[ath9k-devel] Hack ath9k driver for wifi time of flight measurement

[ath9k-devel] Hack ath9k driver for wifi time of flight measurement

[王峰]

Hi All,


I am pretty new to ath9k driver. But I read a recent paper by MIT that the channel state information (CSI) of wifi can
be used for precision positioning. They used Intel 5300 wifi module which has CSI tool support. Some of the
ath9k devices, such as ar9462, ar9380 etc are also supported by an open source CSI tool. So I want to use them
to reproduce MIT's method.


I have already built a wifi CSI test system with two computers using ath9k driver. They can get the CSI readings
out fine (carrier phase). But I need to be able to control the band (2.4G, 5G) and channel of the wifi. Preferably loop
through 2.4 and 5G bands, and every channel of them. For each channel, I need to read CSI for both computer
while one is transmitting and other is listening then reverse the transmitting and listening. What is the
easiest thing to do to? 


The ath9k CSI tool can be found here:


http://pdcc.ntu.edu.sg/wands/Atheros/


MIT's cool method for accurate wifi positioning can be found here:


http://people.csail.mit.edu/deepak/Slides/NSDI2016_Chronos.pdf


Thanks in advance and Best regards,


Everett






 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.ath9k.org/pipermail/ath9k-devel/attachments/20160506/27b9db72/attachment.htm 

[ath9k-devel] Hack ath9k driver for wifi time of flight measurement

[schnitzbauer at teco.edu]

Hello Everett,
I'm currently heading towards the same experiment setup for another  
usecase. I use two tl-wn722n with ath9k_htc attached to a raspberry pi.

I talked to Mathy Vanhoef (Mathy.Vanhoef at cs.kuleuven.be) and he  
confirmed what i already thought while checking the transceiver block  
diagram and Adrian Chadd wrote in the freebsd spectral wiki page.

You can either get spectral information of demodulate the frame for  
beeing processed through the protocol stack.

He recommends either one of the following setups, which intuitively  
come to mind:
1. Detect the start of packets based on the increased signal strength  
that can be seen in the spectral scan samples.
2. Use two devices as receiver. One decodes the frames, the other  
captures spectral scan. Synchronize the captures based on the MAC  
timestamps produced by both chips.

In my eyes a hybrid solution, doing a measurement start with a frame,  
synchronizing on that and then performing the actual measurement is  
the way to go. Starting the measurement from within the firmware on  
the chip is necessary in my eyes for reliable synchronization due to  
the unpredictability and delays if events are processed on linux  
processor.
You may know this paper: dl.acm.org/ft_gateway.cfm?id=2000029 .  
Domenico Giustiniano did a busy loop within the driver code to solve  
the problem with unpredictable scheduling. He is currently working on  
a broadcom platform with a custom firmware

There is a tx99 mode  
http://www.spinics.net/lists/linux-wireless/msg113353.html, which may  
be the TX operation i would prefere.
Having the tx99 mode in ath9k_htc, too would be helpful.


Greetings,

V. Schnitzbauer


Zitat von ?? <everteq@163.com>:

> Hi All,
>
>
> I am pretty new to ath9k driver. But I read a recent paper by MIT  
> that the channel state information (CSI) of wifi can
> be used for precision positioning. They used Intel 5300 wifi module  
> which has CSI tool support. Some of the
> ath9k devices, such as ar9462, ar9380 etc are also supported by an  
> open source CSI tool. So I want to use them
> to reproduce MIT's method.
>
>
> I have already built a wifi CSI test system with two computers using  
> ath9k driver. They can get the CSI readings
> out fine (carrier phase). But I need to be able to control the band  
> (2.4G, 5G) and channel of the wifi. Preferably loop
> through 2.4 and 5G bands, and every channel of them. For each  
> channel, I need to read CSI for both computer
> while one is transmitting and other is listening then reverse the  
> transmitting and listening. What is the
> easiest thing to do to?
>
>
> The ath9k CSI tool can be found here:
>
>
> http://pdcc.ntu.edu.sg/wands/Atheros/
>
>
> MIT's cool method for accurate wifi positioning can be found here:
>
>
> http://people.csail.mit.edu/deepak/Slides/NSDI2016_Chronos.pdf
>
>
> Thanks in advance and Best regards,
>
>
> Everett

[ath9k-devel] Hack ath9k driver for wifi time of flight measurement

[Robert Felten]

Hi,

2016-05-07 9:16 GMT+02:00  <schnitzbauer@teco.edu>:
> ...
> 1. Detect the start of packets based on the increased signal strength
> that can be seen in the spectral scan samples.

Are you sure that this is possible? In my observation your spectral
sample stream get interrupted by an decoded frame.
The start of the frame is not visible and there is a small gap between
the last spectral sample and the start of the decoded frame.

If you merge the power information of the spectral samples and frame
it looks like http://imgur.com/xAWleMY

X-axis: subcarrier frequencies 1 per px,  y-axis: time, 1 pixel =
115us, captured tl-wn722n with ath9k_htc v1.4
Not sure about the overlap at the end of the frame, probably my
tx_time calculation is not accurate enough.


> 2. Use two devices as receiver. One decodes the frames, the other
> captures spectral scan. Synchronize the captures based on the MAC
> timestamps produced by both chips.

Please share how it is possible to convince the hardware not to
decoded the frame and give only spectral samples instead.


> You may know this paper: dl.acm.org/ft_gateway.cfm?id=2000029 .
> Domenico Giustiniano did a busy loop within the driver code to solve
> the problem with unpredictable scheduling. He is currently working on
> a broadcom platform with a custom firmware

The full text paper is also available:

http://dl.acm.org/citation.cfm?id=2079306
http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.381.1215&rep=rep1&type=pdf

> ...

Best,

Robert