What I've learned from MobiCom2013

Really learned a lot from this conference. I list a few notes here for record and I'd like to share them with you guys as well. 

1. Best paper. 

WiSee: Whole-Home Gesture Recognition Using Wireless Signals. Out of question, the presentation is the best. But seems that the TPC judge the quality purely by the paper. It does shed some lights to lead some promising research directions: can we utilize current wireless signals to do something cool? Answer is yes. 

Check the best paper in 2011 "Detecting Driver Phone Use Leveraging Car Speakers" as well. A growing number of app&wireless type of papers in MobiCom and they really manage to solve problems in daily life no matter the scenario is big or small. 

What I learned from WiSee's presentation: a. make a good beginning for your presentation by making clear the problem you are solving. the speaker use 1/3 amount of time to do this and the feedback is good b. Do not make any assumption like the audience knew the technique you use. c. pictures need to be big enough in slides to illustrate scenarios. 

2. Best apps. 

PiCode: http://www.logo-code.com/ QR code is ugly. PiCode makes it good-looking. I talked to the author and they said the processing time is less than 1 second. They did not think through whether they want to make it open source or not. 

iSleep: sleep quality analysis(not just duration as we did) comes into app. the idea is controversial. somebody wonders how does the phone measure different individuals on a same bed..

VideoBee: cool app. there are indeed a lot of wireless video performance problems to be solved. 

3. Student researcher competition winner. 

Tan Zhang. He is a good example for all the chinese Ph.Ds. Asking questions, presenting his ideas fluently and talk to the senior attendees. I should learn from him. 

Yan Yang. (Measuring Human Queues Using WiFi Signals). Interesting work. Same as "Detecting Driver Phone Use Leveraging Car Speakers", it comes from the WINLab from Rutgers(www.winlab.rutgers.edu/docs/research/research.html). It does not require users to  install mobile phone apps but predict the length of waiting queue. 

4. Related work. 

"ParkSense". Ironically the author cannot answer the energy issue very well(wifi keep scanning) but the paper is in the "energy" session. The paper is good example to incorporate mobile sensing and wireless network. But the author did not answer the questions very well. 

"Exploring the Potential in Practice for Opportunistic Networks amongst Smart Mobile Devices. " Good talk, very intuitive. As Henry points out, opportunistic network is retiring technique for a reason that the information gain from it is not comparable as the "energy and time" consuming cloud services. The value of this paper is the dataset itself(two years, 200 freshmen, primary phones). 

5. Localization Panel. 

The panel speakers shared a lot of insight on in-door localization. The debate was 1) what would be the future direction of research in this area, and what would be a good criteria for good research in this topic since there are already tons of papers talking about it. Victor claims to set a threshold a detection accuracy because the ultimate goal is cm-scale positioning. Romit does not think so and claimed the community should still promote novel positioning system with a relative loose metric of positioning accuracy. I don't think the ideas conflict - as Romit pointed out - we may combine various positioning methods to achieve the ultimate goal. 

Another insight was whether we need that accuracy in the first place because the penalty of positioning error is much less indoor compared to the outdoor environment. What are the killer applications out there to push us to pursue that high accuracy? The closing remark is maybe we can shift the gear from in-door localization to in-door discovery. This space involves evacuation, social spaces or the landmark concept. 

6. About research. 

Victor's :

question the assumptions. (because technology evolves)

pursue your research to the point of irrefutability (don't publish paper that you know the limitations and you can overcome them)

publish only when you are done. 

persevere

don't be afraid of failing. 

..........and go change the world to the best. 

Suman's :

push your research as deep as possible so that people can use. 

don't expect someone else out there can industrialize your research. 

Henry's:

90% of the IR papers are overly complicated and not scalable. There is a gap between the mobile system community and data mining community to make the mobile sensing in big data happen. 

7. About SigMobile Community. 

I really admire all the seniors' effort to make SigMobile a place close to the industry. And to make research more impactive, young researchers not only need to do like Suman said, push the work to reality as deep as possible, but only need to think in a scale of community - what's the direction of this community, what need to be addressed as a community. Only this way can we get important tasks solved.

8. New problems. 

Mobile web performance(from Matt Welsh): intelligent web content preload? end-to-end? cross-layer?

Bottleneck for big data generated from mobile sensing : 90% energy consumption on wireless network(from Henry). He estimates the community can reduce the energy cost by 100 times at minimum. 

Fanglin