Inquiry and Paging for Bluetooth Piconet and Scatternet Formation in NSProject OverviewProject GoalsLink Controller State MachineAccomplishmentsAccomplishments (cont)Problems EncounteredExample of a specific problemExample of a specific problem(contd)Implications of the problemOn a brighter side...Some Simulation ResultsInquiry and Paging for Bluetooth Piconet and Scatternet Formation in NS Denis PerelyubskiyVadim OlshanskyVitaliy DykhneProject Overview•Implementation of Bluetooth Inquiry and Paging procedures in ns-2 network simulator based on Bluetooth Standard version 1.0b/1.1Project Goals•Implement Link Controller state machine•Implement Frequency Hopping•Simulate node discovery and connection setup to evaluate performance of Bluetooth technology in various scenariosLink Controller State MachineAccomplishments•Done–Frequency Hopping–Inquiry Procedure–Paging Procedure (pending resolution of several serious issues)–Limited Performance Analysis of Inquiry Procedure–3000 lines of codeAccomplishments (cont)•Not Done–Scatternet Formation–Paging Procedure Performance AnalysisProblems Encountered•Our biggest problem is the standard:–General and vague–Insufficiently precise on critical issues such as frequency hopping–Changed recentlyExample of a specific problem•Master needs to operate in both transmit and receive modes while waiting for an ID packet from the bluetooth slave•In page mode “the RX hops are selected according to the page_response hopping sequence” (BT 1.1, 3rd paragraph, page 98)•X control input to the frequency hopping computation is Xprm =[CLKE*16-12+ Koffset* + (CLKE*4-2,0-CLKE*16-12) mod 16 + N]mod32–here, CLKE* is the “frozen” clock, and k-offset is a “frozen” offset. “The values are frozen when the slave ID packet is received” (section 11.3.3.2 of the BT 1.1 standard)Example of a specific problem(contd)•Problem: if the ID packet has not been received from the slave yet, we will not have frozen clock/offset values — what do we do?Implications of the problem•While we can compute some kind of hopping sequences for both nodes involved, it takes them a very long time to randezvous, which hints at the wrong values of inputs to the frequency-hopping function•Using random sequences of some sort during connection establishment is difficultOn a brighter side...•We did implement and test the frequency hop computation–validated using sample data at the end of the specification–this means that our function works utilizing “garbage-in-garbage-out” principle: given good inputs, it will produce proper output ; this means to fix 99% of our problems, we just need to clarify the inputs to the function at several controversial states.Some Simulation ResultsDevice Discovery using Inquiry Procedure 1 Master 1 Slave01020304050601471013161922252831343740434649525558TrialsDelay