1
00:00:00,000 --> 00:00:04,480
It's late at night and your city gets an empty intersection waiting for the red light to turn green.

2
00:00:04,480 --> 00:00:08,240
You begin to debate the moral ramifications of just driving through the red light,

3
00:00:08,240 --> 00:00:11,920
breaking the societal norms that separate human from beast when the thought occurs to you.

4
00:00:12,560 --> 00:00:15,040
Why is this light even holding you up in the first place?

5
00:00:15,840 --> 00:00:22,400
The answer lies in the large number of ways that traffic lights decide whether to turn red or green.

6
00:00:22,400 --> 00:00:28,960
In lightly traveled areas, some of them are on simple timers, a technology that dates back to 1922,

7
00:00:28,960 --> 00:00:34,480
especially during the wee hours when very few people are on the road, kind of like in 1922,

8
00:00:34,480 --> 00:00:38,480
which explains why you're sometimes stuck at a red for no apparent reason.

9
00:00:38,480 --> 00:00:42,080
But fortunately, most lights are far smarter than that.

10
00:00:42,080 --> 00:00:46,800
Probably the most common way that lights try to react to traffic conditions in some sensible way

11
00:00:46,800 --> 00:00:51,840
is through the use of an induction loop. You've probably seen these when you pull up to an intersection

12
00:00:51,840 --> 00:00:55,760
and you can see what looks like an outline in the pavement near the stop line.

13
00:00:55,840 --> 00:01:00,240
The loop is really nothing more than a piece of wire with a current running through it.

14
00:01:00,240 --> 00:01:05,600
When a car drives over it, the metal in the car's body decreases the impedance of the wire.

15
00:01:05,600 --> 00:01:10,160
A sensor detects this and lets an electronic controller know that a car is waiting,

16
00:01:10,160 --> 00:01:16,800
meaning that the light will turn green before long. Unless, like me, you drive a motorcycle and you miss the inductor.

17
00:01:16,800 --> 00:01:20,800
And this doesn't do much when it comes to managing traffic over a larger area,

18
00:01:20,800 --> 00:01:24,000
where heavy traffic at multiple intersections and close proximity

19
00:01:24,000 --> 00:01:29,280
requires coordination with each other to ensure that the streets don't become a snarled mess.

20
00:01:29,280 --> 00:01:33,360
Sometimes you can see a simple example of this in busy downtown areas,

21
00:01:33,360 --> 00:01:36,960
where several lights in succession are timed to turn green at once,

22
00:01:36,960 --> 00:01:44,160
allowing cars to flow freely for several blocks, but then also timed to stop you at some point to disincentivize speeding.

23
00:01:44,720 --> 00:01:49,200
Now, unsurprisingly, large cities often have some kind of central computer network

24
00:01:49,200 --> 00:01:56,160
to monitor and coordinate traffic lights. One great example of this is Sydney, Australia, which uses a system called SCAPS

25
00:01:56,160 --> 00:01:59,600
to route over 100 million vehicles every day.

26
00:01:59,600 --> 00:02:02,720
Their system works by gathering real-time traffic information,

27
00:02:02,720 --> 00:02:06,560
which is fed into those controller boxes that you might have seen at street corners.

28
00:02:07,200 --> 00:02:10,960
Although much of the system relies on the aforementioned induction loops,

29
00:02:10,960 --> 00:02:14,560
the controller can measure the gaps between vehicles

30
00:02:14,560 --> 00:02:18,160
and use that information to determine when it's time to change the lights.

31
00:02:18,880 --> 00:02:27,280
But all that data gets uploaded to a regional server, which applies algorithms to that data to control large numbers of intersections over a wide area.

32
00:02:27,920 --> 00:02:31,360
In this way, the system can adapt to real-time conditions,

33
00:02:31,360 --> 00:02:35,280
improving traffic flow compared to a series of individually controlled lights.

34
00:02:36,000 --> 00:02:43,520
In Sydney, the government of New South Wales claims that SCAPS has resulted in a 28% overall reduction in travel time,

35
00:02:43,520 --> 00:02:48,720
a 25% reduction in stops, and a 12% reduction in fuel consumption.

36
00:02:49,280 --> 00:02:52,640
Not bad considering that the city is home to over 5 million people.

37
00:02:53,280 --> 00:02:59,440
SCAPS has been deployed in many other cities worldwide as well, but unsurprisingly, much of the future in smart cities

38
00:02:59,440 --> 00:03:05,440
may instead revolve around predictive models that use machine learning to anticipate problems.

39
00:03:05,440 --> 00:03:08,560
We've already seen one version of this installed in Pittsburgh,

40
00:03:08,560 --> 00:03:12,800
where a system called SurTrack allows intersections to communicate with each other

41
00:03:12,800 --> 00:03:17,200
to generate a signaling model based on current traffic conditions.

42
00:03:17,200 --> 00:03:21,520
And of course, work is being done on training traffic lights with machine learning.

43
00:03:21,520 --> 00:03:29,120
At Aston University in Birmingham, England, a recent study that used a neural network for reacting to simulated traffic via cameras

44
00:03:29,120 --> 00:03:34,320
showed that an AI model outperformed other current real-world methods of traffic management.

45
00:03:34,320 --> 00:03:39,440
And AI will likely become even more important once we have truly self-driving cars

46
00:03:39,440 --> 00:03:43,040
that will need to communicate with traffic signals to avoid clogs

47
00:03:43,040 --> 00:03:47,200
or, well, other undesirable outcomes at intersections.

48
00:03:47,200 --> 00:03:51,040
If you enjoyed this video, hit the like button if you didn't. Oh, hey, there's that other button too.

49
00:03:51,040 --> 00:03:54,800
You can check out our other videos if you want more, and don't forget to subscribe and leave a comment below

50
00:03:54,800 --> 00:04:00,080
if you have suggestions for future TechWiki videos not fast as possible.