ICACT20210129 Question.1
Questioner: gueye.ibrahima1@esp.sn 2021-02-09 오후 1:20:25 |
ICACT20210129 Answer.1
Answer by Auhor mrobaei.dev@gmail.com 2021-02-09 오후 1:20:25
Chrome  Click!! |
| Hello,
what performance you bring compared to what existed |
Thank you for your question. The pilot overhead reduction is the main performance gain in this paper. Channel estimation using compressed sensing requires 60 to 100 measurements to estimate the channel. We observed that the proposed channel tracking could reduce the pilot overhead by more than 30% on average while preserving spectral efficiency. A weaker convergence in mean square can decrease pilot overhead further with a cost of less accurate channel estimation and spectral efficiency loss. |
ICACT20210129 Question.2
Questioner: yfujii@kke.co.jp 2021-02-08 오후 4:54:51 |
ICACT20210129 Answer.2
Answer by Auhor mrobaei.dev@gmail.com 2021-02-08 오후 4:54:51
Chrome Click!! |
| Thank you for your very interesting presentation. On the slide.07, you assumed that the maximum speed Tx and Rx is 3.6kmph, 1m/sec, slower than pedestrian's move. Could you tell me what is the rationale for the limitation? |
Thank you for your question. 3,6kmph is a rough and a little bit conservative number from the 3GPP (document TR 38.901). 3GPP recommends inequality (speed*time interval<1m) at page 50 for spatial consistency. Also, for UMI and UMa scenarios, 3GPP recommends 3kmph in Table 7.2-1 page 20. As I mentioned, 3.6kmph is a number in the range. Here is the reference for the 3GPP.
[1] 3GPP, “Technical specification group radio accehttp://icact.org/images/b_write.gifss network; study on channel model for frequencies from 0.5 to 100 GHz (Release 15),” TR 38.901 V15.0.0, June 2018.
|
ICACT20210129 Question.3
Questioner: tetsuya-iye@kke.co.jp 2021-02-09 오전 10:43:10 |
ICACT20210129 Answer.3
Answer by Auhor mrobaei.dev@gmail.com 2021-02-09 오전 10:43:10
Chrome Click!! |
| Thank you for your nice presentation. This may be a very basic question, but what kind of situation do you expect in the time variation in channel because of an antenna misalignment. |
If we consider the dynamics of the environment, mobility of the transmitter (if not fixed), and mobility of the receiver, then antenna misalignment could lead to time-variable error in the AoA/AoD. The variable error may change the spectral properties ( eigenvalues and eigenfunctions) of the channel. The spectral variation leads to a time-variant channel. |
ICACT20210129 Question.4
Questioner: xian236024@163.com 2021-02-09 오후 5:38:43 |
ICACT20210129 Answer.4
Answer by Auhor mrobaei.dev@gmail.com 2021-02-09 오후 5:38:43
Chrome Click!! |
| Hello,it's a pleasure to enjoy you presentation.Have you measured the effectiveness of channel estimation?How reliable is it? |
Thank you for your question. Since channel tracking uses prior knowledge obtained from channel estimation, we expect that channel tracking NMSE to follow the channel estimation NMSE. In other words, an accurate channel estimation should be reflected in the channel tracking in terms of NMSE.
We measured the effect of channel estimation on channel tracking as a function of SNR and the number of measurements, as shown in slide 12.
First, as the SNR increases, the channel estimation NMSE decreases. The figure on the left shows this effect, where the channel tracking NMSE (determined with different gamma values) follows the shape of the channel estimation NMSE and decreases as the SNR increases.
Second, as the number of measurements increases, the channel estimation NMSE decreases. We expect the channel tracking NMSE to decrease as the number of measurements increases for the same reason we discussed above. The figure on the right in slight 12 shows this effect.
In conclusion, the channel tracking NMSE follows the channel estimation NMSE with some degradation. Compared to other papers (discussed in the numerical section), the proposed method is more reliable. For example, at 0dB, the channel tracking NMSE is about -10dB, which is comparable with most of the channel estimation methods in the literature. However, I believe we need to apply it to real data for actual reliability analysis.
|
