WEBVTT 00:00.210 --> 00:07.530 In this additional lesson on the section on ultrasonics, Senso, I'm going to show you how you can 00:07.530 --> 00:12.160 improve the behavior of your ultrasonic sensor with the software. 00:12.180 --> 00:17.230 OK, if you have an erratic behavior, maybe your sensor is working just fine. 00:17.250 --> 00:18.280 So that's not a problem. 00:18.600 --> 00:21.060 Maybe you're using the simulation in that case. 00:21.090 --> 00:25.680 Well, simulation is always working, but if you have the same problem as me, as you could see in the 00:25.680 --> 00:31.950 previous lesson, then it is good to understand maybe why you have this problem or get some ideas of 00:31.950 --> 00:35.070 why and then be able to fix this in the code. 00:35.280 --> 00:40.500 So this kind of issue is actually something you have to be prepared to have a lot when you work with 00:40.500 --> 00:40.900 hardware. 00:40.970 --> 00:46.990 OK, because real life and the real world always contains parts of unknown. 00:47.010 --> 00:50.410 And this is true for everything, especially hardware. 00:50.600 --> 00:55.080 OK, it's not because it's working fine in simulation that is going to work fine in real life. 00:55.530 --> 01:01.800 And it's also not because it's working fine in real life, once or twice that every time you do the 01:01.800 --> 01:04.550 same setup, it's going to work fine also. 01:05.010 --> 01:05.290 Right. 01:05.320 --> 01:07.740 That's one of the main issue with hardware. 01:07.750 --> 01:08.670 It's not really an issue. 01:08.670 --> 01:09.910 It's just how it works. 01:10.410 --> 01:14.480 So you have to be prepared to face that kind of issue in many of your project. 01:14.490 --> 01:17.380 But that's a normal part of any project. 01:17.700 --> 01:23.550 So if you have a problem with how do it, you can either fix the hardware and sometimes you can also 01:23.550 --> 01:25.950 try to create a fix with software. 01:26.130 --> 01:32.320 OK, so here I'm going to talk about both and not that it's also a little bit more advanced, OK? 01:32.340 --> 01:34.710 This is, again, not a big you. 01:34.920 --> 01:38.540 Listen, this is more of an intermediate listen. 01:39.060 --> 01:42.820 So what could go wrong with the hardware? 01:43.050 --> 01:48.590 First, I have shown you that maybe the connections on your breadboard and your arduino are not correct. 01:48.600 --> 01:55.640 So make sure you check that 100 percent, maybe change some wires, change some connectors, connect 01:55.650 --> 01:59.000 the sensor to a different place to see if you can solve the issue. 01:59.210 --> 02:04.920 OK, if it doesn't work well, maybe it's simply because you have the bad quality sensor. 02:05.190 --> 02:10.800 And actually, if you have a starter kit like the starter kit that I have that I'm using right now, 02:11.280 --> 02:18.060 if you have both the cheapest sensor you can find when the thing is that it depends a lot on the manufacturer 02:18.060 --> 02:19.320 of the hardware component. 02:19.370 --> 02:24.990 OK, if you buy the cheapest one, you can maybe expect that disordering of the components on the sensor 02:24.990 --> 02:31.440 is not well made, that maybe some parts that needs to be aligned are not completely aligned. 02:31.780 --> 02:38.790 OK, that's the firmware running on the sensor is maybe not a good quality one either and many different 02:38.790 --> 02:39.210 things. 02:39.210 --> 02:40.320 OK, that can go wrong. 02:40.410 --> 02:46.350 So basically when you buy cheap hardware, well, don't expect it to work 100 percent of the time. 02:46.540 --> 02:52.620 So if you have the chance to get another ultrasonic sensor, maybe it's better to try the second one 02:52.620 --> 02:53.300 if you have one. 02:53.640 --> 02:58.320 And in the future, whenever you want to buy hardware component, if it's very, very, very cheap, 02:58.320 --> 03:00.240 I advise you maybe to buy two or three. 03:00.610 --> 03:05.440 OK, so in case you have one that is not working correctly, you can try the other one. 03:05.610 --> 03:11.520 So if you think it is working well and your component is a bad quality, well, the only option you 03:11.520 --> 03:12.480 have is to change. 03:12.480 --> 03:17.720 The components are also maybe to try to make a fix from the software, which we are going to do here, 03:17.820 --> 03:22.440 and also a third cause that maybe is simply the environment. 03:22.440 --> 03:24.990 OK, where are you operating the sensor? 03:25.290 --> 03:30.950 Because DICENZO is actually sending ultrasonic waves in the environment. 03:31.200 --> 03:36.330 So for my specific example, I am actually in a winning recording here. 03:36.510 --> 03:42.720 I am in a small room with a lot of glass walls and a lot of stuff in the wall. 03:42.960 --> 03:49.050 So maybe the ultrasonic waves are going to bounce everywhere and come back to the ultrasonic sensor, 03:49.050 --> 03:52.000 which will read wrong data from time to time. 03:52.410 --> 03:59.700 OK, you may want to try in a bigger environment, for example, in the outside or just in a bigger 03:59.700 --> 04:02.120 room to try to see if you have the same problem. 04:02.340 --> 04:02.580 Right. 04:02.640 --> 04:06.600 So now you have a few of the reasons that maybe it's not working. 04:06.600 --> 04:10.470 If it's not working all right, if it's working correctly, you don't need to worry about it. 04:10.710 --> 04:13.770 But now what can you do if you are in the same situation as me? 04:13.770 --> 04:16.850 So you have correctly checked your how do I set up? 04:17.370 --> 04:22.080 Well, you just have one sensor so you can change it and you can change on the. 04:22.170 --> 04:28.740 And so I'm doing this on purpose so I can show you how to fix hardware with software. 04:29.100 --> 04:31.590 You can modify the software a little bit. 04:31.590 --> 04:36.690 And that's what we are going to do when we get the ultrasonic distance here. 04:36.960 --> 04:42.240 So we get the duration in Microcircuit of the Pulse and then we compute the distance. 04:42.510 --> 04:48.780 I'm going to apply a few filters here so we can regulate a little bit the values from the software. 04:48.780 --> 04:52.860 So this is not, of course, a one hundred percent solution to the problem. 04:53.040 --> 04:59.300 OK, but this is a nice fix that will allow us to continue and progress with this sensor and the circuit. 04:59.490 --> 04:59.700 So. 05:00.510 --> 05:08.820 What I'm going to do is I'm going to upload these programs activity 15, OK, where we put on a different 05:08.820 --> 05:11.300 energy from the distance we get. 05:11.550 --> 05:14.010 And also because here we print. 05:14.130 --> 05:18.060 OK, listen, make sure you have the same print along with distance. 05:18.960 --> 05:24.600 So I'm going to I can open the science monitor and you can see the different values we get. 05:24.660 --> 05:25.130 OK. 05:26.550 --> 05:33.330 So we have some values and you can see we have some weird values from time to time to get a better visualization 05:33.330 --> 05:33.540 here. 05:33.570 --> 05:36.390 I'm going to open the cell clutter, OK? 05:36.690 --> 05:38.280 And let's see, what do we get? 05:38.310 --> 05:44.700 So here you can see I am in 140, OK, this is the wall behind me. 05:45.040 --> 05:49.400 I'm going to put my hand and do some movement in front of the sensor. 05:49.710 --> 05:56.060 And you can see we OK, first thing we see is that we have some values that go way too high. 05:56.130 --> 06:03.030 OK, from time to time and the maximum value we could get here is 400 centimeters, OK, because this 06:03.030 --> 06:05.300 is the max range for the sensor. 06:06.390 --> 06:15.720 But you can see we get the value up to more than 3000 centimeters, which is well, which is definitely 06:15.720 --> 06:16.450 not correct. 06:16.890 --> 06:24.330 So the first thing maybe we can do is that anytime we get a value that is above four meters or 400 centimeters, 06:24.780 --> 06:27.420 well, we just don't take this value into account. 06:27.750 --> 06:34.860 So to do that, I'm going to create a new viable here that I'm going to call double previous. 06:36.800 --> 06:40.880 Distance and let's initialize it to 400. 06:43.340 --> 06:49.010 OK, and I'm going to go here, so I'm reading the duration and computing the distance, and I'm going 06:49.010 --> 06:52.700 to filter I'm going to add some filters here for the distance. 06:52.730 --> 06:53.570 Let's see if. 06:54.650 --> 06:55.850 Distance is greater. 06:56.950 --> 06:58.030 And 400. 07:00.640 --> 07:05.950 Then what I'm going to do is I'm not going to take this into account, I'm simply going to return to 07:06.610 --> 07:09.010 their previous distance. 07:09.010 --> 07:17.050 And if the distance is lower than 400, so if the distance is correct, I'm going to the previous distance 07:17.050 --> 07:20.710 is equal to the distance and then return the distance. 07:21.130 --> 07:23.530 So what this will do, let's let's make an example. 07:23.560 --> 07:25.450 So first, we have a pretty decent 400. 07:25.450 --> 07:27.570 We just initialized with a venue. 07:27.910 --> 07:29.880 So let's say we have a correct value here. 07:30.220 --> 07:31.120 So we get this done. 07:31.180 --> 07:32.620 Let's say one hundred forty. 07:33.460 --> 07:35.960 The distance is lower than four hundred. 07:36.190 --> 07:41.620 So now we said the previous distance, which become the distance to the previous distance, becomes 07:41.620 --> 07:43.120 one hundred and forty. 07:43.330 --> 07:46.030 OK, then we return one hundred and forty. 07:46.750 --> 07:53.030 The next thing we read, let's say we have a with the disease like three thousand so we get three thousand. 07:53.050 --> 07:54.450 Here we check that. 07:54.460 --> 07:57.020 OK, three thousand is actually greater than 400. 07:57.250 --> 08:04.060 So instead of using this value, we just return the previous one, the previous one, which is now one 08:04.060 --> 08:04.950 hundred forty. 08:05.320 --> 08:07.270 So let's upload this program here. 08:09.050 --> 08:18.950 And let's open the CIA again and you will see, OK, I get 140, you see, I still get many variations 08:18.950 --> 08:27.770 in the volume, OK, that I never get any value that is above one hundred forty or any value that is 08:27.770 --> 08:34.340 above actually 400, OK, because 400 is the maximum the cap for the sensor. 08:34.890 --> 08:41.390 So I can be sure by using this filter that I don't get any value above four hundred. 08:41.390 --> 08:43.850 So that's already a good first step. 08:44.240 --> 08:48.370 So first, eliminate any value that is outside of the range of the sensor. 08:48.770 --> 08:53.560 And no, I'm going to do another thing and this is called a complementary filter. 08:53.990 --> 08:55.820 What is a complementary filter? 08:56.240 --> 08:57.650 So a complementary filter. 08:57.650 --> 09:03.140 I'm not going to go into too many details here if you want to check that outside of the code. 09:03.170 --> 09:06.890 OK, this is a mathematical filter that we are going to use. 09:06.890 --> 09:12.950 And this is very often used to smooth out some data that you read from the sensor. 09:13.340 --> 09:20.860 So the complementary filter is actually quite simple, is instead of just returning the new distance, 09:20.870 --> 09:26.750 OK, we are going to return an average of the previous distance and a new distance. 09:27.050 --> 09:32.750 OK, let's let's actually write the code and then give some explanation to the new distance. 09:32.750 --> 09:39.620 We become previous distance, let's say, multiplied by zero point five. 09:40.160 --> 09:40.700 Let's. 09:41.760 --> 09:48.160 Distance multiplied by zero point five, but we can't the distance here. 09:48.270 --> 09:53.550 OK, we have the previous distance from the previous run of that function and then we do. 09:54.150 --> 09:57.250 So this is an example with zero point five and zero point five. 09:57.690 --> 10:01.290 So the distance becomes so the distance we have just measured. 10:01.740 --> 10:04.780 We take half of this value and half of the previous one. 10:05.340 --> 10:09.480 So to make a complementary filter, you simply have to store the previous value. 10:09.510 --> 10:14.390 So you have to, of course, the data at a certain frequency. 10:14.400 --> 10:21.470 So he attempts, OK, and so you get you still the previous distance to the previous data. 10:21.720 --> 10:26.430 And then when you get the new data, you make a mix of the previous data and the new data. 10:27.060 --> 10:33.630 So, for example, if you have just one spike, which is very high, this spike would actually be smoothed 10:34.110 --> 10:37.130 and you will not see so much weird behavior. 10:37.740 --> 10:42.610 So we can use the distance with zero point five and distance with zero point five. 10:42.960 --> 10:47.350 The actual important thing is that this and this equals to one. 10:47.400 --> 10:54.810 OK, so we could use, for example, the zero point seven and zero point three here and the bigger the 10:54.810 --> 10:56.910 number for the previous data. 10:57.310 --> 10:59.930 OK, the bigger the resistance to change. 11:00.240 --> 11:02.530 And actually, let's mix some examples. 11:02.550 --> 11:06.130 OK, I'm going to go with zero point five and zero point five. 11:06.390 --> 11:09.660 Let's upload that program so you can see just one additional line. 11:10.810 --> 11:18.750 And I'm going to happen to say a lot, and you can see so we still have some weird variations, OK, 11:18.940 --> 11:20.710 but not not that much. 11:21.150 --> 11:24.700 It's actually much better if you can see that the. 11:26.120 --> 11:28.280 Signal is a little bit smoother. 11:29.680 --> 11:37.750 All right, and now let's add, let's say here on nine and point one, which means that whenever we 11:37.750 --> 11:44.740 get to a new distance, we are going to keep 90 percent of the previous distance and only add 10 percent 11:44.740 --> 11:45.670 of the new distance. 11:45.980 --> 11:47.410 So he that's applaud. 11:48.590 --> 11:55.920 You will see that he's very smooth, but also in shape, put my hand here. 11:57.020 --> 12:02.050 It's very smooth, but also very low key to detect changes. 12:02.060 --> 12:03.160 I remove my hand. 12:03.710 --> 12:08.780 You can see it really takes some time to go through here to 140. 12:09.140 --> 12:18.520 I just put my hand again now and you can see it's really low, really slow, but it's much smoother. 12:19.250 --> 12:22.190 So now you have the formula for the complementary filter. 12:22.190 --> 12:26.060 And what you need to tinker is basically those two diameters. 12:26.270 --> 12:32.720 So the percentage, 90 percent, 10 percent here, make sure that the total is one hundred percent on 12:32.720 --> 12:33.620 just one. 12:34.040 --> 12:42.200 OK, so the bigger the coefficient for the data, the smoother the values you are going to read. 12:42.740 --> 12:43.490 But also this. 12:43.580 --> 12:46.340 The data will be updated a little bit late. 12:46.480 --> 12:52.640 OK, so if you need to be very reactive to what's happening in the environment, you don't want to just 12:52.640 --> 12:54.890 get the one of the three segments, OK? 12:54.890 --> 12:57.620 You want to get the warning as fast as you can. 12:57.650 --> 13:05.810 So you need to find a tradeoff between how smooth you want the data to be and how reactive you want 13:05.810 --> 13:09.030 to get the new available data. 13:09.380 --> 13:12.020 So you have two ways actually doing that. 13:12.020 --> 13:15.390 You can change the equation here and you can also increase the frequency. 13:15.410 --> 13:18.890 So if you read faster, this is going to be updated faster. 13:19.190 --> 13:22.570 OK, and instead of 100, I'm going to put 60. 13:22.850 --> 13:29.420 60 is the minimum value recommended by the manufacturer for this component. 13:29.780 --> 13:32.060 I'm going to go to 60 and not lower. 13:32.100 --> 13:36.050 OK, but with this, we are going to read the data first. 13:36.260 --> 13:42.490 And I'm going to use, let's see, 70 here and 30 percent. 13:43.280 --> 13:48.200 So every 60 milliseconds, I'm going to read the new data from the sensor, compute the distance and 13:48.200 --> 13:49.680 check that the distance is correct. 13:49.790 --> 13:54.040 If not, I wrote on the previous one and then apply a complementary filter. 13:54.480 --> 13:58.850 OK, 70 percent of the previous one, 30 percent of the new one. 13:59.750 --> 14:04.450 So let's applaud that and let's see what we get here. 14:05.420 --> 14:12.470 Let's open the CIA plotter and I'm going to put my hands in front. 14:13.350 --> 14:14.760 Of the SENSEOR. 14:16.170 --> 14:25.500 OK, and you can see now, OK, the data we read is much smoother and Andy here is actually going to 14:25.890 --> 14:29.420 be yellow now and I remove my hand and it's going to green. 14:29.760 --> 14:35.180 So it's already quite fast for us to measure the correct distance. 14:35.190 --> 14:36.930 I put my hand again yellow. 14:37.170 --> 14:38.980 I go closer and red. 14:39.270 --> 14:43.590 OK, so we have some accurate data, which is quite smooth. 14:43.700 --> 14:48.860 You can see on the cell plotter and it's also quite reactive. 14:49.560 --> 14:49.880 Great. 14:49.890 --> 14:56.400 So you have seen in this lesson how to actually fix the problem you have with the hardware by actually 14:56.400 --> 15:02.550 applying a few lines of software to filter the data you receive from the sensor. 15:02.900 --> 15:05.520 So this is, of course, a little bit more advanced. 15:05.940 --> 15:11.100 It's OK if you don't understand everything now, you might have to do some more research on the complementary 15:11.100 --> 15:16.080 filter and will know that we have our sensor working correctly. 15:16.770 --> 15:22.020 We can go to the next section and actually we will come back to DICENZO and we are going to use it a 15:22.020 --> 15:22.420 lot. 15:22.950 --> 15:25.340 Well, the final project of this class.