0:00:13 | no |
---|
0:00:14 | to that i'm not going to talk about |
---|
0:00:16 | that a problem and that inverse problem |
---|
0:00:18 | uh i'm actually look at the much simpler problem |
---|
0:00:20 | but i have a perfect set |
---|
0:00:22 | i know exactly but am looking for |
---|
0:00:24 | i put the device in the water |
---|
0:00:26 | okay |
---|
0:00:27 | and then i'm looking for nothing |
---|
0:00:29 | basically i one to reconstruct |
---|
0:00:31 | the what |
---|
0:00:31 | okay |
---|
0:00:33 | um |
---|
0:00:34 | so if i you |
---|
0:00:36 | if i do it before sold december paul inverse problem but they're we'll get its this |
---|
0:00:41 | pretty shocking right |
---|
0:00:43 | it's like finding can walter |
---|
0:00:46 | um |
---|
0:00:47 | in the |
---|
0:00:49 | if you look |
---|
0:00:50 | more careful that if picture of what you see |
---|
0:00:53 | is that with a trace |
---|
0:00:54 | from green to blue |
---|
0:00:57 | we show a basically says that |
---|
0:00:59 | this a in the time of flight |
---|
0:01:01 | probably if we remove |
---|
0:01:03 | this are you know these delay of time of flight we can get |
---|
0:01:06 | to the correct picture do so |
---|
0:01:08 | and what to get |
---|
0:01:10 | is not quite |
---|
0:01:11 | that effect |
---|
0:01:12 | is a lot of uh fluctuations run sense |
---|
0:01:15 | and that this file suggest that the positions |
---|
0:01:18 | um |
---|
0:01:19 | or not correct we have to |
---|
0:01:21 | um estimate this position |
---|
0:01:23 | so or two is to go row uh about it we can send was back to the manufacturer |
---|
0:01:27 | ask them |
---|
0:01:28 | to put them correctly |
---|
0:01:30 | on the circle |
---|
0:01:31 | and um you know it a distance |
---|
0:01:34 | and that they would probably say well uh you know |
---|
0:01:37 | uh on the piece of paper ever these simple but guess what these a physical devices |
---|
0:01:41 | is the best it can |
---|
0:01:42 | so we are |
---|
0:01:43 | start but these do watch |
---|
0:01:44 | okay |
---|
0:01:44 | and then the goal is to find |
---|
0:01:46 | the positions |
---|
0:01:48 | all of these sense |
---|
0:01:50 | um |
---|
0:01:52 | uh so |
---|
0:01:53 | how can we do that |
---|
0:01:56 | um |
---|
0:01:56 | if you are you know |
---|
0:01:58 | homogeneous medium |
---|
0:02:00 | do the simple religion should be didn't time of points and the pay was the senses |
---|
0:02:04 | basically |
---|
0:02:05 | um |
---|
0:02:06 | through this a constant C zero |
---|
0:02:10 | so if you know the time of flight |
---|
0:02:12 | you know the there was sense |
---|
0:02:14 | and what do so than we talk about their what distances |
---|
0:02:17 | um because |
---|
0:02:19 | there is uh |
---|
0:02:20 | very nice there i mean out to a nine john you that says |
---|
0:02:23 | if you put and |
---|
0:02:25 | a a points |
---|
0:02:26 | on the surface |
---|
0:02:28 | and if you for this particular metrics |
---|
0:02:31 | distance squared metrics |
---|
0:02:32 | which is basically |
---|
0:02:34 | the pairwise distances raised to the square |
---|
0:02:36 | that's all only the rank of this matrix is going to be for |
---|
0:02:40 | independent of N |
---|
0:02:42 | it is very easy to actually prove this is always like to three long |
---|
0:02:46 | but um so we're not result how can we use it to is another celebrate celebrated |
---|
0:02:50 | algorithm called multi dimensional scaling |
---|
0:02:53 | which basically through this uh |
---|
0:02:55 | matrix L |
---|
0:02:56 | if you put it on the left and right hand side of this and scored matrix |
---|
0:03:00 | uh applied |
---|
0:03:01 | no single value decomposition can find the exact positions |
---|
0:03:06 | of the sense |
---|
0:03:09 | so uh |
---|
0:03:10 | when i say that you can find exact positions what i really mean |
---|
0:03:14 | is that a can find them all to |
---|
0:03:16 | rigid transformation |
---|
0:03:17 | basically |
---|
0:03:18 | if you only have a pairwise was distances |
---|
0:03:21 | there is no difference between your topology |
---|
0:03:23 | and the one that is reflected |
---|
0:03:26 | okay |
---|
0:03:27 | well the one that is translate there |
---|
0:03:30 | or |
---|
0:03:31 | the one that is rotated |
---|
0:03:33 | so um |
---|
0:03:36 | it seems that the problem is solved |
---|
0:03:38 | because |
---|
0:03:40 | um |
---|
0:03:40 | i know the time of flight |
---|
0:03:42 | i can't D do use the um |
---|
0:03:44 | the pair was is |
---|
0:03:46 | a can use M the S |
---|
0:03:48 | i find a position |
---|
0:03:49 | actually continue stock |
---|
0:03:52 | well apparently um i should |
---|
0:03:54 | and or a couple of challenges head of first |
---|
0:03:56 | which prevents also |
---|
0:03:57 | from using go |
---|
0:03:59 | this met |
---|
0:04:01 | okay |
---|
0:04:05 | first thing all you know |
---|
0:04:07 | the or in you know and |
---|
0:04:08 | oh joint to do signal processing and in signal processing not think he's less |
---|
0:04:13 | so well time of flight are actually not |
---|
0:04:16 | that the first |
---|
0:04:17 | and source of |
---|
0:04:18 | certainty |
---|
0:04:19 | that's not |
---|
0:04:20 | actually |
---|
0:04:21 | the big deal because um |
---|
0:04:24 | well and the F is actually a robust against a a not |
---|
0:04:27 | but the more interesting ones are come |
---|
0:04:31 | the second |
---|
0:04:32 | source of on L certain certainty actually structured missing in trees |
---|
0:04:37 | and what happens here is that |
---|
0:04:39 | when the transmitter here a signal |
---|
0:04:42 | the ones |
---|
0:04:43 | which are in its vicinity can not hear the signal |
---|
0:04:47 | this is one of the limitations of this what |
---|
0:04:50 | so |
---|
0:04:52 | all of these red lines are going to be |
---|
0:04:54 | you right |
---|
0:04:55 | we don't have this information |
---|
0:04:58 | if i put it in a mid trick for |
---|
0:05:01 | well you know what have this use metrics |
---|
0:05:03 | two hundred back to hundred so i can up put actually |
---|
0:05:06 | uh you know numbers here |
---|
0:05:07 | it to take all my |
---|
0:05:09 | so instead i put |
---|
0:05:10 | colours here |
---|
0:05:12 | is top you know these numbers |
---|
0:05:14 | so on the left hand side |
---|
0:05:16 | basically what you know what we should |
---|
0:05:20 | on the right hand thought however |
---|
0:05:22 | the um |
---|
0:05:23 | the central band and these two corners |
---|
0:05:26 | are going to be race and we put zero because we don't know |
---|
0:05:30 | the value |
---|
0:05:34 | another source of uncertainty is actually what we call called right and the missing singing tree |
---|
0:05:39 | what happens is that |
---|
0:05:40 | if you plot the time of flight |
---|
0:05:42 | with the respect to the sensor in texas |
---|
0:05:44 | what you should get |
---|
0:05:46 | um is this news mandy four |
---|
0:05:48 | but |
---|
0:05:49 | what you will get |
---|
0:05:51 | uh |
---|
0:05:52 | we'll have |
---|
0:05:53 | you know a couple of sparks which do not make any sense |
---|
0:05:56 | and these are basically down years |
---|
0:05:58 | you have to discard them |
---|
0:06:00 | so you put zero again here |
---|
0:06:02 | no um |
---|
0:06:04 | if i you list the game means that some almost |
---|
0:06:08 | um |
---|
0:06:08 | pair was these sense are going to be a randomly this on |
---|
0:06:12 | well we seen that are going to be a riesz random of like the model let B C |
---|
0:06:15 | okay |
---|
0:06:17 | now um |
---|
0:06:19 | so we all left be such a picture |
---|
0:06:21 | you no longer have all the pair was this |
---|
0:06:25 | and if that's not enough |
---|
0:06:29 | uh well basically you know |
---|
0:06:30 | well in a |
---|
0:06:32 | okay |
---|
0:06:33 | oh you have to wait |
---|
0:06:46 | um |
---|
0:06:47 | so in a matrix form |
---|
0:06:48 | what it means |
---|
0:06:49 | is that you know you will have a couple of dots |
---|
0:06:52 | you know randomly |
---|
0:06:53 | as yet or |
---|
0:06:55 | you're |
---|
0:06:56 | and if that's not enough |
---|
0:06:58 | well you have these on known like that in the beginning of the talk i uh you know |
---|
0:07:02 | oh i i mention |
---|
0:07:05 | so well the reason here is that um but these are electronic get |
---|
0:07:09 | right |
---|
0:07:10 | and uh |
---|
0:07:11 | oh when you fired the transmitter |
---|
0:07:13 | it's not going to transmit the signal immediately each weights uh for a couple of seconds |
---|
0:07:19 | mark second sir |
---|
0:07:21 | and that but we don't know this V so we have to estimate these as well |
---|
0:07:26 | okay |
---|
0:07:27 | um |
---|
0:07:29 | you know just to uh less first um |
---|
0:07:31 | sort with these um |
---|
0:07:33 | a missing entries forget about this that |
---|
0:07:36 | um shift |
---|
0:07:37 | and time of flight |
---|
0:07:39 | um |
---|
0:07:41 | um so we had |
---|
0:07:43 | these amazing result that this stance squared matrix was ranked for we could use M T S |
---|
0:07:48 | we can no longer use it because may of these that |
---|
0:07:51 | there was these sense are missing |
---|
0:07:53 | now |
---|
0:07:54 | uh the question is there were not we can estimate these missing in tricks |
---|
0:07:59 | well uh |
---|
0:08:01 | this is actually a topic of matrix completion that had recently you know there are a lot of risk uh |
---|
0:08:06 | that's been lot of activities in the past to years |
---|
0:08:09 | and the question is |
---|
0:08:10 | um you know pretty five this state here |
---|
0:08:12 | we have |
---|
0:08:13 | a rank K matrix of dimension and by and |
---|
0:08:16 | some of the injuries are random me |
---|
0:08:20 | and then it turns out that on the the road conditions you can actually |
---|
0:08:24 | find the missing it |
---|
0:08:27 | um the one uh so |
---|
0:08:29 | right now they are you know a lot of uh and is out |
---|
0:08:33 | when we started this work the a couple of them |
---|
0:08:36 | so um |
---|
0:08:37 | we actually |
---|
0:08:38 | use um |
---|
0:08:39 | of the space um um |
---|
0:08:41 | the develop point want to now already and his students at stand for |
---|
0:08:46 | and the way that uh |
---|
0:08:47 | um this algorithm works |
---|
0:08:49 | is basically by projecting |
---|
0:08:51 | the uh at the metrics on the space of rank |
---|
0:08:54 | Q mattresses |
---|
0:08:56 | and then uh doing some kind of great great in these |
---|
0:08:59 | now um |
---|
0:09:01 | do is a catch |
---|
0:09:03 | and them in all these out algorithms |
---|
0:09:05 | that i know |
---|
0:09:06 | um you have to use you |
---|
0:09:09 | that the in trees or |
---|
0:09:10 | you raise randomly |
---|
0:09:12 | okay |
---|
0:09:13 | um so probably do true before i guess |
---|
0:09:17 | um i |
---|
0:09:17 | you know probably you know from of for this problem |
---|
0:09:20 | but uh well to the best of my knowledge |
---|
0:09:22 | this was the case |
---|
0:09:24 | and now um so |
---|
0:09:26 | but as i mentioned we have this structured missing trees |
---|
0:09:29 | these are in trees that we know we will never get |
---|
0:09:32 | a any |
---|
0:09:33 | observations about |
---|
0:09:34 | right so |
---|
0:09:36 | um i to space is not going to work for |
---|
0:09:38 | as a T |
---|
0:09:40 | um |
---|
0:09:42 | so we have to redevelop develop again these all the space to make sure that |
---|
0:09:45 | a a when if we have a structure missing trees |
---|
0:09:48 | this is going to work |
---|
0:09:50 | so before like a you know all these error bounds for |
---|
0:09:54 | uh for the classical a to space is no use for a |
---|
0:09:58 | um um and um well the theory a a is actually quite simple and you can find it in you |
---|
0:10:02 | know paper |
---|
0:10:03 | um i'm not going to bore you with that you know details of the proof sets order sartre |
---|
0:10:08 | but uh let me just mention |
---|
0:10:10 | you know the model that use |
---|
0:10:12 | so we no longer assume that the sensors are are actually say sensor circle |
---|
0:10:15 | you seem that the R |
---|
0:10:16 | uh a on these and we |
---|
0:10:19 | we |
---|
0:10:20 | uh be a |
---|
0:10:22 | and the way that we are going to capture the structure missing trees are great are are |
---|
0:10:26 | but a bit through the use um |
---|
0:10:28 | uh a a and with that if |
---|
0:10:31 | there's the transmitter here |
---|
0:10:32 | all the sensor |
---|
0:10:34 | um in flight three kill or not going to your anything |
---|
0:10:38 | okay |
---|
0:10:39 | so |
---|
0:10:41 | if the sensors |
---|
0:10:43 | uh or |
---|
0:10:44 | you know distributed uniformly at random you sound was |
---|
0:10:48 | and if you see assume |
---|
0:10:50 | that um hmmm the time of lights are going to be a random with probability P |
---|
0:10:55 | fix number |
---|
0:10:57 | and for the structure |
---|
0:10:58 | uh missing trees if we assume that the are fine is going to scale |
---|
0:11:02 | like school root of log over and |
---|
0:11:05 | then our our or and reads as follows |
---|
0:11:07 | that |
---|
0:11:09 | the |
---|
0:11:10 | distance |
---|
0:11:11 | between the |
---|
0:11:12 | um |
---|
0:11:13 | this squared matrix and in its estimate is going to be bound but boy these two true |
---|
0:11:19 | um |
---|
0:11:21 | uh what we should mention is that |
---|
0:11:23 | um |
---|
0:11:25 | we we did a assume anything about the noise |
---|
0:11:27 | so the noise can be deterministic |
---|
0:11:29 | random |
---|
0:11:30 | um you know you name |
---|
0:11:32 | so these bound whole |
---|
0:11:34 | you need full generality |
---|
0:11:36 | um |
---|
0:11:38 | the all that thing that they should mention is that |
---|
0:11:41 | this term goes to zero as an goes to infinity all everyone we control controlled easter |
---|
0:11:46 | in many many cases |
---|
0:11:47 | it goes to zero |
---|
0:11:49 | but i'm pretty sure you can come up with example take doesn't |
---|
0:11:52 | for instance |
---|
0:11:53 | for go in noise |
---|
0:11:55 | uh uh that are |
---|
0:11:56 | now a prior we were not |
---|
0:11:58 | you know interested in a find a distance score metrics what you wanted to do side you know finding to |
---|
0:12:03 | positions |
---|
0:12:04 | but as i said we can find the positions of to transformation |
---|
0:12:07 | and |
---|
0:12:08 | we have to make sure that you know what and uh we we we basically one it |
---|
0:12:13 | uh |
---|
0:12:14 | uh |
---|
0:12:15 | one to define the distance between the estimate |
---|
0:12:19 | and the right one |
---|
0:12:20 | in a way that doesn't depend on the rigid transformation it should be in |
---|
0:12:24 | it turns out the right way to do it is basically |
---|
0:12:27 | these four |
---|
0:12:28 | um which is in barrie |
---|
0:12:31 | on the different formation and |
---|
0:12:33 | it's going to be zero these |
---|
0:12:34 | diff the ins distance |
---|
0:12:36 | even all if E X you equals X i |
---|
0:12:39 | now |
---|
0:12:40 | um if we apply |
---|
0:12:42 | and the S |
---|
0:12:44 | after |
---|
0:12:45 | oh the space |
---|
0:12:46 | uh we can actually bounded if then |
---|
0:12:50 | basically the same rate that B D before |
---|
0:12:52 | is going to be it's same expression |
---|
0:12:54 | okay |
---|
0:12:56 | no uh for the uh |
---|
0:12:58 | so we had another other source |
---|
0:13:00 | of uncertainty which was these to like these constant that to have to measure |
---|
0:13:04 | here we assume that is |
---|
0:13:05 | going to be |
---|
0:13:06 | uh for every want for every transmitter is going to be say |
---|
0:13:09 | okay |
---|
0:13:10 | now |
---|
0:13:11 | a there is going to be a need to about it and that's fine |
---|
0:13:14 | these um this T zero |
---|
0:13:16 | um |
---|
0:13:17 | but the for the sake of them are not be true the details of these out with M |
---|
0:13:21 | uh what they should mention is that is probably is nonconvex |
---|
0:13:25 | so um i it's very difficult to find |
---|
0:13:28 | uh you know to you actually out prove the convergence |
---|
0:13:31 | uh we have if you're if the with and it converges numerically or we don't have any pro |
---|
0:13:38 | uh and the idea is again to use this property of the these sense square metrics metric is rank for |
---|
0:13:42 | wanna make sure that you know what we're fine is actually going to be as close as possible to the |
---|
0:13:47 | right form |
---|
0:13:49 | oh okay |
---|
0:13:49 | so |
---|
0:13:50 | um |
---|
0:13:51 | unofficially we had access to real data what |
---|
0:13:54 | um |
---|
0:13:56 | oh i cannot report these these you know |
---|
0:13:58 | these data as here so what we did is just some simulations that maybe the characteristic of uh the real |
---|
0:14:03 | data |
---|
0:14:04 | and then uh |
---|
0:14:05 | um |
---|
0:14:06 | a diffuse basically you know well what you still before it uh |
---|
0:14:10 | you know the the a or is going to be in the twenties and D meter is the number of |
---|
0:14:14 | them or two hundred and then |
---|
0:14:16 | uh |
---|
0:14:16 | the deviation is going to be half from that are is does that was the D is going to be |
---|
0:14:20 | D the metrics |
---|
0:14:22 | to real matrix if this is going to what |
---|
0:14:23 | to to be able to have |
---|
0:14:25 | and the |
---|
0:14:27 | if the you |
---|
0:14:28 | um you know you actually do well our them |
---|
0:14:32 | uh a fee that there are going to be a lot of deviations |
---|
0:14:35 | uh |
---|
0:14:37 | um |
---|
0:14:38 | from the from the circle so |
---|
0:14:41 | if you got yeah that this is the prince function that it have |
---|
0:14:44 | that all of these sense of are going to be on the thing bill but if we want that are |
---|
0:14:47 | with them see that they are not going to be |
---|
0:14:50 | a you know they're got not be to be place exactly the same |
---|
0:14:53 | so |
---|
0:14:54 | this the last |
---|
0:14:57 | um |
---|
0:14:58 | if you the the picture that we started that of |
---|
0:15:02 | if we actually |
---|
0:15:03 | remove the uh |
---|
0:15:06 | do you lace |
---|
0:15:06 | you know these constant you'll lay set we have to find out |
---|
0:15:09 | we'll get if speech or are you it before |
---|
0:15:12 | if we complete the distance a a three |
---|
0:15:15 | with a a a space |
---|
0:15:17 | you'll get is picture |
---|
0:15:19 | it's not very different from the previous picture |
---|
0:15:22 | a but if you find the positions and then you know a sold the inverse problem |
---|
0:15:28 | you get back into one |
---|
0:15:30 | the |
---|
0:15:30 | it's really important to calibrate the system is really important to find |
---|
0:15:35 | position |
---|
0:15:37 | and the |
---|
0:15:38 | even if all you know beforehand the the um |
---|
0:15:41 | the range was from |
---|
0:15:42 | uh one thousand uh |
---|
0:15:44 | four hundred to one thousand six hundred before the close |
---|
0:15:47 | from |
---|
0:15:48 | you know these value this one |
---|
0:15:50 | then you don't see any deviation a |
---|
0:15:52 | so it eight towards um |
---|
0:15:54 | you |
---|
0:15:56 | yeah thank you very much and i'll be have to answer questions if true or german |
---|
0:16:06 | thank you for this okay and |
---|
0:16:08 | we have time for one question |
---|
0:16:09 | please |
---|
0:16:15 | i |
---|
0:16:16 | um we just some on the might like best aging |
---|
0:16:19 | and and was like given a set up and you make get that yeah so you don and i one |
---|
0:16:24 | writing differentiation between you time at like measurement |
---|
0:16:28 | and and just an eight station at the sense that the "'cause" what we found like if a code is |
---|
0:16:32 | getting good trying to fight management |
---|
0:16:34 | it "'cause" i things like multiple and more fundamental fashion |
---|
0:16:37 | um so i actually the you're not have a seen the time of flights measurements so these there is this |
---|
0:16:42 | yeah |
---|
0:16:43 | these guys they have these estimators for the time of flight |
---|
0:16:46 | right |
---|
0:16:46 | and we seen that what were they you know what where we got from the is actually correct |
---|
0:16:51 | as a how likely as i think you know |
---|
0:16:53 | uh uh what is the and that kind of flight measurements |
---|
0:16:56 | i i get to get good ones because you got a dispersive medium |
---|
0:16:59 | um no i actually don't know the yours |
---|
0:17:01 | K can i think you it seems out |
---|
0:17:05 | okay so let's move to the the second work |
---|
0:17:08 | no the to not be in me |
---|