What Does Traffic Look Like?
This is a look at traffic on LA’s 405 South on Thursday, May 17. The top of the image is the beginning of the day, the bottom is the end. Left to right moves from north to south. These readings come from a small subset of the thousands of traffic speed (and volume) sensors installed across California. This data is provided by RIITS, a data integrator/provider for a coalition of local transportation agencies in Southern California.
Here’s a closer look at what drivers faced that morning approaching the 101.
This next view shows how quickly the transition from free-flowing to jam occurs. At Victory Blvd it goes from 56 MPH at 5:50 AM to 11 MPH by 6:15 AM.
Looking at the afternoon, from the 105 through the South Bay, we see striations. These are the “stop and go” compression waves in the congestion.
This is the 60 West on a Sunday. That black band shows there’s an issue with that sensor near Wilcox Ave.
Zooming in, we also see a great example of two congestion jams and an “incident-induced” jam.
The incident-induced jam is solid black, no one is moving, and then spreads back along the road over the next 30 minutes. The congestion jams spread backwards similarly, but aren’t as severe. The two types clear very differently: once the incident is cleared, traffic clears from the incident back, but the congestion jams just shrink in size starting upstream and moving forward.
At TallyGo, we care about traffic because we predict traffic in order to avoid it. We’re a mapping and navigation company with some very interesting patented IP. We provide turn-by-turn directions just like Google, Waze, Apple, etc. The difference is the way we come up with the route. Most mainstream apps use “real time” traffic. We put “real time” in air quotes because they just take a snapshot of the current traffic conditions and then choose the fastest route based on the traffic that they see at that moment.
But the problem with that approach is that traffic changes over time, often dramatically. So any route they provide based on their “real time” data becomes outdated as soon as you start driving — roads that were once open can become clogged, and congestion in other areas can clear up. That’s why you often get notifications that they’ve found a “faster route” — it’s their “real time” system trying to catch up with rapidly changing traffic conditions.
Here’s the 101 South coming into downtown LA.
Let’s look at how a route based on a “real time” traffic snapshot sees the 101 at 6:08 AM starting near De Soto Ave.
The thing is, if you are driving this route, it doesn’t really matter to you what the traffic is like right now, what matters is what it will be like by the time you get there. Just because there is only light congestion at Haskell Ave at 6:08 AM doesn’t mean it will stay that way. Here’s what it looks like to actually drive the 101:
The optimal approach is to not simply use “real time” traffic to decide the best route, but instead to predict the traffic at every point, along every road, and only then choose on the best route based on that future prediction. That’s how we make routes at TallyGo, and that’s why we care so much about understanding traffic — so we can predict it. Our approach is a classic big data solution that uses machine learning and our patented algorithms to derive the fastest possible navigation route.
To learn more about our innovative approach to navigation, visit us at tallygo.com.
By Thomas Haley, TallyGo Data Science