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

Re: [STDS-802-11] AIML TIG teleconferences announcement



--- This message came from the IEEE 802.11 Working Group Reflector ---

Hi Ziyang,


Thanks for your answers.

 

One more question to your algorithm: how does the AI enabled algorithm you proposed integrate with EDCAF?

 

Thanks,

Zinan

 

 

From: guoziyang <guoziyang@xxxxxxxxxx>
Sent: Monday, December 19, 2022 11:17 AM
To: Zinan Lin <Zinan.Lin@xxxxxxxxxxxxxxxx>; STDS-802-11@xxxxxxxxxxxxxxxxx
Subject: RE: AIML TIG teleconferences announcement

 

 

Hi Zinan,

 

Thanks for the questions. Please find my responses to your questions below. Any further comments are welcome.

 

  1. What are the parameters that the algorithm tries to determine? Is the algorithm to determine best idle time slots (corresponding to DIFS in the conventional CSMA/CA operation) or the best backoff time (after channel is idle for DIFS)? Or the algorithm is simply to determine the best time to transmit? Will the algorithm follow the same slot based channel access algorithm defined currently in the spec?

 

The algorithm determines to transmit or to wait when the channel is sensed idle. It follows the slot based channel access algorithm defined currently in the spec, but has no backoff. That is, the STA needs to wait for DIFS and perform carrier sensing before transmitting. If the channel is sensed idle, it uses AI to decide to transmit or wait; if the channel is sensed busy, it should wait for another DIFS before making a decision.

 

  1. Do you consider any TXOP reservation in the simulation? What is the algorithm to obtain the TXOP? Would the AIML enabled STAs obey TXOP reservation/NAV by other STAs? If so, how would that work in the proposed algorithm?

 

Actually, in the simulation, we set every transmission lasts for a fixed number of time slots 120, which is equivalent to a fixed 1.08ms TXOP. The method to obtain the TXOP can be RTS/CTS defined in current spec, i.e., sending RTS/CTS to reserve TXOP when the AI algorithm decides to transmit. I think AIML enabled STAs can obey TXOP reservation by other STA once we set the rules for them.

 

  1. Simulation questions:
  1. I assume that saturated traffic means it is full buffer traffic, i.e., the STA always has a traffic to deliver.  In this case, once a STA grabs the channel, then this STA will always transmit? Did you put any constraint on the time duration used by one STA? I am just wondering why IDQN-1 always grabs the channel and other STAs have zero throughput (after 1s) in Figure 4.a [8]? Similar behavior occurs in Figure 4.c [8].
  2. How do you define the delay in the simulation of 11-22/1942 and [8]?

 

a. In the simulation, we constraint the time duration used by the STA to 120 slots (1.08ms) once a STA grabs the channel. The phenomenon in Figure 4 is just due to that IDQN does not converge to a good policy that achieve good total throughput and fairness. The learning result is that one agent always transmits while others always wait.

 

b. The delay of a packet is defined as the duration from the time when it enters the queue to the time when it is successfully transmitted.

 

 

Best regards,

Ziyang

 

 

Section F, Huawei Industrial Base, Bantian,

 

Longgang District, Shenzhen, 518129 P.R.C

 

 

 

发件人: Zinan Lin [mailto:Zinan.Lin@xxxxxxxxxxxxxxxx]
发送时间: 20221215 6:22
收件人: guoziyang <guoziyang@xxxxxxxxxx>; STDS-802-11@xxxxxxxxxxxxxxxxx
主题: RE: AIML TIG teleconferences announcement

 

Hi Ziyang,

 

Thanks for presenting 11-22/1942.

 

After reading the reference paper [8], we still have the questions on the algorithm you proposed on 11-22/1942.

 

  1. What are the parameters that the algorithm tries to determine? Is the algorithm to determine best idle time slots (corresponding to DIFS in the conventional CSMA/CA operation) or the best backoff time (after channel is idle for DIFS)? Or the algorithm is simply to determine the best time to transmit? Will the algorithm follow the same slot based channel access algorithm defined currently in the spec?

 

  1. Do you consider any TXOP reservation in the simulation? What is the algorithm to obtain the TXOP? Would the AIML enabled STAs obey TXOP reservation/NAV by other STAs? If so, how would that work in the proposed algorithm?

 

  1. Simulation questions:
  1. I assume that saturated traffic means it is full buffer traffic, i.e., the STA always has a traffic to deliver.  In this case, once a STA grabs the channel, then this STA will always transmit? Did you put any constraint on the time duration used by one STA? I am just wondering why IDQN-1 always grabs the channel and other STAs have zero throughput (after 1s) in Figure 4.a [8]? Similar behavior occurs in Figure 4.c [8].
  2. How do you define the delay in the simulation of 11-22/1942 and [8]?

 

 

Thanks,

Zinan

 

 

 

From: Xiaofei Wang <00001995ce968e76-dmarc-request@xxxxxxxxxxxxxxxxx>
Sent: Tuesday, November 15, 2022 10:37 PM
To:
STDS-802-11@xxxxxxxxxxxxxxxxx
Subject: [STDS-802-11] AIML TIG teleconferences announcement

 

--- This message came from the IEEE 802.11 Working Group Reflector ---

 

Dear all,

 

I am announcing two AIML TIG teleconferences on the following dates at 10 am – 12 pm ET (two hours):

 

Dec 12, 2022

Jan 9, 2023

 

Best regards,

Xiaofei Clement Wang

Principal Engineer | InterDigital

T: (631) 622.4028

E: Xiaofei.wang@xxxxxxxxxxxxxxxx

 


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


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