Lecture 2

(1) What is more privacy-preserving: network-based or terminal-based positioning (according to the taxonomy in Fig. 6-2)? Explain in 1 sentence.

(2) Is GPS the former or the latter?

(3) How does a GPS receiver distinguish between the signals received from different satellites?

Lecture 3

Answer the following questions in one short sentence each:

(1) Classify the Cricket location system according to the taxonomy of the “Positioning” reading (from last week’s Location-Based Services, Chapter 6) shown in Figure 6.2 on page 127. Is it terminal-based or network-based? Is it integrated or stand-alone?

Lecture 4

Provide a short answer for each of the following questions:

(1) In WiTrack, how does FMCW compute the time of flight?

(2) In WiTrack, what is the difference between static and dynamic multipath?

(3) List 2 advantages and 2 disadvantages for WiTrack over the RADAR paper

Lecture 5

No questions this week :)

Lecture 6

Suppose we want to send data from node S to node T in a multi-hop wireless network. The probability of successfully receiving an ACK for a packet sent from S to T directly is q. The other choice is to use an intermediate node, N; the probability of successfully receiving an ACK for a packet from S to N is p, the same as the probability of successfully receiving an ACK for a packet sent from N to T. Using ETX, under what conditions would the direct link from S to T be preferred to the path S — N — T?

Lecture 7

How does Caraoke count cars when multiple E-toll transponders respond at the same time?

Lecture 8

Describe two settings in which inertial sensing can be superior to/preferred over GPS for navigation applications.

Lecture 9

A simple method to detect potholes would be to look at time windows during which the absolute magnitude of the acceleration vector is above some threshold. This will lead to a number of false positives, as described in the paper. After reading the paper, list two sources of such false positives, and how the paper proposes to address them.

Lecture 10

(1) Why can't we build robust crash rate models by simply using the locations of prior crashes alone?

(2) What key inputs does the paper use to build crash rate maps?

Lecture 11

Please read the vital radio paper and answer the following questions:

(1) We learned that WiTrack's accuracy is about 10cm. Further, we learned that the resolution of FMCW is C/2B, where C is the speed of light and B is the bandwidth. VitalRadio uses the same FMCW radio that WiTrack uses. So how is it possible for VitalRadio to measure movements as small as the chest inhale-exhale motion?

(2) VitalRadio can measure sub-centimeter motion. Can one use that technique to improve trajectory resolution in WiTrack to localize a person to within sub-centimeter accuracy? Explain your answer.

Lecture 12

(1) Why does Mapster outperform prior approaches in Boston and Chicago, but perform comparably in Salt Lake City?

(2) How does the cGAN refine inferred data? Why do we need to refine it in the first place?

Lecture 13

(1) Alice and Ben have invited a few colleagues to an important private meeting. They are however worried that one of their colleagues may use his cell phone to record the conversation at the meeting. Alice claims that she and Ben can use the technique in the paper to prevent such recording. Is Alice right? Explain your answer.

(2) List two advantages for the technique in the paper in comparison with past work on near-ultrasound.

Lecture 14

In 2 sentences, explain why Glimpse is better than doing the entire processing in the cloud.

Lecture 15

(1) Why does PAB require much less power than existing underwater communication technologies?

(2) How does the design deal with collisions?

Lecture 16

(1) How does Farmbeats combine aerial imagery with sparse sensor data to generate a fine-grained sensor map?

(2) What are the two constraints that Farmbeats needs to satisfy for duty cycling its IoT operation?