By Lance Eliot, the AI Trends Insider

I was in an airport and on my way to give a speech about self-driving cars when I realized that my smartphone was getting low on its battery charge and I knew that the flight itself would not have any power outlets at the seats. I looked around the gate area to see if there was a place to plug-in my phone to charge it. Most airports seem to have charging bars where you can plug-in your phone, but this airport was apparently stuck in the 1980s and there weren’t any designated areas for phone charging. My next plan of attack was to find an electrical outlet, preferably one that was near some seats so that I could sit down and be right next to my phone as it was getting charged up.

Disappointingly, this airport wasn’t just in the 1980s and instead, worse still, was in the Neanderthal era because there weren’t any electrical outlets anywhere. In fact, it almost seemed like the airport had purposely tried to hide or close-off any chances of finding an electrical outlet. There were some desperate business people that had reached behind a candy dispensing vending machine and had plugged in their phones there (they even had worked together to push the vending machine a few inches from the wall to then get access to the precious electrical outlets behind it).

I was getting desperate as my smartphone dropped below 10% and at any moment it would be entirely without power. I turned it off to preserve what I considered “emergency energy” in the battery and continued my search. After walking endlessly in the airport, I finally decided there wasn’t any place available to charge my phone and had a revelation that I could use my laptop to charge my phone. It meant though that I’d be depleting my laptop charge, but it was worth it to get my smartphone back into the green zone of available charge.

I’m sure that you’ve encountered similar dilemmas of trying to find a place to charge your smartphone and having to be cognizant of how much battery you’ve got left to go.

In my case, it was admittedly not a life-or-death matter as to whether I had a well-charged phone or not — in theory, I could have just gotten onto the flight and slept the whole way to my destination. Like most people, I just feel better to know that my phone is charged and ready for use. Without a charged-up phone, it’s almost as though I had left my wallet at home and would miss being able to use my smartphone as much as missing having my wallet with me. You might say I am overly dependent on my smartphone, but I dare say there are a lot of people that feel the same way about having a charged and usable smartphone with them.

Electrical Power can be a Life Saver

There are some occasions where the amount of battery charge in an electrical device of some kind could be a life-or-death matter. A notable scene in the movie “Apollo 13” showcased a famous life-or-death moment when the astronauts on the Apollo 13 space craft had to deal with a dicey situation involving electrical power and battery charges. The day was April 13, 1970 and the Apollo 13 spacecraft was over 321,000 kilometers on the way to the moon, when unexpectedly one of their oxygen tanks exploded, they no longer could produce electricity with their service module fuel cells (those cells needed oxygen to aid in producing electrical power).

The astronauts and NASA ground control realized in those tense moments that they would need to quickly power down the spacecraft and try to conserve power. This was especially important since they would need sufficient power leftover to run the spacecraft once they were getting back toward earth. If they didn’t have enough power remaining by the end of the trip, they’d likely die as they tried to make that last step to get back here. Powering down though meant they would need to operate without much power during the bulk of the return trip, and this included not being able to run the heaters in the spacecraft (turns out, this led to the interior of the spacecraft becoming bone chillingly cold and dangerously wet, which is not a normally prudent approach and is illustrative of how desperate the situation had become).

Even the power-up sequence had to be carefully orchestrated. Down on earth at NASA there was a scramble to use the spacecraft simulator to figure out the best power-up procedure for the astronauts to use. Different sequences would consume differing amounts of battery charge. By experimenting with a myriad of sequences, they were able to find a series of steps that would help to minimize the electrical charge consumed at reboot.  As you know, fortunately the Apollo 13 crew was able to get home and it turned out to be an incredible indication of the savviness of our space efforts that such a dire situation was successfully overcome. Sidenote: If you do a search on the web for historical documents about the Apollo 13 mission, you can find a copy of the actual notes that the astronauts made on their spacecraft instruction manual as they were dealing with the power issue.

I freely admit that whether my smartphone has a sufficient charge or not would probably be less consequential than what the Apollo 13 crew faced, but nonetheless the amount of battery on my smartphone is important to me. We all tend to watch our smartphones and keep aware of how much charge we have left at any time.

Cars and Their Energy Needs

On a related topic, here’s an interesting question for you, do you pay attention to how much battery charge there is in your car? In other words, do you put as much attention to the charge of your car battery as you do of your car phone? I’d dare say that if you drive a gasoline powered car that you rarely think about your car battery charge. You just assume that your car battery is working and properly charged. Until you have a moment where you try to start your car and it won’t start and makes that irritating noise like the battery is dead, you assume the battery is good.

In contrast, those of you that have an electrical car, you are likely aware of the battery charge in your car. You need to make sure that you have sufficient charge to go wherever it is you are aiming to go. This requires knowing how much total charge capacity your battery has, how long it takes to charge it, and know of places that you can stop to get a charge. Now, I don’t want to overstate this aspect since I realize that electrical car owners point out that you can argue the same thing about gasoline powered cars — gas powered car drivers need to know how much gas is in the car tank, they need to know how miles they can go on the remaining gas, and they need to know where they can find a gas station to fill-up.

For a conventional electric car, some drivers even try to help reduce the power consumption by turning off their infotainment system and their air conditioning whenever they get toward the last trickle of electricity left in their car battery (yes, gasoline powered car drivers often do the same to reduce gas usage). The amount of charge available determines the distance an electric car can drive, and no one wants to get stuck on the roadways with a car out-of-charge (akin to being out-of-gas). I guess we all face our own Apollo 13 moments when driving our cars, having to be aware of either our remaining battery charge or our remaining gasoline amount.

Overall, it is important for those dependent upon a car to be aware of:

•         Energy consumption (how much energy your car uses up)
•         Energy storage (how much energy your car can store or has in storage)
•         Energy generation (how much energy your car can create or generate)

Energy Vital to AI Self-Driving Cars

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

At the Cybernetic Self-Driving Car Institute, we are exploring ways to conserve power consumption via the AI of the self-driving car. Just as the Apollo 13 ground crew and spacecraft crew figured out how to best use their electrical power, the AI of the self-driving car can be doing likewise.

You might be wondering why electrical power for a self-driving car is an issue at all.

Here’s why.

For an AI self-driving car, there needs to be an abundance of specialized sensors on the car to be able to sense the world, including video cameras, radar units, ultrasonic units, LIDAR, and so on. Each of these are electrical devices that consume electrical power. Furthermore, the AI needs lots of computer processors and on-board computer memory, and other related electrical devices, all of which consume electrical power too. That’s one layer of electrical power needs for the self-driving car. In addition to the crucial power needs of the AI self-driving car specialized components, there are the other commonplace electrical gulping aspects of the car.

And, you need to consider that many AI self-driving cars will be running non-stop 24×7, which is partially due to the desire to use them as ridesharing service money makers. There is also the notion that since we’ll be inside of AI self-driving cars a lot, being in a car more so than we do today, there will be elaborate infotainment systems inside of the AI self-driving car (I’ll describe this in more detail herein shortly). The odds are too that there will be numerous Internet of Things (IoT) devices inside of an AI self-driving car, either ones that come with the self-driving car or IoT devices that the consumer brings into the AI self-driving car, which can also devour power provided by the self-driving car.

In essence, you need a miniature power plant to be able to operate all of the add-on’s that make an AI self-driving car into being an AI self-driving car, along with a robust power storage system that can provide a needed buffer of electrical power.

Sufficient power needs to be available to operate the self-driving car components at all times needed, which of course involves when the self-driving car is in motion, but also when it is idle, and even to some degree when the AI self-driving car is no longer considered in active use for a driving journey. Suppose you opt to park your AI self-driving car at the airport for a week-long business trip. Some aspects of the AI self-driving car might very well be trickling energy and depleting the electrical power stored in the battery while you are away. It could be that when you return from your business trip that your AI self-driving car won’t start due to a dead battery.

If your AI self-driving car won’t start, it’s a multi-ton paperweight. Worse still, without any power, you are not going to be able to even communicate with it, such as trying to tell the AI that you want the car to start-up and meet you at the airport airline terminal when you arrive.

An equal if not even greater concern is whether there is enough electrical power to sustain the AI self-driving car once it has gotten underway. Starting it is one thing but making sure that the AI is well-fed with electrical power and that the sensory apparatus is well-fed with electrical power is perhaps even more vital during a driving journey.

Without sufficient power available, the self-driving car would effectively become either blind to the world around it and be unable to sense the surroundings, or the AI itself would be cut-off if the processors had insufficient power. The electrical power for an AI self-driving car is its lifeblood and needs to be carefully orchestrated and managed, doing so for the viability of the AI self-driving car and the safety of humans either occupying the AI self-driving car or those that might come in contact with the AI self-driving car (such as pedestrians or humans riding in other nearby cars).

Some Details About Energy and Self-Driving Cars

For any conventional car, electrical or gasoline, you need electrical power to do the normal things for powering a car, including providing start-up power to get the engine going, power for the headlights and signals of the car, etc. That’s a fair amount of power right there. Plus, people are nowadays expecting to plug their smartphones into their car and often use laptops too, all of which consume more power. Then, we’re adding a huge additional layer of automation and electrically powered devices to do the self-driving of the car. An AI self-driving car becomes one big power-hungry rolling monster.

For Level 5 self-driving cars, which are ones that need no human driver and the AI is supposed to be able drive in whatever manner a human can drive, we are anticipating that the human occupants in the self-driving car will likely be consuming power in new ways.

One concept is that the inside of the self-driving car might have large LED displays throughout the inner walls of the car, allowing people to watch movies while the AI drives the car for them. Various interactive devices like laptops, but maybe something beyond laptops and that hasn’t been invented yet, will be inside the self-driving car for purposes of work or play. Nobody really knows yet what the inside of the futuristic AI self-driving car will be, but you can pretty much bet that whatever it is will require the consumption of electrical power.

Some even say that the exterior of the self-driving car will also have electronic displays. We might have outside electronic boards that help communicate to other drivers what our self-driving car is doing, such as it might display that it wants to get over into the next lane. Or, the electronic displays might be used for advertising purposes and the owner of the self-driving car could make extra money by agreeing to have ads displayed. As an aside, there are others that say this doesn’t make much sense because people will be working or playing inside the cocoon of their self-driving car and will rarely if ever look outside. If they aren’t looking outside, they won’t see ads being displayed on the outside of other cars (though, pedestrians could see it, admittedly).

Anyway, there’s pretty much no disagreement that AI self-driving cars are going to crave electrical power.

Until we invent some other novel power source, all of these needed devices are going to be consuming electricity. It’s as though you need a nuclear power plant inside of your AI self-driving car. That being said, you might at first glance assume that an all-electric car would somehow then be the best candidate for AI self-driving cars, since it’s all about producing electrical power.

Ironically, there is a view that a hybrid car is maybe the better choice for AI self-driving cars.

The belief is based on the idea that in case you run out of electrical power to run the car, you can at least switch over to the gasoline and keep your car going for a while. Not everyone buys into this logic. They say that if you consider the gasoline as though it is a virtual reserve of electrical power, why shouldn’t you just do the same as to your actual battery and electrical power? In other words, just make sure to keep the same amount of electrical power in reserve. One counter-argument is that the ability to get gasoline is generally much easier today, since there are gas stations on almost every street corner, while electrical charging stations and locations are scarce. This might change in the future, but for today, certainly it is relatively easier to get gas than it is to get your car charged.

With AI self-driving cars, the belief is that you’ll put your self-driving car to use all day long, each day, for a usage of near to 24×7. When you are at work, you’ll hire out your AI self-driving car as a ride sharing service. When you are at home sleeping, your self-driving car will be out making money for you doing nighttime ride sharing. You might also use your self-driving car to do errands for you, such as sending your self-driving car over to the grocery, where the store clerks have pre-assembled your ordered items and are ready to place them into your car, which your obedient self-driving car then dutifully brings those groceries back home to you.

If your self-driving car is continually on the move and going all day long, it means more power consumption than a conventional car, by far, and also raises the issue of how will your AI self-driving car be able to keep charged-up. Presumably, we’ll have gas stations and charging stations that will accommodate self-driving cars, including ones that have no human occupants inside at the time the self-driving car comes to do a fill-up. There might at first be human attendants at the gas and charging stations, but we will ultimately likely see unattended stations, meaning that your unattended AI self-driving car can go into an unattended gas station or charging station and fill-up without any needed human intervention.

The experimental AI self-driving cars that are being tested on roadways are prone to the same issue of consuming a whole lot of electrical power, and so you might assume that they’ve figured out how to deal with the power consumption crunch. Not really. Many of these experimental self-driving cars are not being run all day long, and they are running at only low speeds, and they are used only for short trips, and otherwise are being used in a manner which is contrary to what will likely happen once they are released into the wild (i.e., used by the general public).

Furthermore, today’s AI self-driving cars are adroitly pampered by the auto maker or tech firm and so we don’t hear about the power consumption issue (the auto maker or tech firm has an entire human pit crew at-the-ready). If a self-driving car has a dedicated team of engineers, they are taking care of the power aspects and no one else would realize that the power might be a problematic aspect. Once these cars get into the hands of consumers, the consumer perception of having to deal with the power consumption could potentially undermine their joy for these electricity swigging vehicles.

Another twist involves the over-the-air (OTA) updates that are anticipated to be needed for AI self-driving cars. There is a communications capability included in most AI self-driving cars that allows for the AI system to communicate with a cloud-based system setup by the auto maker or tech firm. The self-driving car uploads data and the cloud downloads data and programs into the self-driving car. For some of the self-driving cars, they are made to mainly do this only when the self-driving car is sitting still, stopped, and when the car is not being charged. This will cut into the available time for charging the self-driving car.

By the way, there’s another angle to these many electrical devices on your self-driving car, namely the heat issue.

You might have used a laptop that generated heat while it was running, doing so lightly when the laptop was under minimal use, and then getting quite hot when playing that latest online game and really exercising the processor and electrical components. Imagine that your self-driving car has all of these exotic electrical devices for radar, cameras, and the like, and the numerous computer processors too, and when they are all going full blast it generates a tremendous amount of heat.

The AI self-driving car has to be designed to properly dissipate the heat, otherwise the electrical devices will fry. You can’t afford to have your radar unit sizzle out while the self-driving car is hurtling down the freeway at 80 miles per hour.  Ironically, the systems that will dissipate the heat often require electrical power themselves and thus consume more of the battery of the self-driving car.

There’s also a question of where to put all of these various AI self-driving car devices. Some of the devices need to go in certain areas of the self-driving car. For example, the cameras need a clear line of sight and are mounted at the front, sides, and back of the car. The processors and related electrical devices can be mounted elsewhere in the car, but you can only hide them in so many places, and right now some auto makers are putting these processors into the trunk. Consumers though aren’t going to be happy with a self-driving car that has zero trunk space.

Being Smart About AI Self-Driving Car Power

Overall, as much as possible, the AI self-driving car makers are going to need to find ways to reduce the amount of electrical power consumed (using the power smartly and not just by happenstance), along with finding innovative ways to generate power where feasible.

You might argue that we just need to increase the ability of the self-driving car to generate and store electrical power, which certainly is an equally strident goal, and so in that case it doesn’t really matter how much power is consumed. A self-driving car that has a tremendous sized battery and an ability to generate power while underway will help to mitigate the power consumption problem.

As an innovative example of generating power, there are efforts underway to try and harvest the energy that arises from the tires as they rub across the roadway surface. Research indicates that you can potentially turn this triboelectric effect into storable energy via the use of nanogenerators that you would have on-board the self-driving car. Some argue that you’ll be losing a lot of the energy while trying to collect it, but others point out that as long as the energy gained exceeds the energy used to grab the added energy, the equation works to the favor of the self-driving car.

Advances in car design, battery design, and electrical power generation in cars is ongoing and I’m sure we’ll gradually see marked improvements.

Meanwhile, we need to assume that it is vital to find ways to conserve power for the advent of AI self-driving cars.  You can liken this to smartphones. Early smartphones were battery hogs and only had small batteries. Power management became a big topic of how to decrease the smartphone use of power. Meanwhile, batteries got better. But, at the same time, smartphones got more features and increased how much power they use. It’s an ongoing cat-and-mouse game, improving the batteries and how power is consumed, and the two will go hand-in-hand in the same manner for AI self-driving cars.

One of the most obvious ways to try and reduce power consumption is the nature of the electrical devices themselves that are on the self-driving car. Few of the makers for the radar units, cameras, and other sensory devices have been pushed hard on the power consumption issue up until now, and the push has been mainly for those devices to get better at what they do and get smaller in size. The smaller size is good for self-driving cars since it reduces typically the weight and the bulk of where it will sit in the car. Gradually the power consumption issue has become a bigger issue that they are trying to solve.

Same can be said about the computer processors and memory on-board of the self-driving car. We already know that for purposes of making laptops lighter and smaller, there has been a push towards processors and memory that are smaller and use less power. Nvidia recently produced an 8-core CPU that they claim can do 30 TOPS (trillion operations per second) and only consume 30 watts (called Xavier), and they have another processor that supposedly does 320 TOPS at 500 watts (it’s called Pegasus), all especially earmarked for AI self-driving cars. These kinds of advances are encouraging signs of reducing the electrical power needs of a self-driving car.

One of the ways that some are now describing the power consumption of the AI self-driving car is to liken it to some approximated number of laptops. In other words, suppose your AI self-driving car had (let’s say) 50 to 100 laptops that were all working at the same time and at their peak energy needs and were inside of your AI self-driving car. That provides an overall sense of the magnitude of the amount of power being consumed just for the AI and self-driving car aspects of the car.

As mentioned earlier, one hope for reducing the power consumption involves having the device makers try to design the devices to use as little power as possible. Another variation of this idea would be to have the device itself moderate how much power that it needs. In essence, have a device that is “smart” and knows when it can reduce its power consumption. Again, somewhat similar to what our laptops today do, your laptop goes into a sleep mode when it thinks there’s a moment to do so, cutting down on its power use, and then wakes up when it needs to do so.

For example, you might have processors on the self-driving car that go into a sleep mode, reducing their power consumption. They then kick back into use once the processors need to do something. We could try the same trick with the sensor devices. Maybe the radar unit for example goes into sleep mode, reducing its power consumption, and then when needed it wakes-up and at that time the power consumption rises.

On paper, this seems like a good idea. In reality, have you had your laptop go into sleep mode and then it balks sometimes at waking up? Imagine if your self-driving car is going along the highway at 65 miles per hour and the radar decides to go to sleep because it hasn’t been used in the last few milliseconds. Great, a tiny savings on electrical consumption. But, then suppose the radar is urgently awoken, but freezes up, just when it is needed to detect that a large truck might be veering toward the self-driving car. Do you want your life dependent upon an electrical device that maybe or maybe not can come out of sleep mode?  It’s unnerving, for sure. In whatever manner this is to be undertaken, it needs to be smartly managed.

AI Providing Needed Smarts for Power Management

Can the AI itself help with the power consumption issue?

We believe that the AI can and indeed should be involved in the power consumption solution.

Let’s start with the idea of a device that is “smart” enough to do its own power management. If you allow an individual device to decide whether it can go into a sleep mode, the device itself might or might not choose a good time to do so. The AI of the self-driving car will have a better idea of when it might or might not need a particular device to be working. It is the AI that has the overall plan of what the self-driving car is doing and what it will do next. Individual devices usually do not have that wide of a scope of awareness, and are only focused on their specific task at-hand.

Therefore, it makes sense to have the AI aid in determining when a specific device can take a siesta, one might say, and reduce its power consumption. The AI then also has to be smart enough to realize how this blinds or otherwise adversely impacts the self-driving car. And, the AI needs to be aware of how long it will take to wake-up the device and reengage the device.

The point here is that the AI must do power management in a savvy manner. It cannot just arbitrarily start telling devices to shut down or go into low power modes. Instead, the AI has to be determining when to have any device do its power consuming efforts, and what impact it has to the driving mission and journey. Plus, the AI must anticipate the likely amount of time that will be required for the device to wake-up and function and be calculating the odds that maybe the device doesn’t wake-up in time and so have contingency plans in case it does not.

Also, the AI must know what kind of power consumption each device has, along with in what circumstances it uses that power. Shutting down momentarily five tiny power consumers might not be as advantageous as shutting down a single power consuming hog. There are some devices too that the AI might rarely if ever allow to shut down because they are mission critical to the driving task, and others of a lesser importance that it is willing to shut down all the time during a driving journey.

Having the AI be able to do this is considered an “edge” problem by most of the auto makers and tech firms. An edge problem is one that is not considered as core to the driving task of the self-driving car. It is a lesser important task. In this case, the power consumption is not yet perceived by most of the auto makers and tech firms as a significant problem, which is true because as mentioned earlier these self-driving cars aren’t in the hands of consumers as yet, and so the AI to do this power management is not yet being given much attention. Ultimately, it will need to get attention.

What makes this somewhat tricky too is that the devices themselves are being constantly updated and improved. Thus, if the AI was setup for device X and power management aspects about it, and then you replace that device X with a device Y, and maybe the device Y has different power consumption aspects than did device X, you need to then have the AI adjust accordingly.

Another important factor to keep in mind involves the interaction between the AI and the device.

The device might have its own power management capabilities. It should be informing the AI about what it wants to do about power management. The AI then can either agree or disagree. The two would need to negotiate as to what should take place. Meanwhile, the AI might instigate an indication to the device that it can go ahead and shift into low power mode. The device might have some reason why it thinks doing so is not sensible at that time, and so it might let the AI know that it disagrees with the command. This kind of give-and-take between the device and the AI will need to be fluid in nature and figured out as to how these kinds of discussions or arguments are to be settled.

If you include the fact that the self-driving car might have dozens upon dozens of electrical devices (hundreds, perhaps thousands), the AI could get preoccupied by the power management aspects. And, of course, the AI itself is using up processors and on-board memory to undertake the power management task, which again uses up more electrical power. On the balance, though, the AI should be saving more than enough power by the power management overall, such that the overhead of doing the power management is worthwhile for the power conservation efforts.

Generally, too, the top priority for the AI still involves driving the car, and so the power management is considered a secondary task in comparison to the driving task.  A self-driving car that has done a tremendous job of minimizing power consumption, but that runs into an obstruction because the AI was not paying attention to the driving task, well, that’s not an astute way to design an AI self-driving car. The AI for power management is a separate but integrated module for the overarching framework of the AI of the self-driving car. It would work relatively independently of the rest of the AI, though tap into the rest of the AI as part of the essentials of knowing what the self-driving car is up to, and therefore how and when the devices on the self-driving car need to be in full working status.

Electric cars have had battery packs of about 24 kWatt-hours that allowed for less than 100 hundred miles on a full-charge, and then they shifted toward 60 kWatt-hours capacity batteries allowing for maybe 200 miles ranges, and some are in the 100 kWatt-hours capacity allowing for just over a 300-mile range. James Watt, for whom the electrical consumption unit was named, would probably be flattered to know that for modern and futuristic AI self-driving cars that we are still thinking about electrical power aspects. We do indeed need to find ways to increase the power producing and power storing capabilities of AI self-driving cars and find ways to reduce power consuming aspects of the devices and systems, if we want to have true AI self-driving car.

We hopefully won’t get ourselves into an electrical power bind – imagine that while in an AI self-driving car, one that lacks AI for power management, the AI suddenly announces to you, a human occupant inside the self-driving car, it cannot go much further because it is about to run out of power.

Suppose further that this happens while the self-driving car is on the open highway going 80 miles per hour, and that the power will run out in the next five seconds and you are in the middle of the desert. Not good. In that case, I seriously doubt that allowing the AI to tap into our personal laptop to draw power (which, recall, I had done with my cell phone while at the airport), in order to try and make it to the next service station, will be a reassuring means of power management.  

The advent of AI self-driving cars and electrical power are integral and intertwined. No power, no viable AI self-driving car. With today’s limited trials of using AI self-driving cars on our roadways, the power aspects have not yet become widely known. Once the emergence expands, the electrical power lifeblood will help determine the viability of AI self-driving cars and can either inhibit or promote that viability. I vote that we focus on the power management aspects now, so that we’ll know beforehand that the viability will be promoted and not inhibited. Count me in for power management!

Copyright 2018 Dr. Lance Eliot

This content is originally posted on AI Trends.