1
00:00:00,000 --> 00:00:03,293
(Calming Music Playing)

2
00:00:03,293 --> 00:00:06,420
(Calming Music Playing)

3
00:00:06,420 --> 00:00:09,300
(Calming Music Playing)

4
00:00:09,300 --> 00:00:10,530
- Thanks to technology,

5
00:00:10,530 --> 00:00:14,470
TxDOT has improved the
way it monitors roadways.

6
00:00:14,470 --> 00:00:18,270
This new and enhanced road
condition data is the basis

7
00:00:18,270 --> 00:00:21,120
for a major maintenance
project conducted by

8
00:00:21,120 --> 00:00:23,900
researchers at the University of Texas.

9
00:00:23,900 --> 00:00:27,810
It's designed to help TxDOT be
more accurate in determining

10
00:00:27,810 --> 00:00:30,570
and predicting maintenance needs.

11
00:00:30,570 --> 00:00:34,320
- So they need to predict what's the

12
00:00:34,320 --> 00:00:37,370
estimated payment condition
in the next year or

13
00:00:37,370 --> 00:00:38,780
two or three or four.

14
00:00:38,780 --> 00:00:42,450
The key is to predict on an accurate

15
00:00:42,450 --> 00:00:47,120
manner and to minimize the
uncertainty of this predictions.

16
00:00:47,120 --> 00:00:51,710
- This project was designed
to take the payment data

17
00:00:51,710 --> 00:00:55,300
and make updates of the
payment prediction models.

18
00:00:55,300 --> 00:00:58,900
So that way, we have a
more accurate prediction of

19
00:00:58,900 --> 00:01:01,530
the future payment conditions. And so

20
00:01:01,530 --> 00:01:05,820
we can better allocate the
money, limited money to the

21
00:01:05,820 --> 00:01:09,400
right sections and the right time to get

22
00:01:09,400 --> 00:01:12,020
most cost effective results.

23
00:01:12,020 --> 00:01:15,410
- The data analysis techniques
have evolved enormously

24
00:01:15,410 --> 00:01:17,260
in the past 10, 20 years.

25
00:01:17,260 --> 00:01:21,690
So we were able to use
data-mining techniques,

26
00:01:21,690 --> 00:01:25,920
very advanced statistical models
and artificial intelligent

27
00:01:25,920 --> 00:01:29,060
tools as well to develop our models.

28
00:01:29,060 --> 00:01:33,870
And one of the advantage of
these new models that we develop

29
00:01:33,870 --> 00:01:36,540
is that in some cases for some payments,

30
00:01:36,540 --> 00:01:39,940
we managed to reduce this uncertainty.

31
00:01:39,940 --> 00:01:41,720
- Now we have a ARVR,

32
00:01:41,720 --> 00:01:44,180
what we call automated, uh,

33
00:01:44,180 --> 00:01:46,710
3D laser technology

34
00:01:46,710 --> 00:01:50,580
uh, equipped vehicles,
and drive down the road at

35
00:01:50,580 --> 00:01:53,820
highway speed and collect
the data continuously.

36
00:01:53,820 --> 00:01:57,670
- On the one hand, we
have better data quality.

37
00:01:57,670 --> 00:02:01,610
On the other hand today, as
compared with 10 years ago,

38
00:02:01,610 --> 00:02:03,970
there is many more

39
00:02:03,970 --> 00:02:05,870
advanced data analysis

40
00:02:05,870 --> 00:02:08,630
techniques that we could put our hands on.

41
00:02:08,630 --> 00:02:12,340
And that is why the models today

42
00:02:12,340 --> 00:02:16,330
can predict, eh, in a
much more accurate way.

43
00:02:16,330 --> 00:02:20,630
- The new model versus old
model, we see the improvement,

44
00:02:20,630 --> 00:02:23,580
uh, but you know, the system is a whole,

45
00:02:23,580 --> 00:02:26,650
and we need to consider not
just the prediction models.

46
00:02:26,650 --> 00:02:29,590
We also need to consider
another important piece,

47
00:02:29,590 --> 00:02:31,410
which is the decision trees.

48
00:02:31,410 --> 00:02:34,860
So that worked together
to make recommendations.

49
00:02:34,860 --> 00:02:38,670
- We, we do hope that this
project received some further

50
00:02:38,670 --> 00:02:42,170
consideration and it's
implemented because as I say,

51
00:02:42,170 --> 00:02:44,580
the research is done,
it has been paid off.

52
00:02:44,580 --> 00:02:48,230
It's finished. So now implementation

53
00:02:48,230 --> 00:02:51,230
could take a couple of
years and check out how

54
00:02:51,230 --> 00:02:54,840
these new models perform
relative to the old ones.

55
00:02:54,840 --> 00:02:58,220
But by looking at the data,
by looking at the models,

56
00:02:58,220 --> 00:03:00,810
you can see that the
predictions are more accurate

57
00:03:00,810 --> 00:03:03,730
and the potential for savings are there.

58
00:03:03,730 --> 00:03:05,410
- For more information, and to find the

59
00:03:05,410 --> 00:03:08,290
publications for this
project, please visit the

60
00:03:08,290 --> 00:03:11,783
TxDOT research library
at the link shown below.