History and Future of Self-Driving Cars in One Book



Image course: https://www.pinterest.com/pin/264938390560462730/

“Autonomy: The Quest to Build the Driverless Car” is a story of the self-driving industry. Lawrence D. Burns, the author of the book, is a former General Motors executive and an advisor to Waymo.

The book is full of fascinating stories about the evolution of a crazy idea of self-driving cars from a handful of small teams to the current state of the industry. Burns was a first-hand participant in many events he describes.

“Talks at Google” with Lawrence Burns

He starts the story from DARPA Grand Challenge, a competition for robocars, which would be held on March 13, 2004 in the Mojave Desert in California. That competition was followed by another one in 2005, and “Urban Challenge” in 2006. It was interesting to read about all the technical challenges encountered by two most prominent teams: the Carnegie Melon University’s “Red Team” led by William “Red” Whittaker and Chris Urmson and Stanford University’s team led by Sebastian Thrun.

He starts the story from DARPA Grand Challenge, a competition for robocars, which would be held on March 13, 2004 in the Mojave Desert in California. That competition was followed by another one in 2005, and “Urban Challenge” in 2006. It was interesting to read about all the technical challenges encountered by two most prominent teams: the Carnegie Melon University’s “Red Team” led by William “Red” Whittaker and Chris Urmson and Stanford University’s team led by Sebastian Thrun.


The Urban Challenge was crazy hard. It was not enough to just follow a precomputed path across the desert. The cars should drive on urban roads shared with other cars and comply with California driving rules! Tasks like negotiating who would go first on a 4-way stop intersection, and unprotected left turns with oncoming traffic were quite hard for autonomous cars. Today, 12 years after the competition, left turns are still reportedly hard for the self-driving cars.

Despite the hard technical problems, a few teams were able to complete the course and demonstrate that autonomous vehicles are feasible. Surprisingly, no car manufacturers jumped on the opportunity.


Enter Google. Larry Page and Sergey Brin knew better about capabilities of technology. In 2008, just two years after Urban Challenge, Google started the self-driving car project. Sebastian Thrun, the leader of the team, hired the best engineers from the teams participated in DARPA challenges. You might recognize the names: Chris Urmson, Bryan Salesky, Anthony Lewandowski, Mike Montemerlo.

Brooks touches a topic of cultural difference between Detroit car manufacturers and Silicon Valley software companies. One very important innovation of Google self-driving car team was testing on public roads.

“The company’s lawyers investigated the state’s driving regulations and discovered that, at that point, nothing explicitly prevented a computer from operating an automobile. The key part of the legislation required a human to be in the driver’s seat — but didn’t specify that the driver had to actually be controlling the vehicle.”

This approach to testing was unthinkable for car companies in Detroit.


I was surprised by the economic models of shared fleet operations in urban areas. The book describes calculation by Columbia’s Earth Institute. They used Ann Harbor for a model. It’s a city with 200,000 privately owned vehicles that make 740,000 trips per day. They focused on 120,000 vehicles that were driven less than 70 miles a day, responsible for 528,000 trips per day.

“Ann Arbor required just 13,000 vehicles to service the city’s internal trips at average periods. To provide nearly instantaneous access for trips in and around Ann Arbor, even during rush hours, the city would require a shared fleet of just 18,000 vehicles.”

“We decided to study several other cities: Salt Lake City, Utah; Rochester, New York; Columbus, Ohio; Austin, Texas; and Sacramento, California. In every case, a shared fleet size equal to about 15 percent of the number of cars owned in these cities would provide responsive and efficient mobility service. Upon further analysis, we concluded that in communities with population densities in excess of around 750 people per square mile, our envisioned mobility system based on autonomous shared vehicles would perform quite well. This density criterion covers most towns and cities in the U.S. In addition, we were encouraged by how robust our results were to changes in key modeling assumptions. For example, you’d need to serve only 10 percent of the trips in an urban area to simultaneously attain high fleet utilization, low empty miles and fast response times. This means you could launch a commercial business with a relatively modest market share of trips.”


I enjoyed seeing the estimates of the potential impact of autonomous cars on the economy. The current transportation system is highly inefficient. American cars sit unused for 95% of the time.

“Less than 30 percent of the energy from the gasoline you put in your car is used to move it down the road.”

“Because typical vehicles weigh around 3,000 pounds and typical people weigh around 150 pounds, only about 5 percent of the gasoline energy translated into motion is used to move the driver, which amounts to just 1.5 percent of the total energy in gasoline.”

People buy cars for rare use cases, e.g., Americans increasingly prefer buying SUV, but they need capabilities of SUV just once or twice a year. 2-person small vehicle would suffice for most of the trips.

Three trends are converging: autonomous driving, electric vehicles, and shared fleets. It’s a huge disruption.

“The expert math modeler Bill Jordan and I calculated in 2011 that the deployment of such an integrated system could reduce the annual costs of automobile travel in the U.S. alone by $4 trillion — about the same amount as the entire budget of the federal government. More to the point, our research suggested that driverless electric vehicles tailor-designed for shared transportation service in U.S. cities could reduce the out-of-pocket and time costs of conventional automobile travel by more than 80 percent (from $1.50 per mile to $0.25 per mile) — while providing safer and more convenient mobility.”

Self-driving makes ride-hailing services more interesting because labor costs are reduced dramatically. Cars were a product, now they become a service.

Car manufacturers were optimizing for cost of a car at the dealership. Cost per mile metric is a more interesting metric for fleets. And when you optimize cost per mile, electric vehicles suddenly look really good. Besides, electric cars get cheaper every year.

Cars become simpler; electric vehicles have an order of magnitude fewer parts than gasoline-fueled cars. The expertise of car manufacturers used to be in internal combustion engines, the most complex part of a gasoline-fueled car. In electric vehicles, it’s gone. The complexity of the vehicles shifts from integration/manufacturing to software development. Traditional car manufacturers need to relearn how to design and build cars.


When the cost of autonomous shared vehicles gets lower than the cost of a personal car, we’ll see large changes in transportation, car manufacturing business, car design, urban design and more. Parking gets impacted, dealers get impacted, real estate gets impacted. The entire economy gets impacted.

Switch to electric vehicles may cause even oil prices to drop.

“America’s vehicles consumed about 180 billion gallons of fuel a year — which makes up about half of total U.S. oil consumption.”

“In fact, the oil industry could be looking at a future not unlike what the coal industry has been experiencing over the past several decades.”


Ready to read the book? I bet you’ll enjoy it!

Source: Deep Learning on Medium