WEBVTT Kind: captions Language: en 1 00:00:04.330 --> 00:00:07.040 This is a teach pendant, and traditionally, 2 00:00:07.040 --> 00:00:09.730 it’s how you teach a robot arm to do something. 3 00:00:09.730 --> 00:00:11.110 It’s awkward. 4 00:00:11.110 --> 00:00:13.969 Two of the problems with programming robots the traditional way 5 00:00:13.969 --> 00:00:17.600 are that you need to tell it exactly what you want it to do, 6 00:00:17.600 --> 00:00:18.720 point to point. 7 00:00:18.720 --> 00:00:22.460 The other is that the robot assumes that nothing changes. 8 00:00:22.460 --> 00:00:24.950 You just manually drive the tip of the robot 9 00:00:24.950 --> 00:00:27.220 to where you know you need it to go 10 00:00:27.220 --> 00:00:29.220 and then hit ‘record,’ and remember that point, 11 00:00:29.220 --> 00:00:31.980 then you drive it over to the other place and hit ‘record,’ 12 00:00:31.980 --> 00:00:33.860 and remember that point, and then, when you play, 13 00:00:33.860 --> 00:00:35.100 it goes from one point to the next. 14 00:00:35.100 --> 00:00:36.970 And it can do that over, and over, and over again. 15 00:00:36.970 --> 00:00:39.270 In the event that there is some error or something moves, 16 00:00:39.270 --> 00:00:41.110 the robot has no way of knowing that. 17 00:00:41.110 --> 00:00:44.450 The robot will still go exactly where you told it to go the first time, 18 00:00:44.450 --> 00:00:46.830 and that can result in a crash, or breaking something, 19 00:00:46.830 --> 00:00:49.020 or injuring a person. It’s just like any computer. 20 00:00:49.020 --> 00:00:50.989 It will do exactly what you ask it to do, 21 00:00:50.989 --> 00:00:52.900 even if that's not what you meant! 22 00:00:52.900 --> 00:00:54.440 In the early days of computers, 23 00:00:54.440 --> 00:00:57.970 writing code generally meant using what’s called 'assembly language'. 24 00:00:57.970 --> 00:00:59.860 You would have to literally tell the processor 25 00:00:59.860 --> 00:01:02.510 to move things from one memory address to another, 26 00:01:02.510 --> 00:01:03.510 or to tweak some values, 27 00:01:03.510 --> 00:01:06.290 or to change specific pixels on the screen. 28 00:01:06.290 --> 00:01:08.100 Like giving instructions to this arm, 29 00:01:08.100 --> 00:01:12.370 you would be telling the chip every single thing that it had to do. 30 00:01:12.370 --> 00:01:16.150 As time moved on, programmers got higher level languages. 31 00:01:16.150 --> 00:01:18.050 Write some words and some brackets, 32 00:01:18.050 --> 00:01:21.560 the system works out the boring details for you. 33 00:01:21.560 --> 00:01:25.030 If you’ve ever worked with a spreadsheet, then that counts. 34 00:01:25.030 --> 00:01:28.170 Writing a formula in Excel, that's enough. 35 00:01:28.170 --> 00:01:31.530 In the end, sure, it all gets converted to assembly language. 36 00:01:31.530 --> 00:01:34.840 For most programmers, they never even have to think about it. 37 00:01:34.840 --> 00:01:39.050 The team here at Autodesk's Pier 9 in San Francisco are 38 00:01:39.050 --> 00:01:41.600 working on more or less the same thing, 39 00:01:41.600 --> 00:01:45.340 just for making physical things. 40 00:01:45.340 --> 00:01:48.570 Basically, the biggest robot that we have in the lab, 41 00:01:48.570 --> 00:01:51.560 which we call ‘Ash,’ is essentially a big, 42 00:01:51.560 --> 00:01:54.850 robotic 3D printer that's printing in stainless steel. 43 00:01:54.850 --> 00:02:00.110 We have a MIG welder that is depositing stainless steel onto a metal plate. 44 00:02:00.110 --> 00:02:04.950 By activating the MIG welder while moving the robot, we’re building a weld bead, 45 00:02:04.950 --> 00:02:07.260 and then we’re building beads on top of each other. 46 00:02:07.260 --> 00:02:10.310 You would typically use welding to stitch two pieces of metal together. 47 00:02:10.310 --> 00:02:13.230 What we’re doing is using the same technology, 48 00:02:13.230 --> 00:02:15.470 but stacking the metal on top 49 00:02:15.470 --> 00:02:18.350 in order to produce a separate piece of finished material. 50 00:02:18.350 --> 00:02:20.630 Now, ultimately, the result is the same. 51 00:02:20.630 --> 00:02:23.510 The motors in the robot arm are told, “move this way.” 52 00:02:23.510 --> 00:02:26.930 It’s just that human’s original instructions are a bit more abstract, 53 00:02:26.930 --> 00:02:29.470 and they're filtered through another couple of layers. 54 00:02:29.470 --> 00:02:34.200 Using a teach pendant would be pretty impractical for these complex curves. 55 00:02:34.200 --> 00:02:37.450 Normal 3D printers do basically 2½D. 56 00:02:37.450 --> 00:02:40.000 They go to an X and a Y, and then up and down. 57 00:02:40.000 --> 00:02:42.220 This robot can point in various directions. 58 00:02:42.220 --> 00:02:44.450 The robot needs to know not only where it is, 59 00:02:44.450 --> 00:02:47.300 but how it should point when it gets there. 60 00:02:47.300 --> 00:02:48.530 We give it a piece of geometry, 61 00:02:48.530 --> 00:02:51.930 and the software figures out the instructions set for the robot 62 00:02:51.930 --> 00:02:54.780 that will result in a print, that is what we want. 63 00:02:54.780 --> 00:02:58.520 One of the things we’re developing is a closed loop feedback system, 64 00:02:58.520 --> 00:03:01.120 where the robot is actually aware of what it’s doing. 65 00:03:01.120 --> 00:03:03.010 Before it completely fails a print, 66 00:03:03.010 --> 00:03:06.460 it’s keeping track of the quality of the print that it’s doing. 67 00:03:06.460 --> 00:03:10.090 It will actually correct, reprogram itself in real time, 68 00:03:10.090 --> 00:03:12.250 in order to avoid an actual failure. 69 00:03:12.250 --> 00:03:14.750 What we’re working on is a vision system, 70 00:03:14.750 --> 00:03:16.570 where between vision and a couple of other sensors, 71 00:03:16.570 --> 00:03:19.780 we can monitor and supervise the status of the prints. 72 00:03:19.780 --> 00:03:23.050 If the welder runs out of wire, or if something else happens, 73 00:03:23.050 --> 00:03:25.030 the robot would traditionally have no way of knowing. 74 00:03:25.030 --> 00:03:27.020 Now, the robot can know that. 75 00:03:27.020 --> 00:03:31.280 Not all the robot’s movements are being directly controlled by a person. 76 00:03:31.280 --> 00:03:33.650 If it goes wrong, it’s a bit different 77 00:03:33.650 --> 00:03:35.970 than just having an error message pop up. 78 00:03:35.970 --> 00:03:37.800 This arm here weighs two tonnes, 79 00:03:37.800 --> 00:03:40.620 and when it wants to, it moves fast. 80 00:03:40.620 --> 00:03:42.280 The only reason I'm allowed this close is because 81 00:03:42.280 --> 00:03:46.080 I'm literally holding the emergency stop button in my hand, 82 00:03:46.080 --> 00:03:47.319 just in case. 83 00:03:47.319 --> 00:03:51.300 Our robot currently has no way of sensing us. What currently happens is 84 00:03:51.300 --> 00:03:54.390 when a person that shouldn’t be near a robot gets close to the robot, 85 00:03:54.390 --> 00:03:55.970 you shut everything down. 86 00:03:55.970 --> 00:03:58.209 It would be great, and we’re interested in a future 87 00:03:58.209 --> 00:04:01.019 where the robot can know that person is there, 88 00:04:01.019 --> 00:04:03.390 with vision or some other kinds of sensing, 89 00:04:03.390 --> 00:04:05.580 and then actively avoid that person, 90 00:04:05.580 --> 00:04:08.400 and continue doing what it’s doing. 91 00:04:08.400 --> 00:04:10.959 Self driving cars, hospital heart monitors, 92 00:04:10.959 --> 00:04:14.050 basically everything electronic: ultimately, 93 00:04:14.050 --> 00:04:16.340 the code in it is just 1s and 0s. 94 00:04:16.340 --> 00:04:19.959 The more levels of abstraction between the programmer and the bare metal, 95 00:04:19.959 --> 00:04:21.419 the easier it is to write code, 96 00:04:21.419 --> 00:04:23.610 but when something goes wrong, 97 00:04:23.610 --> 00:04:26.330 fixing it might be out of your hands. 98 00:04:26.330 --> 00:04:28.319 Thank you very much to all the team at Autodesk, 99 00:04:28.319 --> 00:04:29.990 and to their applied research lab, 100 00:04:29.990 --> 00:04:31.360 here at Pier 9 in San Francisco. 101 00:04:31.360 --> 00:04:33.300 Go and check out their YouTube channel, 102 00:04:33.300 --> 00:04:36.400 or pull down the description for some links to see the amazing projects they're working on.