1 00:00:00,503 --> 00:00:03,170 (graphic blips) 2 00:00:06,000 --> 00:00:08,280 - [Narrator] Artificial intelligence has shown positives 3 00:00:08,280 --> 00:00:10,230 for organizations in breaking down 4 00:00:10,230 --> 00:00:12,120 large sums of information. 5 00:00:12,120 --> 00:00:14,790 The Center for Transportation Research is investigating 6 00:00:14,790 --> 00:00:17,430 the practical advantages of integrating AI 7 00:00:17,430 --> 00:00:20,220 into operations from the Texas Department of Transportation, 8 00:00:20,220 --> 00:00:22,410 examining potential benefits in this new 9 00:00:22,410 --> 00:00:24,120 and vast data landscape. 10 00:00:24,120 --> 00:00:26,550 - So this was a four-year long project. 11 00:00:26,550 --> 00:00:28,950 So during the first two years, 12 00:00:28,950 --> 00:00:32,490 it was a broad exploration project, 13 00:00:32,490 --> 00:00:36,480 trying to help TxDOT understand what the landscape was 14 00:00:36,480 --> 00:00:39,570 in terms of artificial intelligence and transportation. 15 00:00:39,570 --> 00:00:42,300 So this was five years ago when there were a lot 16 00:00:42,300 --> 00:00:46,740 of emerging data sources, vendors based on machine learning, 17 00:00:46,740 --> 00:00:49,860 artificial intelligence, and we were trying to get 18 00:00:49,860 --> 00:00:50,790 the lay of the land. 19 00:00:50,790 --> 00:00:53,370 The second two years, based on what we found 20 00:00:53,370 --> 00:00:56,100 during the first two years, we realized 21 00:00:56,100 --> 00:00:58,230 that there were some low-hanging fruits for things 22 00:00:58,230 --> 00:01:02,850 that we could do with data that TxDOT was already acquiring, 23 00:01:02,850 --> 00:01:05,400 and that could derive more value from that data. 24 00:01:05,400 --> 00:01:08,220 - They found that there is a great potential 25 00:01:08,220 --> 00:01:13,220 for AI to help our transportation system operators 26 00:01:15,990 --> 00:01:19,620 and owners to address our goals. 27 00:01:19,620 --> 00:01:23,163 For example, in our cases include safety, 28 00:01:24,210 --> 00:01:29,210 alleviate the congestion and include the air quality. 29 00:01:30,090 --> 00:01:32,908 We got tons of data, and the AI technology definitely 30 00:01:32,908 --> 00:01:37,908 would help us to make more informed, data-driven decisions. 31 00:01:39,360 --> 00:01:42,930 - For the first period, the first two years, 32 00:01:42,930 --> 00:01:44,820 we did small prototypes. 33 00:01:44,820 --> 00:01:49,350 So we have some insights concerning the use 34 00:01:49,350 --> 00:01:53,460 of machine learning to identify safety hotspots, 35 00:01:53,460 --> 00:01:56,040 to understand travel patterns, 36 00:01:56,040 --> 00:02:00,330 to do signal timing plans in real time. 37 00:02:00,330 --> 00:02:02,400 And we actually prototype this 38 00:02:02,400 --> 00:02:05,010 short term travel time estimation. 39 00:02:05,010 --> 00:02:10,010 And so for the models that we used over the last two years, 40 00:02:10,830 --> 00:02:13,110 we found that actually machine learning 41 00:02:13,110 --> 00:02:14,940 can help us estimate travel times 42 00:02:14,940 --> 00:02:18,420 up to 40% more accurately 43 00:02:18,420 --> 00:02:19,980 than what you would normally do. 44 00:02:19,980 --> 00:02:21,690 So in these days, 45 00:02:21,690 --> 00:02:25,890 we typically use the current travel time as an estimate 46 00:02:25,890 --> 00:02:28,200 of the travel time that you're going to experience. 47 00:02:28,200 --> 00:02:31,650 And that could be up to 40% wrong in some corridors 48 00:02:31,650 --> 00:02:33,420 and at some time of the day, 49 00:02:33,420 --> 00:02:35,940 and you can do it better using machine learning. 50 00:02:35,940 --> 00:02:40,500 - Some of the models tested in the project show promise 51 00:02:40,500 --> 00:02:45,168 in improving TxDOT's ability to help drivers identify 52 00:02:45,168 --> 00:02:50,046 the shortest route among competing alternatives. 53 00:02:50,046 --> 00:02:53,520 They model, we tested 54 00:02:53,520 --> 00:02:56,100 and gave us some pretty reliable results. 55 00:02:56,100 --> 00:03:01,100 So that would help the drivers make more informed decisions. 56 00:03:01,380 --> 00:03:04,863 And I think that's benefits 57 00:03:04,863 --> 00:03:07,511 that this project will bring to us. 58 00:03:07,511 --> 00:03:09,960 - So if this moves to implementation, 59 00:03:09,960 --> 00:03:13,350 potentially we could be providing more accurate estimates 60 00:03:13,350 --> 00:03:16,140 of travel times in such a way 61 00:03:16,140 --> 00:03:19,680 that drivers may save a few minutes when crossing Austin. 62 00:03:19,680 --> 00:03:24,300 And at the same time, by switching to an alternative route, 63 00:03:24,300 --> 00:03:26,370 all old congestion may get relieved. 64 00:03:26,370 --> 00:03:29,283 So everyone may experience travel time savings. 65 00:03:30,669 --> 00:03:32,700 And also from a traffic management perspective, 66 00:03:32,700 --> 00:03:36,240 maybe having more accurate travel time information 67 00:03:36,240 --> 00:03:39,600 will allow decision makers to maybe, 68 00:03:39,600 --> 00:03:42,330 when they have the ability to adjust timing plans, 69 00:03:42,330 --> 00:03:44,820 if they foresee a long delay on I-35, 70 00:03:44,820 --> 00:03:47,700 that may divert more traffic to the fronts roads 71 00:03:47,700 --> 00:03:50,583 that could improve management in general. 72 00:03:51,610 --> 00:03:55,410 - We can utilize this big amount of data 73 00:03:55,410 --> 00:03:59,160 to help us make more informed decisions. 74 00:03:59,160 --> 00:04:03,930 That's the direction we're going. 75 00:04:03,930 --> 00:04:06,000 One of the important tools 76 00:04:06,000 --> 00:04:10,410 that will help us decipher was including the data, 77 00:04:10,410 --> 00:04:13,620 is the artificial intelligence models. 78 00:04:13,620 --> 00:04:18,620 So that way, we can come up with more reasonable, 79 00:04:18,900 --> 00:04:22,120 accurate estimate of travel times 80 00:04:23,371 --> 00:04:27,180 which would provide a better experience 81 00:04:27,180 --> 00:04:29,730 to our traveling public. 82 00:04:29,730 --> 00:04:30,930 - [Narrator] For more information 83 00:04:30,930 --> 00:04:33,330 and to find the publication for this project, 84 00:04:33,330 --> 00:04:35,670 please visit the TxDOT Research Library 85 00:04:35,670 --> 00:04:36,873 at the link shown below.