Lance Eliot, CIO/CTO, Techbrium Inc.
Lance Eliot
Techbrium Inc.

Dr. Lance Eliot is CIO/CTO at Techbrium Inc., overseeing all digital systems efforts. Previously, Dr. Eliot was a CIO at a major Venture Capital firm, a global retail services company, and a top tier home healthcare firm, and has been a successful entrepreneur having launched, run, and sold several startups. He is a frequent speaker at major industry events, author of over 20 books and 300+ articles, and a recognized thought leader on the subjects of Artificial Intelligence, Machine Learning, and also Autonomous Vehicles. He had been a professor at USC and UCLA, and holds a PhD from the University of Southern California.

By Dr. Lance Eliot, CIO and CTO, Techbrium

This article is by Featured Blogger Lance Eliot from his AI Trends column. Republished with the author’s permission.

Years ago, I was driving up to Silicon Valley from Southern California, cruising along at top speed on the freeway, and ended-up getting a speeding ticket in a somewhat unusual manner. Here’s what happened.

I was driving along and minding my own business. Other cars were zooming past me, going much faster than the posted speed limit.  Mile by mile, I gradually let my speed climb. I still wasn’t going as fast as those other cars, but admittedly was now going faster than the speed limit. I knew of course that there was some risk in getting a speeding ticket. But, it was a clear sunny day, I could see a great distance behind me and in front of me, and figured that I would certainly spot a police car and then could reduce my speed before they would realize that I was speeding. There were posted signs on the freeway that said they used planes to spot speeders, so I was even keeping my eyes aimed upward in anticipation of spotting any low flying planes in the empty blue sky.

After a few hours of driving above the speed limit, my being on-watch had waned. Guess I am not much of a criminal in that sense. Suddenly, I noticed back in the distance a highway patrol car behind me and it was coming up very fast. Yikes! I was tempted to pump my brakes, but then I realized it might suggest to the patrol officer that I was trying to reduce speed and therefore must be speeding. So, instead, I just let up on the accelerator and prayed as I watched my speed inch its way down to the allowed speed limit.

Maybe he was too far away to know I was speeding. Maybe he would be more interested in the other cars that had been passing me, since they were going much faster than me. They were the true law breakers! Well, he came up to me, got into the lane to my left, drew in alignment with my car. I assumed at this point that he wasn’t going to pull me over and I had gotten lucky about the whole situation. Surprisingly, he looked over at me, and then made a motion with his two hands of the kind that you would use when scolding a child. He then accelerated rapidly and rushed up ahead.

I decided that a scolding was fine and better than having gotten an actual traffic ticket. My nerves settled down and I opted to remain driving at the speed limit. About five miles later, I could see ahead to the right of the road that there were several cars stopped and parked. This seemed odd. This was a stretch of road that had nothing on it. No reason to be stopped. From time to time, I’d see a car stopped with its hood up, apparently suffering from engine troubles. But I had never seen a slew of cars parked on this open stretch of highway.

As I neared the line of cars that were parked on the shoulder of the roadway, I slowed down due to being concerned that maybe something was amiss up ahead. Road out? Bridge down? Eight-hundred-pound gorilla in the roadway? I then noticed that at the front of the line of parked cars was the highway patrol car that had gone past me earlier. Furthermore, the police officer was standing outside next to his patrol car.  He was waving at me and motioning for me to pull over.

I pulled over and got behind the other parked cars on the shoulder. Whatever was going on, it seemed serious. It was also dangerous because being parked on the shoulder of a highway that had big trucks and cars traveling at high speeds is not a safe place to be. Anyway, I waited to see what was going to happen next.

The officer came up to my car. I rolled down my window. He had those tough guy sunglasses on and looked like someone not to mess with. He then proceeded to tell me that he was pulling me over for speeding. Me, and about a dozen of my fellow criminals. He had gone to the head of the pack of cars, pulled that first car over, and then one by one pulled the rest of us over. At first, I was dumbfounded and had never realized that such a procedure existed. I wondered too how he had known of my speeding. Turns out he claimed that their police plane had been hovering over the highway and spotted all of us speeding.

To this day, some of my friends say that I should have ignored his motioning for me to pull over. My friends claim that I should have just kept driving and that he would have been unlikely to come after me, given that he had about a dozen other scofflaws already in-hand. I could have maybe claimed that I didn’t see his motioning or misunderstood his motioning. On the other hand, I tell my friends that turning a speeding ticket into an all-out police pursuit would probably not have been a wise move.

Why do I tell this story? I tell it because one aspect of self-driving cars that we need to consider involves having a self-driving car pull over when a police request is made to do so. This is what we at the Cybernetics Self-Driving Car Institute call an “edge problem” of self-driving cars (see my column on Edge Problems for self-driving cars). This is something that does not happen very often, but it can happen, and when it does happen then the self-driving car should be able to respond.

Some of you will argue that there should never be a cause for a self-driving car to be pulled over by the police. In this view, a self-driving car is always going to be a legal driver. I have debunked this in prior columns. Self-driving cars are going to potentially do actions that can be construed as an illegal act and so legitimately be pulled over. Even if the self-driving car is being driven legally, suppose the front license plate is missing from the car and it’s a state that requires a front license plate. You can get pulled over for that. Or, suppose that the self-driving car has been used in the committing of a crime, like say it was used by a bank robber. The police might want to pull over the self-driving car, even though the self-driving car itself has not done anything wrong.

I hope this convinces you that it is possible that a police officer would want to pull over a self-driving car. Do not cling to the naïve view that this won’t ever happen.

The next aspect is then what is to be done when a police officer does want to pull over a self-driving car. How will they signal to the self-driving car, and how will the self-driving car know what the signal means?

Those of you living in the future would say that the advent of V2V (Vehicle-to-Vehicle) communication will allow for a police car to send an electronic communication to the self-driving car and tell it to pull over. Though this is certainly possible and a likely future, we are many years away from that future. It will be a long time before today’s cars are outfitted with V2V. I wouldn’t hold my breath about the idea of assuming that self-driving cars will be using V2V with police cars anytime soon.

Another response to the being-pulled-over aspect is that it would be the responsibility of the human driver in the self-driving car to do so. In other words, the human driver in the car should be watching for a police officer that wants them to pull over, and then the human should take over the controls from the self-driving car to enact the pull-over. This is kind of sufficient for the level of self-driving cars of 0 to 3, but not what I believe should happen for level 4, and definitely is not what is intended to happen at level 5 (see my column on the Richter scale for self-driving cars).

In short, we do need to have some means for a self-driving car to do this:

a) Detect that a police instruction to pull-over is taking place

b) Signal to the police that the self-driving car is going to comply

c) Find a means to safely pull-over and come to a stopd) Adjust the stopping and location as instructed by police

e) Alert or inform the occupants of the self-driving car of the situation

f) Come to a safe stop and turn-off the engine as warranted

Let’s take a close look at these aspects.

Detection that a police officer is possibly wanting you to pull over will usually occur by a police vehicle that has its flashing lights going and possibly is blaring its siren. They will normally come up behind your vehicle, directly behind it, and this is an indication they want you to pull over. Note that this is different than when a police car is traveling on alert as an emergency vehicle (see my column on Emergency Vehicle Awareness). In that case, the police car is going past you, while in this case they are coming up directly to you. This can be a somewhat tricky detection because they might come up behind you and really want to get around you, but have not yet found an opening to do so.  In any case, it’s a pretty solid sign that they want you to pull over.

Using the rear camera and radar, along with LIDAR, you can detect the police car and its flashing lights by the images and the speed and location of the vehicle. Though very few self-driving cars are currently audio equipped to listen for sound, we have added this sensory capability to our self-driving cars. The sound of the siren is another clue to the approaching police car that wants to pull you over.

Indeed, often a police car will get behind you and then give a verbal command to pull over, doing so via their loud speaker. Self-driving cars that don’t have any listening devices won’t know this is taking place. The comprehension of what is being said is a hard problem to solve, even though there are solutions such as Alexa and Siri as voice language processing, but in this case it is outdoors, it is mumbled, it is via a loudspeaker, noisy, and so discerning intelligible speech is hard. On the other hand, nearly anything spoken can be considered another likely clue of being pulled over.

This brings up another aspect about self-driving cars. Many self-driving car makers are assuming that the occupants of the car are not interacting with the self-driving car. Our viewpoint is that it is essential that occupants can communicate with the self-driving car. In this case, as a case in point, the occupants could tell the self-driving car that there is a police officer wanting the self-driving car to pull over. The self-driving car can either take their word for it and abide, or it might already have detected other clues such as the police car following them and thus combine the two indications together to confirm that indeed the self-driving car needs to pull over.

Some would say that if the occupants can communicate with the self-driving car that then there is no need at all for the self-driving car to have to ferret out whether a police car wants them the self-driving car to pull over. Presumably, the occupants can just tell the self-driving car when it should pull over.  We don’t agree with this generalized aspect.

Keep in mind that there are circumstances wherein the occupant might be unaware of the police request, such as if the occupant has gone unconscious. This does though bring up a thorny issue. Suppose the occupants of the self-driving car do not want to be pulled over by the police. This is the scenario of the bank robbers that are in the self-driving car.  Keeping in mind that it is generally lawful for a police officer to pull over a car, it really in some ways can be argued that it doesn’t matter whether the occupants want to be pulled over or not. I certainly didn’t want to pull over when I had the highway patrolman motion me, but I was obligated to do so.

Once the self-driving car has arrived at the notion that it needs to pull over, we then enter into the next stage of this process. If possible, the self-driving car should signal to the police officer that it is going to comply. This is important because otherwise the police officer might become concerned that the self-driving car is trying to make a getaway. A human driver would usually turn on their turn indicator to signal that yes, they are going to get over. Using the conspicuity features of a self-driving car (see my column on this), the self-driving car should signal to the police officer that it is going to pull over.

Next, the self-driving car has to find a suitable spot to pull over. This in one sense should be something that the self-driving car is already considering, even prior to the police coming up. I say this because we believe that a self-driving car should always be calculating where it would be safe to stop. In other words, while driving, it is always possible that a car might need to pull over. Human drivers aren’t usually continually considering this notion, but to some degree their subconscious is kind of aware of their surroundings and generally knows when things look safe or not. We always have this being run in the background by the self-driving car AI.

A self-driving car should not do reckless acts like coming to a stop in the middle of an intersection in response to the police indication to pull over. Alleys are usually bad, as would be any terrain that is rough and not suitable for driving onto. Preferably, the area to stop should be well lit and safe to be parked in, like a wide shoulder or perhaps even exiting a road to get into a parking lot. Once the car has found such a spot, sometimes a police officer gives further direction, such as ordering a driver to drive ahead fifteen feet, or pull into a driveway. The self-driving car tries to interpret such commands, though as mentioned earlier this is currently a hard problem.

During the time that the self-driving car is pulling over, one added aspect would be to potentially call 911 and report that the self-driving car is pulling over at the request of a police officer. The 911 would then relay to the police car that the self-driving car is trying to comply. Human drivers can do the same thing, though it is somewhat rarely done. In rare cases too, the officer tells 911 to tell the driver to take action. This is somewhat a better approach to receiving verbal commands since the amount of noise on the phone line is likely much less than what is commanded via a loudspeaker.

After safely settling into whatever spot has been found to stop the car, the engine would normally be turned-off by a human driver. In the case of a self-driving car, the question arises as to whether or not it makes sense to turn off the engine. If the use of the engine is required to generate power for the self-driving car, then turning off the engine will put the self-driving car AI out of commission. If the AI is powered separately then it is possible to go ahead and turn off the engine of the car.

The rolling down of the window to allow for the officer to talk with the occupants could be an automatic feature, or it could left to the occupants to manually do (or verbally command the self-driving car to do). Providing of a car registration will be for now still likely a paper-based exercise, though in the future it is info that can be readily available electronically and transmitted to the officer in that manner.

Trying to talk you way out of a ticket is not something that we envision a self-driving car to do, but, hey, maybe that’s a future feature that consumers will want. In any case, seriously, the ability of a self-driving car to pull over when instructed by a police officer is a useful and usable aspect for a self-driving car to know how to do. We covered here the aspects of a police vehicle that comes up from behind to signal to pull over, but as my story indicated earlier there are other ways this can happen too.

Also, another aspect about self-driving cars will be all of the data they are collecting while driving, and whether or not this info should be given to the police. If a police officer claims your self-driving car was speeding, and you refute the claim, you might want to dig into the data being collected by the self-driving car as support for your claim that you weren’t speeding. This also then raises issues about the security of the data and whether you could have doctored it for purposes of supporting your claim.

Originally published on AI Trends.