Driverless Cars Must Be Wary Of Human Drivers Nearby That Try To Be Traffic Guardians

Dr. Lance Eliot, AI Insider

Image for post
Image for post

Holiday traffic can be difficult to navigate. I had just gotten done doing some shopping at a popular mall and was pulling out of the mall parking lot, trying to enter into traffic onto a rather busy street that had lots of cars zipping along. Though you might think that people would be in a good mood because the holidays were just around the corner, it seemed that most drivers were crazed and driving as though it was their last day on Earth. Drivers were snarling at each other, cutting each other off, and otherwise driving as if it were a dog-eat-dog occasion.

Suddenly, a car came to a stop just before the mall exit that I was waiting at and offered to let me into the stream of traffic. I assumed at first that the driver was probably trying to come into the mall at this juncture and so was letting me go first, getting my car out of the way, so that they could then come into the mall. But, the driver didn’t have a turn signal on, which would have indicated they were going to turn into the mall. Furthermore, I knew this was considered an exit from the mall and though you could drive into the exit area, it tended to make for confusion of drivers in the mall that were usually all jockeying at the exit to get out of the mall.

I smiled at the driver and did a quick wave with my hand, suggesting that they should go ahead and continue forward on the street and not continue to hold-up traffic behind them.

The driver was acting as a traffic guardian, trying to assist other traffic.

You might even be wondering why I would not go ahead and pull out into the street, taking advantage of the car that was holding back traffic.

My first point of hesitation was that the driver that was holding back traffic did not look particularly trustworthy to me. This is a judgement call, I realize. I was somehow slightly suspicious that I might go ahead forward into traffic and the driver might suddenly speed-up. I’ve seen this happen before.

My second point of hesitation was that the traffic on this particular street was moving along at a fast clip and there was a tremendous volume of traffic. As such, some cars were opting to go past the stopped driver and then cut into the lane that the driver was blocking. If I went into the spot that the driver was blocking for me to get into the street, there was also a chance that another car might go around that driver and then try to whip into “my” lane at that moment, and we’d end-up in a potential collision.

The Nature Of Human Driver Traffic Guardians

I’m guessing that you’ve likely had similar kinds of driving circumstances.

Another driver opts to try and do something for you that allows you to get your car into traffic, and you need to decide whether to proceed or not. The other driver is actually ceding the right-of-way to you.

I’ve seen novice teenage drivers fall for a kind of right-of-way ceding trap. These novice drivers are just learning how to best navigate the streets safely. They often assume that the more seasoned drivers on the road know what they are doing. I don’t make that assumption.

A driver that appears to grant you the right-of-way might opt to suddenly retract the offer and catch you midstream trying to take the right-of-way. Or, other drivers that don’t realize what is happening might make moves that end-up clobbering you and/or the right-of-way granting driver.

A novice driver tends to consider these situations to be a godsend.

If you’ve ever tried to turn down the offer from a traffic “guardian” such as my mall exit situation, you know that sometimes the guardian driver is insistent that you take their offer. The driver often assumes that for sure you need to take their offer. It even becomes a kind of game of insistence.

My mall exit example is merely one exemplar of these kinds of traffic guardian moments. There are lots of driving situations involving a driver that decides to become a traffic guardian. They take on the role of directing traffic. In some cases, their efforts are laudable. In other instances, their efforts are misguided.

AI Autonomous Cars And Driver Traffic Guardians

What does this have to do with AI self-driving driverless autonomous cars?

At the Cybernetic AI Self-Driving Car Institute, we are developing AI software for self-driving cars. One aspect for the AI is that it needs to be able to contend with driver traffic guardians.

Allow me to elaborate.

Some AI developers might say that they don’t need to figure out whether someone is a driver traffic guardian. All the AI needs to do is react to whatever driving situation happens to arise.

I consider this rather narrow thinking on the part of the AI developers and I assert it puts the AI itself into a (at best) novice teenage driver mindset, meaning that the chances of a car accident and the danger of injury or killing humans is going to go up if the AI cannot realize what is happening.

Complexities Of Coping With Driver Traffic Guardians

Let’s take the mall exit example.

What would the AI do?

You might suggest that since the AI was intending to leave the mall, and if the stopped traffic allows for the AI self-driving car to get out of the mall, it should go ahead and proceed. No fuss, no particular complexity involved.

Suppose though that the other traffic in the street opts to suddenly swerve over into the stopped lane and collides with the AI self-driving car as it enters into the stopped lane? I’m sure that some AI developers would claim that’s the fault of those other drivers and the AI was “in the right” and thus nothing else matters. But, if the AI allows the self-driving car to get into a situation for which the odds of a collision happening is relatively high, wouldn’t we all agree that this is something the AI ought to be trying to anticipate? I would say so.

There are AI developers that cling to the notion that as long as the “fault” is on the heads of the other drivers, it somehow means that it is Okay for the AI to make rather short shrift decisions that get the AI self-driving car into bad or even dire circumstances.

That’s not the real-world. If AI self-driving cars are going to be on our real-world streets, you cannot somehow expect that the rest of the world is going to gingerly and respectively give AI self-driving cars their own special safety cushion.

Rather than trying to change the rest of the world to accommodate AI self-driving cars, I emphasize that the AI self-driving cars need to be able to be designed and built to fit to the nature of the real-world.

How AI Autonomous Cars Need To Cope

Overall, I would argue that the AI needs to be able to contend with the driving traffic guardian.

The first element of the AI involves it being able to detect that a driving traffic guardian aspect is arising. Detection is key.

Once the matter is detected, the rest involves prediction of what might happen and being able to prepare an AI action plan to deal with the matter.

In the case of the mall exit, the AI via its sensors would have likely been able to detect that the car in the traffic lane had come to a stop. The question arises or should arise as to why has the car stopped there? If the traffic was flowing unabated, and if there is nothing apparent in the front of the stopped car, what else might account for the car having come to a stop?

If the AI can rule out most other relatively common reasons as to why a car would suddenly stop in traffic, this allows the AI to then consider that the car might be stopped due to a driver traffic guardian. This is not a certainty that the situation entails a driver traffic guardian. Just as humans can only guess when they experience such a moment, likewise the AI is also going to be “guessing” that the situation involves a traffic guardian.

I mention this aspect about “guessing” since there are some conventional systems developers that are not particularly used to dealing with uncertainties and probabilities in terms of their coding practices. They write programs that they expect will repeatedly and always do whatever they do without any notion of the “chances” of things occurring or not occurring. In AI development, and especially a human endeavor such as driving a car, you have to alter your programming mindset to include the use of uncertainties and probabilities.

Also, in terms of identifying traffic situations that might include the act of a traffic guardian, it is possible via analyzing large-scale datasets of traffic data to be able to more readily spot such moments. Using Machine Learning (ML) and deep learning, usually consisting of Artificial Neural Networks (ANN), you can deeply analyze thousands upon thousands of traffic situations and train the neural network to be able to identify the driver traffic guardian aspects. When a driving situation begins to arise, the ML can potentially participate in making the detection of it.

The context of the detected driving situation will then shape what action the AI self-driving car should next take. It could be that the AI will opt to proceed and make use of the traffic guardian’s actions. Or, it could be that the AI will opt to wait it out and not exploit the traffic guardian’s actions. The AI might even opt to take some other action such as had I been able to maneuver in my mall exit predicament, I might have decided to not even exit onto the street at all, and perhaps proceed to drive in the mall parking lot to another exit that allowed for a less chancy way of getting out of the mall. The AI needs to consider this possibility too of escaping the situation altogether.

Here’s another twist for you. So far, I’ve focused on having the AI self-driving car detect and react to a driver traffic guardian. Perhaps what’s good for the goose is good for the gander.

In essence, would we want to have the AI be able to act as a driver traffic guardian?

There are some AI developers that bemoan the idea of having to craft a capability to be able to contend with driver traffic guardians. They point out that they perceive this matter to be an edge problem. An edge problem is a portion of a larger problem and it is considered at the edge or corner of the mainstay of the problem being solved. We’re not so convinced it is an edge problem.

The aspect of the AI self-driving car being able to on its own instigate a driver traffic guardian action does fit more so into the edge problem category.

Generally, we assert that the detection and contending with a traffic guardian should be a core part of the AI’s capabilities for driving a self-driving car. The next advancement of having an AI self-driving car that can leverage as a driving tactic the initiation of being a traffic guardian is not quite as essential and therefore could be lumped into the edge problems bucket.

Either way, the traffic guardian elements are vital to driving safely.


Driver traffic guardians, love them or hate them.

You likely have faced driving situations wherein a traffic guardian has helped you out and you were heart warmed that someone would care enough to give you an opportunity to make a driving maneuver that you otherwise felt jammed on. There are probably many other occasions wherein a misguided traffic guardian created a bottled-up nightmare and you figured the person was an utter dolt.

Let’s make sure that the AI of a self-driving car can contend with traffic guardians, both the appropriate ones and the dolts, and ensure that the self-driving car is able to act in a responsible manner that helps ensure safety for all. That’s the kind of automated driving guardians we need on our roads.

For free podcast of this story, visit:

The podcasts are also available on Spotify, iTunes, iHeartRadio, etc.

More info about AI self-driving cars, see:

To follow Lance Eliot on Twitter: @LanceEliot

For his blog, see:

For his Medium blog, see:

For Dr. Eliot’s books, see:

Copyright © 2019 Dr. Lance B. Eliot

Written by

Dr. Lance B. Eliot is a renowned global expert on AI, Stanford Fellow at Stanford University, was a professor at USC, headed an AI Lab, top exec at a major VC.

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store