Boxes-on-Wheels Becoming Jewel of the AI Self-Driving Cars Boon

Dr. Lance B. Eliot, AI Insider

Image for post
Image for post
Grocery clerk placing goods for driverless delivery to online shopper

Watch out, the rolling boxes are on their way. Many call them a box-on-wheels. That’s referring to the use of AI self-driving car technology to have a vehicle that would be driverless and would deliver goods to you or more.

At the Cybernetic AI Self-Driving Car Lab, we are developing AI software for self-driving cars, and are also including into our scope the use of AI systems for boxes-on-wheels. I offer next some salient aspects about the emerging niche of boxes-on-wheels.

Let’s start with a typical use case for a box-on-wheels.

You could potentially order your groceries online from your neighborhood grocer, and a little while later those groceries pull-up in front of your house as contained in a so-called box-on-wheels. You walk outside to the vehicle, enter a special PIN code or some other form of recognition, and inside are your groceries. You happily carry the grocery bags up to your apartment or house and do so without ever having to drive your car. The vehicle drives off to deliver groceries to others that have also made recent orders from that grocery store.

Notice that I mentioned that this is considered a use of AI self-driving car technology. It is not the same as what most people think of as an AI self-driving car per se. I say that because the vehicle itself does not necessarily need to look like a passenger car. A box-on-wheels can be a different shape and size than a normal passenger car, since it is not intending to carry humans. It is intended to carry goods.

If you ponder this aspect of carrying goods, you’d likely realize that it would be best to design the vehicle in a manner intended to carry goods rather than carrying humans.

Consider first what it’s like to carry goods inside a passenger car. I’m sure you’ve tried to pile your own grocery bags into the backseat of your car or maybe on the floor just ahead of the passenger front seat. The odds are that at some point you had those bags flop over and spill their contents. If you made a quick stop by hitting the brakes of the car, it could be that you’ve had groceries that littered throughout your car and maybe had broken glass from a smashed milk bottle as a result. Not good.

Don’t blame it on the passenger car! The passenger car is considered optimized to carry people. There are seats for people. There are armrests for people. There are areas for people to put their feet. Etc.

Thus, I think you can understand the great value and importance of developing a vehicle optimized for carrying goods, of which it is not bound to the design of a passenger carrying car. There are a wide variety of these designs all vying to see which will be the best, or at least become more enduring, as to meeting the needs of delivering goods.

The ones that go on the sidewalks need to especially be honed to cope with pedestrians and other aspects of driving on a sidewalk, plus there often is the need to get regulatory approval in a particular area to allow a motorized vehicle to go on sidewalks. Having such a vehicle on a sidewalk can be a dicey proposition. If you are wondering why even try, the notion is that it can more readily get to harder to reach places due to its smaller size and overall footprint, and in neighborhoods where they restrict the use of full sized cars it could potentially do the delivery (such as retirement communities), even perhaps right up to the door of someone’s adobe.

Some designers are going to the opposite extreme and considering boxes-on-wheels that are the size of a limo or larger. The logic is that you could store even more groceries or other goods in one that is larger in size. This could cut down on the number of trips needed to deliver some N number of goods to Y number of delivery spots.

The downside of the limo sized box-on-wheels is whether it can readily navigate the roads needed to do its delivery journey.

Indeed, let’s be clear that there is no one-size-fits-all solution here. There are arguments about which of the sizes will win out in the end of this evolving tryout of varying sizes and shapes of boxes-on-wheels.

Box on Wheels Free-for-All Today

That doesn’t though suggest that all of the variants being tried today will survive. I’m sure that many of the designs of today will either morph and be revised based on what seems to function well in the real-world, or some designs will be dropped entirely, or other new designs will emerge once we see what seems to work and what does not work.

Let’s consider an example of a variant being tried out today. Kroger, a major grocer, has teamed up with Nuro, an AI self-driving vehicle company, for the development of and testing of delivery vehicles that would carry groceries. The squat looking vehicle has various separated compartments to put groceries into. There are special doors that can be opened to then allow humans to access the compartments, presumably for the purposes of putting in groceries at the grocery store and then taking out the groceries when the vehicle reaches the consumer that bought the groceries.

This kind of design makes a lot of sense for the stated purpose of transporting groceries. You want to have separated compartments so that you could accommodate multiple separate orders.

There has to be also be some relatively easy way to access the compartments. Having a lockable door would be essential. The door has to swing or hinge in a manner that it would be simple to deal with and allow you access to the compartment readily and fully.

I mention these aspects because the conventional notion is that the box-on-wheels will most likely be unattended by a human.

If you had a human attendant that was inside the vehicle, they could presumably get out of the vehicle when it reaches your home, they could open the door to the compartment that contains your groceries, and they might either hand it to you or walk it up to your door. But, if the vehicle is unattended by a human, this means that the everyday person receiving the delivery is going to have to figure out how to open the compartment door, take out the groceries, and then close the compartment door.

This seems like a simple task, but do not underestimate the ability of humans to get confused at tasks that might seem simple on the surface, and also be sympathetic towards those that might have more limited physical capabilities and cannot readily perform those physical tasks. Presumably, the compartment doors will have an automated way to open and close, rather than you needing to physically push open and push closed the compartment doors (though, not all designs are using an automated door open/close scheme).

This does bring up some facets about these boxes on wheels that you need to consider.

First, there’s the aspect of having a human on-board versus not having a human on-board:

  • Human attendant
  • No human attendant

I’ve carefully phrased this to say human attendant. We don’t need to have a human driver in these vehicles since the AI is supposed to be doing the driving. This though does not imply that the vehicle has to be empty of a human being in it. You might want to have a human attendant in the vehicle. The human attendant would not need to know how to drive. Indeed, even if they knew how to drive, the vehicle would most likely have no provision for a human to drive it (there’d not be any pedals or steering wheel).

Why have a human attendant, you might ask? Aren’t we trying to take the human out of the equation by using the AI self-driving car technology? Well, you might want to have a human attendant for the purposes of attending to the vehicle when needed. For example, suppose the grocery carrying vehicle comes up to my house and parks at the curb in front of my house. Darned if I broke my leg in a skiing incident a few weeks ago and I cannot make my way out to the curb. Even if I could hobble to the curb, I certainly couldn’t carry the grocery bags back into the house and hobble at the same time.

The friendly attendant instead leaps out of the vehicle when it reaches my curb. They come up to my door, ring the doorbell, and provide me with my grocery bags.

Having a human attendant can be a handy “customer service” aspect. They can aid those getting a delivery, they can serve to showcase the humanness of the grocer, they can answer potential questions that the human recipient might have about the delivery, and so on.

Some say that the box-on-wheels should have a provision to include a human attendant, but then it would be up to the grocer to decide when to use human attendants or not. In other words, if the vehicle has no provision for a human attendant to ride on-board, the grocer then has no viable option to have the human attendant go along on the delivery.

So, why not go ahead and include a space in the box-on-wheels to accommodate a human attendant? We’re back to the question of how to best design the vehicle. If you need to include an area of the vehicle that accommodates a human attendant, you then are sacrificing some of the space that could otherwise be used for the storing of the groceries. You also need to consider what must the requirements of this space consist of.

This brings up another aspect about the box-on-wheels design, namely whether it can potentially do driving in a manner that would be beyond what a human would normally do. Assuming that the groceries are well secured and packaged into the compartments, the box-on-wheels could make sharp turns and brake suddenly, if it wanted or needed to do so. If there’s a human attendant on-board, those kinds of rapid maneuvers could harm the human, including perhaps some kind of whiplash or other injuries.

Also, if the box-on-wheels somehow crashes or gets into an accident, if you have a human attendant on-board there needs to be protective mechanisms for them such as air bags and seat belts, while otherwise the only danger is to the groceries.

Let’s shift attention now to the nature of the compartments that will be housing the goods.

Grocery Bags in Compartments of the Box-on-Wheels

For the delivery of groceries, it is so far assumed that the groceries will be placed into grocery bags and that in turn those grocery bags will be placed into the compartment of the box-on-wheels. The grocery bags are quite handy in that they are something we all accept as a means of grouping together our groceries. It has a familiar look to it.

For the moment, assume that the grocery store will indeed use a grocery bag for these purposes. You would want the grocery bag to be sturdy and not readily tear or fall apart — imagine if the box-on-wheels has no human attendant, arrives at the destination, and the human recipient pulls out their bag of groceries and it rips apart and all of their tangerines and other goods spill to the ground.

The odds are that the grocery store will use some kind of special cloth bag or equivalent which is durable and can safely hold the groceries and be transported. Likely the grocery store would brand the bags so that it is apparent they came from the XYZ grocery store.

You might say that the human recipient should put the grocery bags back into the box-on-wheels after emptying the grocery bags of their goods.

That’s a keen idea. But, you probably don’t want the box-on-wheels to be sitting at the curb while the human recipient goes into their home, takes the groceries about of the bags, and then comes out to the box-on-wheels to place the empty grocery bags into it.

This shifts our attention then to another important facet of the box-on-wheels, namely the use of the compartments.

We’ve concentrated so far herein on the approach of delivering goods to someone. That’s a one-way view of things.

Suppose though that the compartments were to be used for taking something from the person that received delivery goods. Or, maybe the compartment never had anything in it at all and arrived at the person’s home to pick-up something. The pick-up might be intended to then be delivered to the grocery store. Or, it could be that the pick-up is then delivered to someone else, like say Sam. As mentioned earlier, Sam lives some blocks away from you, and perhaps you have no easy means to send over something to him, and thus you use the grocery store box-on-wheels to do so.

The possibilities seem endless. They also raise concerns. Do you really want people to put things into the compartments of the box-on-wheels?

Some grocers are indicating that the recipients will not be allowed to put anything into the compartments. This is perhaps the safest rule, but it also opens the question of how to enforce it. A person might put something into a compartment anyway. They might try to trick the system into carrying something for them. Ways to try and prevent this include the use of sensors in the compartment to try and detect whether anything is in the compartment, such as by weight or by movement.

This does bring up an even more serious concern. There are some that are worried that these human unattended box-on-wheels could become a kind of joy ride for some. Imagine a teenager that “for fun” climbs into the compartment to go along for a ride.

Besides having sensors in the compartments, another possibility involves the use of cameras on the box-on-wheels.

There could be a camera inside each of the compartments, thus allowing for visual inspection of the compartment by someone remotely monitoring the box-on-wheels. You can think of this like the cameras these days that are in state-of-the-art refrigerators. Those cameras point inward into the refrigerator and you can while at work via your smartphone see what’s in your refrigerator (time to buy some groceries when the only thing left is a few cans of beer!).

We can enlarge the idea of using cameras and utilize the cameras on the box-of-wheels that are there for the AI self-driving car aspects. Thus, once the box-on-wheels comes to a stop at the curb, it might be handy to still watch and see what happens after stopping. Presumably, you could see that someone is trying to put a dog into a compartment. The box-on-wheels might be outfitted with speakers and a remote operator could tell the person to not put a dog into the compartment.

The use of remote operators raises added issues to the whole concept of the delivery of the goods. You are now adding labor into the process. How many remote operators do you need?

On the topic of remote operators, here’s another twist for you. Suppose the box-on-wheels arrives at the destination address. Turns out that the curb is painted red and presumably the box-on-wheels cannot legally stop there. The street is jam packed with parked cars. There is no place to come to a legal stop. What should the AI of the box-on-wheels do?

We all know that a human driver would likely park temporarily at the red curb or might double-park the delivery vehicle. But, do we want the AI to act in an illegal manner?

Some believe that with a remote human operator you might be able to deal with this parking issue by having the remote operator decide what to do. The remote operator, using the cameras of the AI self-driving vehicle, might be able to see and discern where to park the box-on-wheels.

Would the remote operator directly control the vehicle? Some say yes, but if that’s the case then the question arises as to whether they need to be licensed to drive and opens another can of worms. Some therefore would say no, and that all the remote operator can do is make suggestions to the AI of where to park (“move over to that space two cars ahead”).

Here’s another facet to consider. How long will the box-on-wheels be at a stopped position and allow for the removal of the goods?

From the grocer viewpoint, you would want the stopped time to be the shortest possible. For every minute that the box-on-wheels sits at the curb and is waiting for the delivery to be completed, it is using up time to get to the next destination. Those further along in the delivery cycle are all waiting eagerly (or anxiously) for the box-on-wheels to get to them.

Suppose a person comes out to the box-on-wheels, opens the compartment designated for their delivery, and for whatever reason rummages around in the grocery bag, maybe doing an inspection to make sure the bag contains what they ordered. They decide to then slowly remove the bag and slowly walk up to their home and slowly put the bag inside the home. Meanwhile, they have four other bags yet to go that are sitting in the compartment. They walk out slowly to get the next bag. And so on.

What the AI is Doing for the Box-on-Wheels

This takes us to a crucial part of this discussion, namely what the AI is doing for the box-on-wheels.

The usual aspects of the AI involve the driving of the vehicle. Once the groceries are loaded into the compartments, it is given a “proceed ahead” indication at the grocery store. It then drives the vehicle to each of the destinations. At each destination, it allows for the compartments to be opened and then closed and needs to ascertain when to continue along on the journey.

In our Lab, we are adding to the AI by including the aspects about the delivery aspects, which a normal AI self-driving car has no concern about and no provision for. In essence, the typical AI self-driving car will remain dormant during the time that the box-on-wheels is stopped. All it cares about is when it should start driving again. There is no provision to communicate any further or take any other actions until it is cleared to continue driving.

Ideally, the AI would be aiding the delivery moment. This includes the detection of a human or humans that are coming to the box-on-wheels to pick-up the goods.

You can also add into this list of tasks the potential arduous parking aspects. Have you had a human driver that came to delivery something and you met them at the curb and told them to go ahead and park up ahead at the corner? I’m sure you have. The AI of the self-driving car can potentially interact with human(s) during the parking stage to help ascertain a place to pull over for the disgorging of the goods.

One other aspect about the box-on-wheels involves the kinds of goods that it is intended to carry. If there are frozen food items, you’d presumably want the compartment to be refrigerated so that the frozen items would not melt during the journey. You cannot know for sure the length of time to undertake the deliveries, given the vagaries of traffic and also the vagaries of the time during the delivery moment, and thus you can’t just hope that the food will remain in proper shape during the journey. Using conventional air conditioning might not be enough to keep the food at the proper temperature.

You might be tempted to say that only certain kinds of groceries can be delivered via the box-on-wheels. Yes, you could make that constraint, but you’ve now made for a dilemma for the customer. If I cannot get my frozen fish and frozen pizza from the grocery store, I’ll need to make my own trip there. If I am going to make my own trip there, why futz with the box-on-wheels delivery service?

This also logically takes us to another consideration about boxes-on-wheels. If true AI self-driving cars become prevalent, would I even need to use a box-on-wheels? In other words, if I owned a true AI self-driving car, which is considered a Level 5, I could just tell it to go to the grocery store and pick-up my goods. No need to use the box-on-wheels.

The counter-arguments are that not everyone is necessarily going to have a true AI self-driving car, and will be relying instead on using other people’s AI self-driving cars to get around.

I’ve mentioned the idea of keeping food cold, but there’s also the potential desire of keeping food hot. Perhaps from the grocery store, I order some cooked chicken that the grocery store is selling at their in-store buffet. I’d want the chicken to remain hot during the journey to me. Thus, the compartments might need refrigeration and they might also need some form of heating capability.

This also brings up the recent efforts by Domino’s Pizza and by Pizza Hut to consider using AI self-driving vehicles to delivery pizza. Pizza Hut has teamed-up with Toyota and opted to try and get ovens closer to the door of the customer. These kinds of boxes-on-wheels are potentially going to either be keeping the pizza warmed-up or could possibly even be cooking the pizza during the journey of performing the delivery.

They still face the same issue about having the customer come out to the box-on-wheels to get the goods.

In the case of Domino’s, they teamed-up with Ford and did an interesting experiment. They did a month long test in Ann Arbor, Michigan and had a human driver that was instructed to not interact with the customers at all. The vehicle contained the pizza that was to be delivered, placed in the backseat area and reachable to the customer by the vehicle rolling down the back window, and it was a pretense that there wasn’t any human to interact with, thus, similar to picking up a pizza from an AI self-driving vehicle.

Some of the customers indicated they liked the idea of not having to interact with a human attendant.

As mentioned earlier herein, they discovered the parking problem issue of knowing where to best stop the vehicle to accommodate the customer (recall that they were pretending that the human driver could not interact with the customer — this is somewhat the case for today’s AI, but in the future should not be).

Conclusion

Boxes on wheels. Or, should that be boxes on wheels!

There’s little doubt that in spite of the potential emergence of AI self-driving cars, we’ll still need some kind of specialized vehicles to do deliveries. An AI self-driving car that is optimized for carrying passengers will not be as optimized for carrying goods.

We are still a long way away from having true AI self-driving cars. And, they will not become prevalent overnight.

Thus, there is definitely an opportunity for the advent of boxes-on-wheels. There are many opportunities available in this niche and it provides an exciting source of challenges. The phrase “box on wheels” sounds perhaps demeaning to some, but it has the potential for being a money-making way to undertake deliveries, can reduce the cost of delivery, can aid society by enabling delivery, and is going in the direction of a society that wants to order online and have items delivered to them. Two cheers for box-on-wheels.

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 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