[STDS-802-11] AIML TIG discussion 11-22/1934r1

Thread Links	Date Links
Thread Prev	Thread Next	Thread Index	Date Prev	Date Next	Date Index

Dear all,

Thank you so much for attending AIML meeting yesterday. Below is the summary of comments/questions raised during the presentation of 11-22/1934r0. Attached please find the updated document 11-22/1934r1 incorporated with the comments/suggestions raised in the meeting.

Please let me know if I miss any question in the meeting.

Thanks,

Zinan

Section 2.1 Use case 1 CSI feedback compression

Q1: How often the model needs to be updated if the channel varies
Q2: KPI metrics may need to add

AIML additional overhead, which may include the model parameters
AIML algorithm computation complexity

Q3: KPI metric PER need to consider both MU-MIMO and SU-MIMO cases

In SVD based feedback, the user selection in MU-MIMO may not have significant impact as PMI based MU-MIMO

Q4: Are there any results to show AI based solution vs non AI based solution ( which may be the enhanced version of existing standard)?

One contribution from Samsung

Section 3.1 Requirements:

Q5: Performance consideration: AIML overhead may include data used for AIML model training and AIML model parameters
Q6: the integration and backward compatibility with legacy 802.11 is the outcome not the requirement
Q7: latency due to additional AIML processing (e.g., compression/decompression) may need to be considered
Q8: 10% target PER can be considered as well
Q9: latency can be the metric for KPI/performance evaluation
Q10: generalization of AIML model may need to be considered, fewer update of the model for different configuration

To unsubscribe from the STDS-802-11 list, click the following link: https://listserv.ieee.org/cgi-bin/wa?SUBED1=STDS-802-11&A=1