Chances Of Collisions Involving Urgently Underway Police Cars And AI Self-Driving Cars

Police Cruiser in Rush

The dangers of potential collisions between rushing police cars and roadway self-driving cars.

getty

Code 3.

That’s the usual parlance for indicating that a police car will activate its flashing lights and engage a blaring siren as the vehicle is driven at high speeds while rushing to a crime scene. The police car is apt to weave around normal traffic and ostensibly dodge other everyday cars during the heart-pounding effort.

If you’ve ever seen a police car go past you in such a fast-moving manner, the odds are that you were cognizant of the risks associated with that kind of driving. The officer driving the police car is hopeful that other nearby drivers will get out of the way and enable the rushing vehicle to proceed unimpeded.

The thing is, lots of ordinary drivers don’t seem to take the kinds of avoidance actions that they ought to take and that they are typically required by law to perform.

Years ago, most drivers would readily pull over to the side of the road, doing so the moment they became aware of an upcoming police car that was fervently moving along. Nowadays, it seems like there are fewer and fewer conscientious drivers that do so. Perhaps some aren’t sure of what to do, and thus they merely slow down and remain in their existing lane, figuring that the police car will route around them.

Others might not hear or see the police car and therefore fail to take any avoidance efforts. It could be that today’s drivers are in their own bubble, as it were, ensconced in modern cars that can seemingly dampen noises from the outside and also boast impressive internal stereo systems that fill a car with earsplitting levels of musical sounds that you’d hear at a live rock concert.

Of course, there are also those drivers that are otherwise distracted and not entirely devoted to the driving task at hand. They might be watching those enchanting cat videos and ergo not paying attention exclusively to the roadway. Or they might be on their smartphone and chatting with someone at work or trying to respond to text messages while nonetheless still at the wheel of a multi-ton life-or-death in-motion vehicle.

Some assert that there is also a herd mentality that can overtake drivers. If a driver hears the siren or sees the flashing lights, the driver might first look around to see what other cars are doing. Upon noticing that few other cars are taking any avoidance actions, the driver shrugs their shoulders and figures they won’t either. In fact, the driver might mentally calculate that if they opt to move over, it could create a worse situation, whereby other drivers that aren’t doing so will get confused and possibly collide with the one car that is trying to abide by the lawful need to get out of the way.

The whole matter is quite a conundrum.

A potentially deadly conundrum.

A police car that is in a rush might end up coming to an intersection that presents a red light to the vehicle. Rather than coming to a stop, the officer is likely to proceed into the intersection and do so under the belief and hope that cars coming from the other sides will not intervene. This has got to be one of the most unnerving moments for an officer. A car that has the green light might proceed into the intersection and plow broadside into the speeding police car.

All told, you undoubtedly get the gist that when a police car is rapidly responding and underway, there are lots of outsized risks for the officers inside the police car, along with added risks to those driving perchance in the same path as the urgently proceeding emergency vehicle (as a side note, the same could be said for urgently responding ambulances, fire trucks, and the like, though usually, those tend to be larger vehicles and possibly proffer some added protection, though in any case still pose grave risks to all).

An added twist and a boatload of controversy entail a car chase situation. When a police car or a series of police cars are in avid pursuit of a car that is driving recklessly and at high speeds, there is a complex calculation that needs to be made. One aspect involves the risks to the general public during the course of the car chase. Another aspect includes the risks of allowing the car being chased to get away and thus potentially allow a criminal or presumed criminal to be on the streets. Different authorities and experts have various viewpoints on whether or when such pursuits should be undertaken and when those pursuits should be curtailed.

Let’s do a quick overall recap.

There are instances whereby a police officer presses ahead with a designated Code 3 and undertakes what is usually depicted as a hot response. During that effort, the officer driving the speeding vehicle is tasked with getting around everyday traffic and also potentially taking outsized chances when crossing intersections. This means that there is a heightened chance that the police car might ram into a nearby car, and likewise a heightened chance that a nearby car might ram into the police car.

On top of this, intentions need to be weighed too.

By and large, when a police car and an ordinary car collide during a Code 3, this was unintentional. Neither the officer driving the police car and nor the everyday driver of the ordinary car was intending to run into each other. It happened as a result of the circumstances.

In contrast, during a hot police pursuit throughout a frenetic car chase, there is a chance of intentional rammings, such as a police car attempting to ram the fleeing car to end the chase, or the fleeing driver opting to ram a police car as part of their desperate escape efforts.

You can categorize these scenarios into two major facets entailing a police car that collides with an ordinary car versus an ordinary car that collides with a police car. Plus, we can add to this categorization the notion of intentionality versus unintentionality.

We then would have these four general classifications:

·        A police car collides with an ordinary car, doing so unintentionally.

·        An ordinary car collides with a police car, doing so unintentionally.

·        A police car collides with an ordinary car, doing so intentionally.

·        An ordinary car collides with a police car, doing so intentionally.

Most people probably don’t know how dangerous Code 3 can be for police officers.

Per statistics reported by the CDC, the motor vehicle-related line-of-duty deaths for law enforcement officers are calculated as being on average approximately one officer per week being killed on our roadways. This analysis was based on reported incidents from 2010 to 2019. There was an indicated 53 law enforcement officer deaths per year, which included officers killed due to vehicle crashes (25% of the total) and in other circumstances such as getting struck by a vehicle while on foot.

Shifting gears, consider that the future of cars will encompass self-driving cars.

True self-driving cars are driven by an AI-based driving system, doing so without any need for human assistance during the driving task. For extensive coverage about nature, pace, and the predicted advent of self-driving cars, see my column postings at this link here.

Assume for the sake of discussion that self-driving cars will gradually become viable and we’ll see more and more of them on our roadways. Generally, the belief is that the emergence of self-driving cars will initially be quite gradual and it will take many years, if not decades, for self-driving cars to become a prevalent vehicle on our highways and byways.

Consider this intriguing question: Will the advent of AI-based true self-driving cars suggest that we will no longer have any collisions with police cars?

Let’s unpack the matter and see.

Understanding The Levels Of Self-Driving Cars

As a clarification, true self-driving cars are ones that the AI drives the car entirely on its own and there isn’t any human assistance during the driving task.

These driverless vehicles are considered Level 4 and Level 5 (see my explanation at this link here), while a car that requires a human driver to co-share the driving effort is usually considered at Level 2 or Level 3. The cars that co-share the driving task are described as being semi-autonomous, and typically contain a variety of automated add-on’s that are referred to as ADAS (Advanced Driver-Assistance Systems).

There is not yet a true self-driving car at Level 5, which we don’t yet even know if this will be possible to achieve, and nor how long it will take to get there.

Meanwhile, the Level 4 efforts are gradually trying to get some traction by undergoing very narrow and selective public roadway trials, though there is controversy over whether this testing should be allowed per se (we are all life-or-death guinea pigs in an experiment taking place on our highways and byways, some contend, see my coverage at this link here).

Since semi-autonomous cars require a human driver, the adoption of those types of cars won’t be markedly different than driving conventional vehicles, so there’s not much new per se to cover about them on this topic (though, as you’ll see in a moment, the points next made are generally applicable).

For semi-autonomous cars, it is important that the public needs to be forewarned about a disturbing aspect that’s been arising lately, namely that despite those human drivers that keep posting videos of themselves falling asleep at the wheel of a Level 2 or Level 3 car, we all need to avoid being misled into believing that the driver can take away their attention from the driving task while driving a semi-autonomous car.

You are the responsible party for the driving actions of the vehicle, regardless of how much automation might be tossed into a Level 2 or Level 3.

Self-Driving Cars And Police Cars On Code 3

For Level 4 and Level 5 true self-driving vehicles, there won’t be a human driver involved in the driving task.

All occupants will be passengers.

The AI is doing the driving.

One aspect to immediately discuss entails the fact that the AI involved in today’s AI driving systems is not sentient. In other words, the AI is altogether a collective of computer-based programming and algorithms, and most assuredly not able to reason in the same manner that humans can (see my explanation at this link here).

Why this added emphasis about the AI not being sentient?

Because I want to underscore that when discussing the role of the AI driving system, I am not ascribing human qualities to the AI. Please be aware that there is an ongoing and dangerous tendency these days to anthropomorphize AI. In essence, people are assigning human-like sentience to today’s AI, despite the undeniable and inarguable fact that no such AI exists as yet.

With that clarification, you can envision that the AI driving system doesn’t natively somehow “know” about the act of driving. This is a capability that needs to be programmed as part of the hardware and software of the self-driving car.

Let’s dive into the myriad of aspects that come to play on this topic.

First, assume that police cars will remain human-driven for quite some time to come.

In theory, we will ultimately have self-driving cars that are used as police cars, but this is a further out step in the emergence of self-driving cars. The current focus by most automakers and self-driving tech firms is to get an ordinary car to function on a self-driving basis. The somewhat extraordinary driving aspects that would be required for a police car are out-of-scope for the time being and sits on a so-called edge or corner case list for now (meaning that this is a low priority at this time).

Next, assume that we will have a mixture of human-driven cars and self-driving cars on our roadways.

Though some seem to think that a simple snap of the fingers will magically turn all conventional cars into self-driving cars, this is wild talk and not sensible. There are about 250 million conventional cars in the United States alone, and they aren’t going to overnight be transformed into self-driving cars. The reality is that self-driving cars will gradually be made available and incrementally become more and more popular.

The proportion of self-driving cars to human-driven cars will initially be quite low. Over time, the proportion will shift. Whether we will ever completely do away with human driving is an ongoing debate, whereby some exhort that you will only take away their driving privileges when you pry their cold dead hands from their steering wheel.

Okay, we will have police cars being driven by human police officers, and we will have a smattering of self-driving cars and a swath of human-driven cars. That’s the picture to keep in mind during this discussion.

A rushing police car will have to contend with both human-driven cars and self-driving cars.

There is also the oft chance of a rushing police car having to contend with another rushing police car, but let’s put that less likely scenario as out of scope for this particular discussion since it is ostensibly a rarity and has its own specifics to be dealt with.

Your first thought might be that the self-driving car would indubitably do the right thing and abide by the speeding police car, namely that, unlike human drivers, the AI driving system would be programmed to consistently and persistently get over when a police car is zipping nearby.

Yes, that is likely to be true, but it also requires deeper digging to ferret out that there are potential gaps or loopholes, as it were.

Let’s start with the key aspect that a self-driving car would need to realize that a speeding police car is nearby.

How will the AI driving system realize that a police car is on Code 3 and within the driving vicinity of the self-driving car?

Humans use their eyes and ears to try and ascertain that a police car is urgently underway and near to their location. You might see the flashing lights. You might hear the sound of the siren. Furthermore, you can likely determine how far away the police car is, such as by the distant sound of the siren or by how far away the flashing lights are on display.

Also, you might watch for movement associated with the flashing lights and the siren. All told you are trying to figure out whether the police car is getting closer to you or perhaps moving further away. Is the police car up ahead of you, or behind you someplace? And so on.

A self-driving car is outfitted with various sensors that are used to detect the driving scene. This includes streaming video cameras, radar, LIDAR, ultrasonic units, thermal imaging, and the like (this varies by automaker and self-driving tech maker). The AI driving system receives data from the sensors and has to try and combine the data to mathematically compute the driving environment. Part of this involves undertaking MSDF or Multi-Sensor Data Fusion, combining the myriad of sensory data into an attempted cohesive whole.

One sensor that has not been heretofore given much attention is the use of microphones or audio sensory input from the outside of the self-driving car. There are likely microphones inside the self-driving car, allowing for the AI to interact via Natural Language Processing with riders, but there has been less focus on microphones externally mounted on the vehicle.

I think you can guess why this sparsity of external microphones can be detrimental when it comes to the police car scenario. Without an externally focused audio pick-up, the odds of detecting a wailing siren are quite low. Sure, there is an oft chance that a microphone inside the vehicle might hear the siren, though this is probably unlikely and certainly not as reliable as would be an externally mounted microphone.

Currently, this means that self-driving cars would tend to rely more so on the detection of flashing lights and the kind of “facial recognition” of detecting a police car. By recognition, I am referring to the taken-for-granted notion that we can all instantly recognize a police car when we see one. As humans, we know that a police car usually is black and white, has markings that say it is a police car, has mounted light bars, and so on.

Realizing that an AI driving system does not inherently “know” this, the AI developers would need to train the AI on recognizing police cars. This is typically done via the use of Machine Learning (ML) and Deep Learning (DL), consisting of computational pattern matching. In brief, lots of pictures of police cars can be fed into the ML/DL and it will compute the features that seem to characterize police cars.

The point to this somewhat technical discussion is that it is not an assured aspect that an AI driving system will natively somehow be able to detect a police car. Furthermore, detection of the existence of a rapidly moving police car is just the start of the matter. The AI driving system would need to calculate where the police car is currently and where it is likely headed.

Generally, an AI driving system maintains an internal database that is a virtual model of the scene surrounding the self-driving car. Upon potentially detecting a speeding police car, the virtual model would need to be marked that such a vehicle is at such-and-such likely position and that the police car is moving at such-and-such speed and potentially headed in a such-and-such direction.

The AI driving system typically contains a planning component that attempts to identify where the self-driving car should go. Based on detecting an urgently moving police car, the AI driving system would probably be programmed to try and move over, when the timing makes sense to do so, and enable the police car to proceed unabated. There are many complexities entailed in simply moving over, since there might be other cars nearby to the self-driving car or perhaps even pedestrians on the streets, etc.

None of this is surefire.

The self-driving car might not detect the police car. In which case, the AI driving system is going to proceed jauntily as though there is nothing else amiss. You might liken this to the human drivers that are unaware of the presence of a speeding police car and therefore take no avoidance actions accordingly.

When coming upon an intersection that is green to the self-driving car, the AI driving system is likely programmed to assume that the self-driving car can proceed through the intersection. Unless by chance the self-driving car detects the approaching police car, the AI driving system would proceed directly into the intersection and not be taking evasive action or coming to a stop.

All told, we know that there are risks associated with a police car that collides with a human-driven car. There are also risks associated with a police car colliding with a self-driving car. It can happen.

There are also risks associated with a human-driven car that collides with a police car. Likewise, there are risks associated with a self-driving car that might collide with a police car. It can happen.

If the AI driving system is unable to detect the presence of the police car, the only hope would be that at the last moment before a collision, there is a chance that the self-driving car would detect that another vehicle is about to collide with the self-driving car. This might or might not include detection that it is a police car, and merely be that another vehicle is moving at a fast rate of speed and mathematically is going to end up colliding with the self-driving car.

Whether there is sufficient reaction time to then avoid the collision will depend upon the situation at hand.

There is an interesting tangent associated with this matter. Human drivers have a mental model about the actions of other human drivers. We tend to rely upon our “theory of mind” about what other drivers are thinking and what they will do. This can be crucial in those split seconds before a collision and trying to decide what you should do versus what the other driver is going to do.

It is unlikely that many police officers have yet formulated a “theory of mind” about what a self-driving car is going to do. I want to emphasize that I am not suggesting the AI driving system has a “mind” and I earlier stated that it is unsavory to anthropomorphize AI. What I am alluding to here is that a police officer might not be aware of how the AI driving system has been programmed, such that it is unclear as to what driving actions the AI driving system are going to take in such extreme circumstances.

Conclusion

One potential solution to this dire issue of rushing police cars and collisions with self-driving cars involves the use of V2V and V2I.

V2V refers to vehicle-to-vehicle electronic communications. A self-driving car outfitted with V2V could alert other nearby self-driving cars that a police car is rushing down the street on 5th avenue and headed northbound. In addition, police cars can be outfitted with V2V and could beam out electronic messages to alert nearby self-driving cars as to where the police car is and where it is heading.

V2I refers to vehicle-to-infrastructure electronic communications. You might already be aware that traffic signals and other elements of the roadway infrastructure are gradually being computerized. A police car can emit an electronic signal that tells a traffic light to change to green for its passage and thus show red to other vehicles. Self-driving cars will also be set up to communicate electronically with the roadway infrastructure.

These and other kinds of electronic messaging solutions would dramatically lessen the risks of collisions between police cars and self-driving cars.

Indeed, there is a ready argument that conventional cars ought to also be equipped with similar devices, thus alerting human drivers too. One can somewhat confidently state that the AI driving systems will programmatically consistently tend to abide by such alerts, while human drivers are apt to drive with less uniformity (nonetheless, such devices for the human-driven car would still be vital and overall advantageous).

One nifty aspect about self-driving cars is that the chances of a self-driving car getting into a wild car chase are rather unlikely (see my column for further discussion). In other words, you aren’t going to likely see a criminal that is riding in a self-driving car taking the police on a high-speed chase (since the AI driving system is usually programmed to not allow such activity).

As a final thought, for now, imagine the world once there are AI driving systems for police cars. In that case, we’ll have urgently driven police cars being piloted by AI, and they will be dealing potentially with human-driven cars and other self-driving cars on the roadways.

I think you can imagine how that type of AI is going to have its hands quite full trying to cope with traffic while attempting to safely fulfill a Code 3.