radar

Bottom Line Up Front Stefano P. Coraluppi suggests we form tracks first, then try to merge those, and that we don’t trust those initial tracks as much as we currently do in MHT. Summary Multi-target tracking (MTT) Sequence of sets of measurements, unknown time-varying number of targets. Overlay of false alarms. Must determine a _set_of trajectories. An intractable posterior probability distribution, both computationally and conceptually. MAP estimation, even if you nailed it, would be useless....

Sampada Deglurkar, Michael Lim, Johnathan Tucker, Zachary Sunberg, Aleksandra Faust, Claire Tomlin. Bottom Line Up Front I rate this 8.1 More, smaller, models trained to their own goals end up being more efficient and robust. This is… close to what I was attempting, but I think my “produce an image” thing was insane. Next step: go find this thing and see how it is proposing particles. I think I may be able to adapt this so that I’m keeping a bunch of unrelated particles alive and just deciding which ones to kill....

Folker Hoffmann, Matthew Ritchie, Francesco Fioranelli, Alexander Charlish, Hugh Griffiths Bottom Line Up Front I rate this 7.9. Super cool experiment, I’m frankly shocked that it works as well as it did. I wish I had radars to play with, and results from slightly less cooperative UAVs. Clearly lots of opportunities for ’next steps’ with this exact hardware setup. I want one. Figure Anarchy I know LaTeX and maybe even IEEE insist on letting figures roam the countryside however they please, but good lord, this paper is just a description of figures and none of them are anywhere near the text describing them....

Alexander Charlish, Folker Hoffmann Bottom Line Up Front This is a chapter from a larger book that I should read before discussing it with a student working under Alexander Charlish. I rate this 8.8 Easily the coolest thing I’ve seen, I’m 100% in love with the formalization of a POMDP, but I disagree with certain design/implementation choices. Took too long to get there, this chapter should be two chapters, the first of which I’d just ignore and the second I’d re-read 12 times....

Piotr Samczyński, Karol Abratkiewicz, Marek Płotka, Tomasz P. Zieliński, Jacek Wszołek, Sławomir Hausman, Piotr Korbel, Adam Ksiȩżyk Bottom Line Up Front I rate this 9.4. I’d never read much about either Passive Radars or 5G, but this paper gave me enough context to feel like I knew what was going on while at the same time doing something cool. Incredibly written, I never felt either overwhelmed or condescended to. Super cool experiment....

Sevgi Zubeyde Gurbuz, Hugh D. Griffiths, Alex Charlish, Muralidhar Rangaswamy, Maria Sabrina Greco, Kristine Bell. Bottom Line Up Front I rate this 4.4. Too much ink spilled on context and definitions. For a Past, Present, and Future paper, it felt like it was about 70% past, about 20% present, and a few tweets anthropophormizing radars tossed in at the end. It takes too much reading between the lines to find “things a researcher could do in this area”....

ECM is Electronic Counter Measures W.D. Blair (cited in a lot of these benchmark things in Jack’s book), G. A. Watson, T. Kirubarajan, Y. Bar-Shalom 1996, published 1998. This is the “Benchmark 2” paper. Thoughts So these are all from the late 1990s, but there’s a bizarre glimmer of hope that I have that I’ll be able to run them myself. I think worst case if I want to steer my masters into “just make a benchmark for RRM”, seeing how other benchmarks were implemented/interfaced with would be beneficial....

6-chapter book by Peter Moo and Zhen Ding. Thoughts: Definitions aren’t consistent in this book, task is sometimes == look, other times not. Sometimes it’s a sensor, a radar, or a node. It’s clear it is always referencing some other work, but it is a lot for somebody coming in with no prior experience. I’m most excited about the waveform selection. In my head, an AI that is tasked with a relatively tight set of waveform selections that is rewarded for correctly tracking targets could be an elegant, if black box, way to just step over a lot of this....

Alexander Charlish, Karl Woodbridge, Hugh Griffiths Bottom Line Up Front Run a quick auction to adjust the sizes of tasks, then dump the resulting tasks into earliest deadline first. It works better than randomly dropping tasks, it costs less computationally than Q-RAM. Summary multiplexed in time and angle time/power budget multidimensional parameter selection problem task revisit interval task dwell duration Part 2 of CDAPS (Continuous Double Action Param Selection) paper has good problem formulation....

Chris N. Potts, Luk N. Van Wassenhove Bottom Line Up Front I rate this 3.2. A period piece only readable in 1985, it did lead me to “Single Machine Weighted Earliness-Tardiness Scheduling Problem” though. I should have considered branching, trying to one-shot it was foolish. Next stop: brushing up on mixed integer linear programming/constraint programming. Make something slow. If too slow, tabu search. Summary Total Weighted Tardiness Problem n jobs, numbered 1 to n, must be processed without interruption on a single machine that can only handle one job at a time....