ECE 6962, Applications of Fading Channels, Fall 2012 Homework 2

Be sure to download the version 1.1 of my Python and Matlab listen and plotting codes. I have a zip file of all of the "v1" code here. The comments at the start of each file describe the modifications I've made to make the programs more robust to missed packets. The data formats have not changed, just how it reacts to a missed packet.

  1. Implement one of the line-crossing detection algorithms presented in [Youssef 2007] or [Kaltiokallio 2011]. Whatever your detector is, it essentially calculates a value (let's call it the "score") that is a function of your current and past measurements, and when that score goes above a threshold, it raises an alarm. Use your sensors to collect a data set in which you cross the link at a known time. A real-time implementation is helpful here because you can plot the score over time and watch the plot as you are walking through or not walking through the link line, and judge how accurate your detector is. In any case, turn in (1) your algorithm code and (2) one figure (using subplot, perhaps) that plots (a) your measured RSS on the four channels and (b) the score of your algorithm.
  2. Download my experimental data files HW2_2a.txt and HW2_2b.txt. These are what I obtained from running
        sudo ./ | tee HW2_2a.txt

    (or "... | tee HW2_2b.txt" for part b). Incidentally, my node numbers are 5 and 6, and my channel numbers are 14,18,22, and 26. The linksToPlot is set to:

        linksToPlot = [ (6,5,14), (6,5,18), (6,5,22), (6,5,26)]

    Your job is to run your code and tell me what times I crossed the line between sensors. You need to tell me how many crossings, and at what time each crossing was. Slot number is the best way to describe time here -- each RSS measurement has a slot number associated with it. The slot numbers on each line are very close. Pick any slot time on the line in the text file at which you think that I was right in the middle of the line between the two sensors. Or, if you want, the average of the slot times on that line. I am telling you that I was NOT crossing the line during the first 1400 slot numbers in either experiment (so for example, in Experiment a, any slot number between 17036 and 18436 indicates a measurement taken with NO person on the link line).

    • Experiment a. For part (a), the two sensors were 4.0 feet apart in an empty room, with nothing in between the two sensors.
    • Experiment b. For part (b), the two sensors were 8.3 feet apart, but they were in two different empty rooms, with a wall in between the two sensors. The wall is about 4 inches thick. One sensor is 6 feet from the wall, and the other sensor is 2 feet from the other side of the wall.

    Clearly, the through-wall experiment is going to have links that are non-line-of-sight and thus more likely to be in a deep fade. may use whatever algorithm you choose - your answer for question 1, or something you make up. You may use two different algorithms for experiments a and b. Describe your algorithm(s) if different from your answer for question 1, and turn in your plot of the score value (what your detector uses to decide if the line is crossed), your threshold, and the times at which you believe the links were crossed. It isn't fair to just guess, or to answer that the links were crossed at some times at which your score is not above your threshold.

Homework solutions are now posted.