JacobSorber

- friends
1 link karma
12 comment karma
send messagemember for
what's this?

TROPHY CASE

papers

subscribe8 readers

Welcome to the public review site of papers appearing in MobiSys 2012. Here you can find the public review written by a PC member for each paper and the authors' rebuttal to it. You can comment on any of them or add your own opinion. Finally you can vote to "like" or "dislike" a paper by clicking the up or down arrows next to its title.

reddit is a source for what's new and popular online. vote on links that you like or dislike and help decide what's popular, or submit your own!

Satellites in Our Pockets: An Object Positioning System using Smartphones by PuneetJainin papers

[–]JacobSorber 0 points1 point ago

sorry, this has been archived and can no longer be voted on

Question: People are inherently lazy, and will want to take as few pictures as possible. Can you tell me when to stop taking pictures? Is this computationally tractable on the phone?

Answer: No, we can't do this all on the phone, at this point. Also, you need at least 4 photos. It depends how far the object is, and the angles of the photos. We were looking at objects that were 20-100m away.

Satellites in Our Pockets: An Object Positioning System using Smartphones by PuneetJainin papers

[–]JacobSorber 0 points1 point ago

sorry, this has been archived and can no longer be voted on

Question: What if the object is moving?

Answer: This won't work, not without significant changes to the algorithms.

Satellites in Our Pockets: An Object Positioning System using Smartphones by PuneetJainin papers

[–]JacobSorber 0 points1 point ago

sorry, this has been archived and can no longer be voted on

Question: How does this approach perform when there is occlusion? Does it work better for certain classes of objects? Just buildings? What if a car drives by?

Answer: This approach doesn't require every key point. We just need enough points. It comes to whether you can gather enough key points.

Satellites in Our Pockets: An Object Positioning System using Smartphones by PuneetJainin papers

[–]JacobSorber 0 points1 point ago

sorry, this has been archived and can no longer be voted on

Question: You use vision to correct the result from GPS. How do you know the vision results are more accurate than the GPS?

Answer: From my experience, our vision techniques either work very well, or not at all. The vision algorithms, when they work, seem to be very structurally accurate.

No Need to War-Drive: Unsupervised Indoor Localization by HeWangin papers

[–]JacobSorber 0 points1 point ago

sorry, this has been archived and can no longer be voted on

Question: Philosophical question. At this point, can we finally say that indoor localization is a solved problem? Can we say this is done?

Answer: I feel that we are at a stage where we have converged on a solution that is an 85% solution. There are still a lot of loose ends remaining. If there were some incentive, I think industry could work out the last 15% and this problem would be solved.

Victor: There is a disconnect between academic research and industry in a big way. I think this is going to be solved by industry without all this fancy stuff.

Let me go back about 20 years...[story about the difficulties of finding value propositions...I can't type fast enough].

We've tried a lot of these solutions, and most of these just don't work.
They use contrived examples, and they just don't work.

Romit: We're not setup to take it 100%. That's not the role of academic research. We're setup to take an idea to 85%, and let industry take it from there.

Roy: [Couldn't hear the beginning.] One of the problems is we don't have a good way of benchmarking one solution against another. One mechanism is to create a repository with test data. I think we should do this.

Romit: I agree. This would be a good thing for many different problems. Not just localization.

You are Facing the Mona Lisa: Spot Localization using PHY Layer Information by SouvikSenin papers

[–]JacobSorber 0 points1 point ago

sorry, this has been archived and can no longer be voted on

Question: The overhead of war-driving seems high. How do you deal with this?

Answer: You don't need to cover the entire space, meter by meter. You just need to war-drive about 35 spots. When not in a spot, you can use coarser-grained WiFi localization.

You are Facing the Mona Lisa: Spot Localization using PHY Layer Information by SouvikSenin papers

[–]JacobSorber 0 points1 point ago

sorry, this has been archived and can no longer be voted on

Question: Do you get the same results with a person holding the receiver?

Answer: Yes, there are more variations, more clusters. But you can just add those clusters into your set of clusters for that spot.

You are Facing the Mona Lisa: Spot Localization using PHY Layer Information by SouvikSenin papers

[–]JacobSorber 0 points1 point ago

sorry, this has been archived and can no longer be voted on

Question: When talking about centimeter accuracy, height comes into play. How much does height impact your results?

Answer: We haven't done much with height. We are not able to vary height when doing experiments with the roomba. We have done some war driving with a student moving around with the receiver on a box at a different height. Even at this different height we didn't notice significant differences in accuracy.

You are Facing the Mona Lisa: Spot Localization using PHY Layer Information by SouvikSenin papers

[–]JacobSorber 0 points1 point ago

sorry, this has been archived and can no longer be voted on

Question: Did you conduct measurements in a room full of people? Museums are often full of people.

Answer: Yes, we took measurements in a student area, and in an area where students gather to eat lunch. [showed some data] Most of the time you can see that it is very robust to people in different environments. There is more information in the paper.

FM-based Indoor Localization by YinChenin papers

[–]JacobSorber 0 points1 point ago

sorry, this has been archived and can no longer be voted on

Question: What about cities with dense antennas or fewer transmitters?

Answer: Our dataset is limited to only a few cities, so I can't generalize. Information about the number and location of FM transmitters is also publicly available for many cities.

FM-based Indoor Localization by YinChenin papers

[–]JacobSorber 0 points1 point ago

sorry, this has been archived and can no longer be voted on

Question: Why use the two combined? Why not just use FM? How do you use this practically? Size of antenna?

Answer: I don't recommend using FM alone. FM is good for differentiating two locations within the same building, but it is sometimes difficult to differentiate between two locations in different nearby buildings. Also, we used an antenna that was twice the size of the phone in our experiments. There are smaller patch antennas that are very small and can easily fit into a phone, but we didn't have these available to include in our experiments.

FM-based Indoor Localization by YinChenin papers

[–]JacobSorber 0 points1 point ago

sorry, this has been archived and can no longer be voted on

Question: Can you explain why FM variations are smaller than WiFi? Are there weather dependencies?

Answer: Most of our datasets were collected over three months, during the Summer. They include a lot of weather variation (rain, sun). We also took continuous FM measurements for certain fixed locations, and the variation seems to be small.

As for why the lower variations: I can't prove it, but my intuition is that the lower frequencies are less susceptible to things like multipath.