radio Links: Assignment of wireless microphone frequencies to available radio frequency spectrum taking into account Service- Level-Agreements (SLA) Information Acquisition (IA): Assignment of tasks to Scanning Controller Subsystem (SCS) Interference rating and risk rating 4

5 CEN: Start Up Sequence (simplified) Challenges Initial frequency planning intermodulation-free frequencies optimization procedure time-consuming process Why is frequency planning not always applicable? highly dynamic environment multiplicity of potential interferers 5

6 CEN: Operating Sequence (simplified) Reaction on interferer Reactive case Interferer is recognized by abrupt degradation of link quality immediate reaction is essential for operational reliability Proactive case CEN detects potential interferer before link quality decreases proactive switchover, previous to occurrence of interference 6

7 Messe Berlin Scenario (simplified) Exemplary Position of scanning receivers in a hall on Messe Berlin in case of signal detection the outcome of SCR is 1, otherwise 0 case Y: three out of nine scanning receivers detect a signal 7

8 Basics of Case Based Reasoning Possible (simple) approach for case representation case = problem+solution case as a vector of attribute-value pairs store previous cases (experience) in the case base compare the new problem with each case and select the most similar case from the case base Example for global similarity measures for case x and case y example constraint: attribute-values are binary case 1 is x= ( ) and case 2 is y= ( ) similarity measure with Hamming Distance: sim H (y, x)=

9 Should we initiate proactive reaction now? Case X 1 Case X 2 Case X n Case Y 9

10 Considered AI Techniques Case Based Reasoning (CBR) Close to human reasoning Relies solely on previous case Requires large case memory Rule Based Systems (RBS) Simple implementation Tedious rule derivation process Requires perfect domain knowledge, which is not always available Artificial Neural Network (ANN) Ability to describe multitude of functions Can identify new patterns Training may be slow, depending on network size No theory to link application with required network Source: He et al., A Survey of Artificial Intelligence for Cognitive Radios, IEEE Transactions on Vehicular Technology,

11 Research and Development Steps 1. Efficient initial frequency planning 2. Base-system with reactive behaviour using Link Quality Indicator trigger 3. Radio ressource management including cognitive algorithms and risk rating of frequencies 4. System with mechanisms for proactive behaviour using sensor data and optimization and decision making algorithms 11

12 Questions Acknowledgement The C-PMSE project is co-funded by the German Federal Ministry of Economics and Technology (BMWi) 12