Tag Archives: av

Rapid Evolution of Autonomous Vehicles

The 2008 animated Pixar movie “Wall-E” follows the refuse-based adventures of a sentient, autonomous trash compactor whose primary function was to clean an abandoned city on a now-deserted planet Earth, long ago having been abandoned by humanity. The movie highlights some of the issues that would likely occur from human beings’ over-reliance on an automated lifestyle – issues such as waste management, obesity and human environmental impact, to name a few. “Wall-E” is set hundreds of years in the future, but some of those issues ostensibly exist in the world we inhabit today.

The transportation sector around the globe is a multitrillion-dollar industry. There’s money and mistakes to be made. While we are probably a ways off from sentient automobiles, the age of vehicle autonomy is well upon us. Every week, another company releases some update, patch or application that nudges autonomous tech in a new direction.

There have been some setbacks – name me a sector that doesn’t have any – but cars that are less reliant on humans are here to stay. This is almost universally viewed as positive, with many examples given to support this position, such as:

  • Fewer accidents.
  • A move away from owned to rented vehicles, lessening the need for parking garages.
  • A productivity increase during commuting time.
  • A reduction in traffic congestion.

There are many more, but the age of connectivity comes with risks. One must exercise caution with any kind of new technology. What happens when things go wrong? Computers malfunction sometimes; we’re all familiar with Windows’ blue screen of death.

See also: Autonomous Vehicles: ‘The Trolley Problem’  

You are turning over your most precious commodity – your family – to a computer. And if that computer fails when you are trusting it not to – let’s say when it is in full autonomous mode – how will that fail affect things? In what manner will it fail? It will likely fail however the lowest-bidding subcontractor designed it to fail.

Even if it does not fail, a computer still needs to be told what to do, at least initially. Computers can learn things and eventually make better iterative decisions based on this learning, but what do you tell a computer it should do when faced with a myriad of input data?

Autonomous vehicles (ones that fly) have been around a long time. Most commercial airliners are autonomously piloted more than 90% of the time. Aircraft, along with the routes they take, are heavily regulated. They essentially all report in to the same system around the world. There is a reason all pilots around the world must communicate in English. There has to be one universal language to avoid miscommunication and errors.

Autonomous automobiles have none of that. There is no central control, no clearing house and no standardization, to the extent that even the levels of autonomy differ by manufacturer. They can, though, roughly be classified in the following manner:

Level 0 — Nothing

The baseline since Gottlieb Daimler traded horse power for horsepower. Level zero applies to all vehicles that rely solely on humans to dictate driving actions. That is my car, and almost every car that has come before it. At best it has cruise control, but it is the “dumb” version that will crash you into a wall if you let it. Example: my 2009 Honda Ridgeline truck.

Level 1 — Driver Assistance

What does this level offer us? Some automation, but not much. For level one, you are looking at adaptive cruise control or lane departure tech to come as standard on your vehicle. While the human driver still supervises everything, the vehicle is capable of some decisions on its own. Example: your eco-friendly neighbor’s 2016 Toyota Prius.

Level 2 — Partial Automation

We get a step up from driver assistance in level two. This combines multiple automated functions such as lane assist, automatic braking and adaptive cruise control to ensure they work in a smooth, coordinated fashion. Anticipating traffic signal changes, lane changes and scanning for hazards are still the domain of the driver. Example: the Audi your boss drives that has Traffic Jam Assist as standard.

Level 3 — Conditional Automation

A car running level three automation can take full control of the vehicle during certain parts of a journey under certain conditions and within certain parameters. The vehicle will, however, turn control back over to the human driver when it encounters a situation it cannot handle or when it cannot interpret input data. The onus is, therefore, on the driver to stay alert because the vehicle may prompt the driver to intervene at any moment. The incident in Tempe, AZ, in March 2018, involving a pedestrian fatality involved a vehicle running level three autonomy. Example: Tesla’s Autopilot.

Level 4 — High Automation

An auto at level four automation does not require a human to ride along during certain journeys, subject to geographic and road-type limitations. These are currently being tested, and we should see them within the next 18 months. Think Amazon last mile and pizza delivery vehicles. Example: Johnny Cab, from the original “Total Recall.”

Level 5 — Full Automation

At level five, absent inputting the destination, which will probably be done via your phone beforehand, there is no driver involvement. You will enter the vehicle, turn on your movie or laptop and that is it until you reach your destination. Example: KITT from “Knight Rider.”

Level 6 — Beyond Full Automation

Well, there is no level six – at least yet. What would level six look like if it did exist? A teleporter? Something that transports you from your bedroom, via the bathroom and kitchen, straight to the office? A flying car? We have returned to Wall-E territory. Example: The Jetsons’ Aerocar.

Technology in vehicles is designed to assist us and make us safer. For good reason, a few of the car and tech companies working on autonomous driving have said they do not want to release anything below level four. Either force people to drive, or let the machine do all of the work. Partial implementation runs the risk of scaring people away from the technology. The more reliant you are on tech, the tougher it is when you do not have it. When, in an instant, the computer turns full control back to you because its inputs are confusing, are you ready?

See also: Autonomous Vehicles: Truly Imminent?  

What does the future look like? We should expect a reduction in the frequency of accidents, but, given the complicated nature of what is now hidden under a fender, accidents will likely cost more (increased severity).

Software updates can be problematic. They do not work well on airplanes, for example. You would not release beta software for an airliner. A recent over-the-air software update by Tesla reportedly disabled the autopilot system. Too much automation in the cockpit or car, and things can go bad when the computer gets an input it does not understand.

Walt Disney promised us self-driving cars back in 1958. They are here – somewhat – but 60 years is a long time to wait in line. As a juxtaposition to that, with robotaxis already hitting our roads, the future has arrived more quickly than most people anticipated.

15 Hurdles to Scaling for Driverless (Part 2)

Will driverless cars (AVs, for autonomous vehicles) live up to the revolutionary potential imagined by many, including me? In part one of this series, I asked whether AVs might develop like the Segway personal transporter and be relegated to narrow niche applications. To avoid going the way of the Segway, AV developers must overcome significant hurdles to scaling, trust, market viability and managing secondary effects.

In this article, I outline the challenges to scaling. Building and proving an AV is a big first step. Scaling AVs into industrial size and strength business operations delivering transportation as a service (TaaS) is an even bigger step. Here are seven giant hurdles related to scaling:

1. Mass production. Hand crafting or retrofitting a few thousand cars with AV technology is good enough for development and testing. Industrialization will require producing hundreds of thousands of cars at scale. But, as Tesla learned the hard way, building cars at scale is more complicated than it looks.

2. Electric charging infrastructure. Almost all AV efforts are being developed on top of electric vehicle (EV) platforms. Before EV fleets can operate at scale in any market, a whole new electric charging infrastructure must be built. This will take time and lots of money.

3. Mapping. The industrialization of the detailed, high-definition (HD) maps on which AVs depend limits where AVs can operate. Even though the cars are loaded with sensors, cameras and software, they need up-to-date maps to figure out where they are and what to do.

See also: Driverless Cars and the ’90-90 Rule’  

4. Fleet management and operations. Industrialization will require flawless maintenance and efficient operation of tens of thousands of AVs widely distributed across large metropolitan service areas. Doing so will entail much more than cleaning windows and vacuuming carpets. It will entail maintaining complex computers on wheels. It will require complex predictive analytics to recharge, dispatch and load balance in response to spiky customer demand. Both public safety and business viability depend on this.

5. Customer service and experience. AV TaaS services are like a hospitality business built on a fleet of mobile hotel rooms with no on-property staff. Even the shortest trip can become arbitrarily messy and unruly—especially because there will be no human supervision in the car. Acceptable service and experience will have to extend to non-customers, as well, because AVs must interact with pedestrians, cyclists, other drivers, emergency personnel, other companies’ AVs and a host of other actors outside the car.

6. Security. Computer security is a challenging issue, in general, and networked armadas of computers on wheels will be attractive hacking targets. There are physical security issues, too. Physical security involving disgruntled drivers and bystanders, pranksters, thugs and others could also create security issues for both passengers and the public at large.

7. Rapid localization. How much of what Waymo and others are learning on the well-marked, well-lit, well-laid-out and relatively polite streets of Phoenix is transferable to the not-so-polite paved cow paths and roundabouts of Boston or the congested, pedestrian-filled city centers of New York, Paris and Beijing? Much—but not all. That is why every developer tests in multiple regions, to understand the peculiarities of local infrastructure, weather, cultural norms, etc. How fast and how well such localization can be done is another hurdle to scaling.

See also: How to Adapt to Driverless Cars  

In part three of this series, I’ll explore the challenges to market acceptance. I will discuss eight industrialization hurdles that deal with trust, market viability and secondary effects.

15 Hurdles to Scaling for Driverless Cars

Will the future of driverless cars rhyme with the history of the Segway? The Segway personal transporter was also predicted to revolutionize transportation. Steve Jobs gushed that cities would be redesigned around the device. John Doerr said it would be bigger than the internet. The Segway worked technically but never lived up to its backers’ outsized hopes for market impact. Instead, the Segway was relegated to narrow market niches, like ferrying security guards, warehouse workers and sightseeing tours.

One could well imagine such a fate for driverless cars (a.k.a. AVs, for autonomous vehicles). The technology could work brilliantly and yet get relegated to narrow market niches, like predefined shuttle routes and slow-moving delivery drones.  Some narrow applications, like interstate highway portions of long-haul trucking, could be extremely valuable but nowhere near the atmospheric potential imagined by many—include me, as I described, for example, in “Google’s Driverless Car Is Worth Trillions.”

For AVs to revolutionize transportation, they must reach a high level of industrialization and adoption. They must enable, as a first step, robust, relatively inexpensive Uber-like services in urban and suburban areas. (The industry is coalescing around calling these types of services “transportation as a service,” or TaaS.) In the longer term, AVs must be robust enough to allow for personal ownership and challenge the pervasiveness of personally owned, human-driven cars.

See also: Where Are Driverless Cars Taking Industry?  

This disruptive potential (and therefore enormous value) is motivating hundreds of companies around the world, including some of the biggest and wealthiest, such as Alphabet, Apple, General Motors, Ford, Toyota and SoftBank, to invest many billions of dollars into developing AVs. The work is progressing, with some companies (and regulators) believing that their AVs are “good enough” for pilot testing of commercial AV TaaS services with real customers on public roads in multiple markets, including SingaporePhoenix and Quangzhou.

Will AVs turn out to be revolutionary? What factors might cause them to go the way of the Segway—and derail the hopes (and enormous investments) of those chasing after the bigger prize?

Getting AVs to work well enough is, of course, a non-negotiable prerequisite for future success. It is absolutely necessary but far from sufficient.

In this three-part series, I look beyond the questions of technical feasibility to explore other significant hurdles to the industrialization of AVs. These hurdles fall into four categories: scaling, trust, market viability and secondary effects.

Scaling. Building and proving an AV is a big first step. Scaling it into a fleet-based TaaS business operation is an even bigger step. Here are seven giant hurdles to industrialization related to scaling:

  1. Mass production
  2. Electric charging infrastructure
  3. Mapping
  4. Fleet management and operations
  5. Customer service and experience
  6. Security
  7. Rapid localization

Trust. It is not enough for developers and manufacturers to believe their AVs are good enough for widespread use, they must convince others. To do so, they must overcome three huge hurdles.

  1. Independent verification and validation
  2. Standardization and regulation
  3. Public acceptance

Market Viability. The next three hurdles deal with whether AV-enabled business models work in the short term and the long term, both in beating the competition and other opponents.

  1. Business viability
  2. Stakeholder resistance
  3. Private ownership

See also: Suddenly, Driverless Cars Hit Bumps  

Secondary Effects. We shape our AVs, and afterward our AVs reshape us, to paraphrase Winston Churchill. There will be much to love about the successful industrialization of driverless cars. But, as always is the case with large technology change, there could be huge negative secondary effects. Several possible negative consequences are already foreseeable and raising concern. They represent significant hurdles to industrialization unless successfully anticipated and ameliorated.

  1. Congestion
  2. Job loss

I’ll sketch out these hurdles in two more parts to come.