A Comment on Black-Box adaptation for simulation methodology January 2002 doc.: IEEE 802.11-02/xxxr0 March 2004 A Comment on Black-Box adaptation for simulation methodology Adrian P Stephens Adrian.p.stephens@intel.com Intel Corporation Adrian Stephens, Intel Corporation John Doe, His Company
The universe of possible adaptation mechanisms March 2004 The universe of possible adaptation mechanisms The universe of possible 802.11n solutions includes: ABL MIMO: Open-loop & SVD Varying rates of adaptation Fast (per-exchange) Slow (multiple exchanges, statistical analysis, long processing time) + other’s this author is not aware of Adrian Stephens, Intel Corporation
Why is adaptation important? March 2004 Why is adaptation important? It is our “job” to use radio resources effectively Current 802.11a/g systems adapt using heuristics based on inaccurate observables (e.g. PER). Current systems find it easy to “adapt down” but harder to “adapt up” This author expects all proposals will include some method to make adaptation to the channel explicit Adrian Stephens, Intel Corporation
How does adaptation work? January 2002 doc.: IEEE 802.11-02/xxxr0 March 2004 How does adaptation work? Adaptation algorithms respond to observed events and retained state: Instantaneous Channel State observations Averaged Channel State Interference environment Statistics of power level vs periodicity Long term per STA PER Previous packet PER, BER, quality Adrian Stephens, Intel Corporation John Doe, His Company
March 2004 The problem of bias This author cannot see how the Black-box proposal could support the universe of reasonable adaptation 802.11n solutions without introducing much of the complexity of the MAC simulation into it – thereby rendering it useless as a simplification and undefined as a standard interface Therefore I see that making this proposal a mandatory requirement of 802.11n introduces bias because it does not permit a reasonable class of solutions to be proposed It is our duty to consider all reasonable solutions without introducing bias in our process Adrian Stephens, Intel Corporation
March 2004 Conclusion In the view of the author, attempting to standardize a mandatory MAC-PHY simulation interface that attempts to include the effect of link adaptation mechanisms would introduce bias into our selection process and is therefore counter to our policies and procedures Adrian Stephens, Intel Corporation