Drivers Use Micro-Movements While Driving, Driverless Cars Don’t, Whoops!

Dr. Lance B. Eliot, AI Insider

Image for post
Image for post
Micro-movements while driving are subtle tells or clues of driving behavior

I was eagerly awaiting making a right turn on a red light at a busy intersection that led onto the always hectic Pacific Coast Highway (PCH). There seemed to be a gap in traffic so I gunned my engine to try and launch into the rightmost lane, when all of a sudden another car opted to swing into the lane, and I had to hit my brakes before I became an interloper that caused a pile-up.

Another driver that was behind me had assumed I was going to make the right turn and so they had followed me forward. When I hit my brakes, the other driver realized they needed to do the same, luckily. No hit, no harm. But it was a close call. Nobody else on planet earth would even realize that this moment had occurred, other than me and the car driver behind me. It all happened in an instant.

I’ve told you this story to bring up something that is quite important and yet often overlooked as an aspect of human driving behavior, namely the use of micro-movements when driving.

A micro-movement is considered a somewhat subtle driving action that exists somewhere in the gray area between doing an overt and obvious driving maneuver and doing essentially no noticeable driving maneuvering at all.

Some Drivers Oblivious to Micro-Movements

Some human drivers are oblivious to the nature of micro-movements. These drivers are blind to noticing micro-movements and can’t readily tell you what a micro-movement looks like. Their attention to the driving task is based on whole-scale movements and maneuvers, ones that are stand out like a brass band.

You don’t necessarily need to be in a car to see the micro-movements of other cars. I ride my bike quite a bit and anyone that is hopeful of surviving the riding of a bike in city traffic is likely to be an expert on the micro-movement evidenced by cars and those wacky human drivers.

It takes a keen eye to watch for telltale signs of what a car and its driver are going to do next. You might see the wheels slightly get angled. You might hear the sound of the engine as it begins to rev up. You might notice the car braking in a series of precursors to a turn or other maneuver.

Here are some the key ways that micro-movements can appear:

  • Angling of the wheels of the car
  • Leftward leaning of the car
  • Rightward leaning of the car
  • Stuttered motion of the car (start/stop)
  • Car lurches (sudden motion)
  • Riding of the brakes
  • Etc.

These aforementioned car movement indications are considered micro-movements when they are done in a subtle manner. It’s a kind of a tease. Just a little bit of showing of the leg, as it were.

Some might liken this to a “tell” in a sport like basketball. When playing basketball competitively, you might give a movement of your head that seems to say you are going to make a run to the basket, and the other player covering you can potentially read the tell and get positioned to block you. This can all happen in an instant.

Your Tell Gives You Away

If you like to play poker, you certainly know about the tell or clues that other players often inadvertently show.

While in college, I used to play poker on Thursday nights with a bunch of friends, and one player had a tell that involved pushing his glasses up his nose and toward his eyes. We all eventually figured out that this was his tell. It meant that he was lying about whatever he might be saying. None of us revealed the tell to him, and we kept it each to ourselves, relishing it. After months of this, one of the pack opted to tell him about his tell.

At first, he was angry that we had noticed the tell. It then dawned on him that if we had become conditioned to his tell, he could use that conditioning to his advantage. For a few weeks, he would purposely push his glasses on his nose, doing so to make us think he was lying, even though he might be telling the truth.

This highlights my earlier point about micro-movements of cars. Some people notice them, some do not. And it also highlights another very important point. You cannot for sure bet on a micro-movement since there is no guarantee it will ultimately graduate into becoming a full-on movement. Not all tells blossom.

Similar to my story about my poker playing friend that ended-up exploiting his tell, there are some savvy drivers that like to use their micro-movements in an exploitative way. These drivers often assume that other drivers will notice the micro-movement and then back-off or otherwise give way to the driver that is using the micro-movements.

This happens a lot in parking lots. When cars are driving around and around trying to find a parking spot, it can become a gamesmanship effort of who suggests they found an open spot first. You’ve likely seen two cars that came open an open spot, and the cars are facing each other head-to-head. Which one will get the spot? If one driver angles their wheels, it can be a kind of assertion that they have tagged the spot and are going to aggressively make their way into the spot.

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

At the Cybernetic AI Self-Driving Car Institute, we are developing AI software for self-driving driverless cars. One aspect that is not yet getting much attention involves the micro-movements of driver behavior, which some consider an edge or corner case problem.

I’d like to first clarify and introduce the notion that there are varying levels of AI self-driving cars. The topmost level is considered Level 5. A Level 5 self-driving car is one that is being driven by the AI and there is no human driver involved.

For self-driving cars less than a Level 5, there must be a human driver present in the car. The human driver is currently considered the responsible party for the acts of the car. The AI and the human driver are co-sharing the driving task.

Another key aspect of AI self-driving cars is that they will be driving on our roadways in the midst of human driven cars too.

Micro-Movements and the Development of AI Self-Driving Cars

Returning to the topic of micro-movements of human driving behavior, let’s consider how these actions come to play in the design and development of AI self-driving cars.

I usually get asked two questions about this topic, which comes up when I am speaking at industry conferences.

First, does it matter whether AI self-driving cars are developed such that they will be able to detect the micro-movements of other cars?

My short answer is that yes, it does matter and AI developers at the auto makers and tech firms should be including this capability into their AI self-driving cars.

Most of the auto makers and tech firms are not yet encompassing the detection of micro-movements. This lack of attention to the topic is due to several reasons, including that they are too busy with other “core” aspects of AI driving, along with a lack of awareness about the importance of micro-movements as a driving tactic and strategy.

If you take a look at many of the AI self-driving cars being tested on our roadways today, you can usually readily figure out that they aren’t using micro-movements to their advantage. The stilted nature of the AI driving is a giveaway that the AI is not versed in detecting, analyzing, and making use of the “tell” of human drivers making micro-movements.

Admittedly, getting the AI to become versed in micro-movements requires some pretty heavy lifting in terms of AI development, plus itis best done when you also include savvy Machine Learning (DL), Deep Learning (DL), and Artificial Neural Networks (ANN) into the mix.

Trying to use the sensors of an AI self-driving car to detect that the wheels of a car ahead are slightly askew, it’s not easy, and you need really good sensors and really good sensor processing software. The same can be said about detecting when nearly any of the typical micro-movements are undertaken. By definition, those movements are subtle. Picking up on the cues or tells of other cars is hard for other human drivers to do, and admittedly hard for an AI system to do.

Training an AI Neural Network for Micro-Movement Driving Patterns

Fortunately, you can leverage Machine Learning and Deep Learning to your advantage. By training a deep or large-scale multi-layer Artificial Neural Network with tons of collected driving related data, based on human driven cars as they are driving in regular human driving traffic, there are patterns of the micro-movement driving that can be trained for.

These trained beforehand Machine Learning and Deep Learning algorithms can then be included into the on-board AI system of the self-driving car. In many respects, this is really just a deeper form of those capabilities. The odds are that Machine Learning and Deep Learning are already being used to examine sensory data for aspects such as street signs, the presence of pedestrians, etc. The micro-movements are essentially a step deeper into that kind of analysis.

You also need to realize that the micro-movement is not merely about detection. Once you’ve detected a micro-movement, it needs to be populated into the AI virtual world model. The AI action planner has to be then be versed in what to do with the indications about the micro-movements. If the AI action planner ignores the micro-movement detection, the whole effort is for naught.

The AI action planner also has to deal with micro-movements that might not be a true tell per se.

In this case, the AI needs to be able to include the micro-movements as part of a larger picture or “understanding” of what is happening in the traffic around the AI self-driving car. You cannot in isolation try to make use of the micro-movements.

It is also important to realize that human drivers are going to try and spoof or prank AI self-driving cars. If you as a human driver know that the AI is going to watch the alignment of your wheels, you could do a kind of basketball-like head fake and shift the direction of your wheels. It might not be due to actually needing to turn the wheels, but instead a ploy to get the AI to perhaps grant you that sought for parking spot or otherwise give you an advantage over the AI system that’s driving the self-driving car.

I’ve tried to answer the first question which was whether or not AI self-driving cars should be outfitted with a capability to detect, analyze, and leverage the micro-movements made by human drivers. The answer is a resounding Yes.

The second question that I get asked is whether the AI self-driving car should make use of micro-movements, doing so in a similar manner to how human drivers use micro-movements.

My answer is yes, the AI ought to also make use of micro-movements. Here’s why.

An AI self-driving car should be sending the same kind of subtle “signals” to human drivers, doing akin to what other human drivers do. This again is part of my overarching belief that if AI self-driving cars are going to be driving among human drivers, which we know to be the case, those AI drivers need to be doing actions that human drivers do.

If the AI lacks the micro-movement showcase capability, it means that those human drivers around the AI self-driving car will no longer have an essential “tell” that can forecast what the AI self-driving car is going to do. Without the tell, the human drivers are going to be caught unawares.

Conclusion

I sometimes liken the micro-movements topic to the nature of dancing. If two people are going to dance together, they need to figure out the subtle movements that indicate whether to turn to the left or turn to the right, and whether to speed-up or slow down the pace.

Each dancer needs to be able to detect the micro-movements of the other, along with knowing how to react.

Furthermore, each dancer needs to be able to exhibit their own micro-movements, in order to allow their dancing partner to detect the movement and be able to react accordingly.

Right now, we have humans that detect the micro-movements of other human drivers, plus human drivers emit micro-movements as part of their driving efforts (by-and-large). We are in the midst of witnessing AI self-driving cars being fielded and tested that don’t detect the micro-movements of other cars, and nor do the AI self-driving cars emit their own micro-movements.

It’s time to make sure that the dancing partners of human drivers and AI drivers are versed in the same kinds of capabilities and tactics of driving. I vote that the auto makers and tech firms put more effort toward the micro-movement elements, which makes sense to ensure that the dancing partners won’t collide into each other and make ghastly missteps.

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

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