Self Driving cars face a long way to become the future of transport

Driverless cars are the future of transport. According to some reports, 10 million vehicles will come into circulation by 2020. They will transport passengers from one place to another, such as driverless taxis. They will transport packages and raw materials from one city to another. And they will deliver groceries, meals and packages to houses and apartments throughout the country.

But despite all the optimism surrounding self-contained cars, there is an equal amount of skepticism and concern.

In a couple of polls published by the American Association Association in January and Gallup in May, 63% of people reported being afraid to travel in a completely autonomous vehicle and more than half said they would never choose to travel in one.

Those feelings have not changed much. Three separate studies this summer - the Brookings Institution, the HNTB think tank and the Autostrade and Automotive Safety Advocates (AHAS) - have found that most people are not convinced of the safety of driverless cars. Over 60 percent said they were not "inclined" to travel in driverless cars, nearly 70 percent expressed "concerns" to share the road with them, and 59 percent expected that driverless cars "no longer existed sure "that controlled human machines.

This is despite the fact that about 94% of motor vehicle accidents are caused by human error and that in 2016 the three main causes of traffic fatalities were distracted by driving, drunken driving and driving excessively. of speed. According to the National Safety Council, the odds that Americans have of dying in a car crash are one in 114. In 2016, vehicle deaths caused 40,000 deaths.

What does it take to convince a skeptical audience that the stand-alone cars are ready for the road? In summary: many other tests.

A typical stand-alone drive configuration consists of three main types of sensors: distance measuring lids, color cameras and radar. On-board computers combine data flows and apply intelligence, which translates into perceptions. The goal is that the system distinguishes a pedestrian from a cyclist, a four-way crossing of a roundabout, dogs for small children and millions of other objects, types of roads and riding styles.

To "teach" systems to navigate safely on public roads and city streets, Waymo, Uber and others recruit safety drivers for driverless racing, during which they take note of unexpected errors and behavior. But since there are no two equal units, they subject the systems to millions of virtual, computerized and highly customizable tests.

Waymo calls his Carcraft simulation platform, after the famous World of Warcraft videogame series. At one point, over 25,000 virtual cars without drivers travel digitized roads from Phoenix, Mountain View, Austin and other cities where Waymo has deployed driverless cars, as well as test tracks. To date, the fleet has driven 5 billion simulated miles.

It's a start, but it's far from the baseline that some researchers believe the cars they drive must reach. The RAND Corporation, for example, estimates that they will have to accumulate 11 billion miles before having reliable statistics on their safety. To make a comparison, the 20 companies that tested driverless cars in California have together registered just over 1 million miles in two years.

Read More :

Sanjoy Baruah, a professor of computer science and engineering at the University of Washington in St. Louis, said the question of whether autonomous car technology is ready for deployment is "very questionable" at the moment.

"The more data they collect, the better [they will be], but we have to take a step-by-step approach," he said. "It is asking too much for an accident rate of zero percent, but it is true that we do not have enough information about how traffic and rare weather events, for example, should be modeled or understood".

The company that is perhaps the most advanced - Waymo - reached 8 million real miles in July and in a February report stated that its cars can travel only 5,600 miles between disconnections. In other words, if you were to give someone with a 10-mile day trip one of his driverless cars, that person should take control once a year.

Cruise, the General Motors driverless car division, registered 131,000 miles in California in early 2018 and managed to reduce the disconnections from once every 35 miles last year to once every 1,250 miles.

All other companies reported a disconnection rate no higher than once every 160 miles.

Chuck Price, CEO of TuSimple, a three-year self-employed trucking company that has 20 autonomous trucks operating in Texas and China, said metrics like miles are not all that is believed. But he admitted that experience, both simulated and real, is one of the best ways to reduce error rates.

"Operating on a motorway is almost something social," he said. "Systems must understand fusion and lane change and mimic that kind of human behavior as much as possible."


Autonomous execution towards the market can have tragic consequences.

In May, a Uber prototype, a self-powered Volvo XC90, fatally clashed with a pedestrian in Tempe, Arizona. And in March, a Tesla Model X in a semi-autonomous Autopilot mode crashed into a concrete barrier, killing Wei Huang, a 38-year-old Apple software engineer.

Mitigating factors contributed to both incidents, Uber reportedly disabled Volvo's integrated collision avoidance system, and Tesla said Huang ignored the automatic warnings previously in the unit. But at least one accident involving a driverless car is due to a system failure: an S model that collided with a truck in Florida in May 2016. The National Administration for Highway Traffic Safety has declared the circumstances of the 'functionality incident' exceeded the performance of [Autopilot]. "

Even basic assistance technologies have proved unreliable. In August, the Insurance Institute for Road Safety warned in a report that the systems of automatic breakage, adaptive speed control and active lane in Tesla, BMW, Mercedes-Benz and Volvo cars "have committed errors that could be fatal without a driver ". intervention. "Some of its faults included drifting in lanes, hitting stationary objects even when they were within the sensor beam and unexpected slowdowns.

Jack Weast, chief architect of the independent guide at Intel, said it was due to the lack of training, sensors and redundant systems.

"One of the challenges of developing autonomous driving systems is ... ensuring that they accurately perceive the environment," he said. "Capture hundreds of millions of video miles, feed them through an algorithm and test the accuracy of the system." As a company, you should think, "How can I show a consumer that my vehicle will always make safe decisions and never have an accident?"

To complicate matters, most of the autonomous systems are "a black box," said Austin Russell, CEO of Linar, Luminar's supplier.

"When rubber hits the road, consumers will be more comfortable if they show that self-employed vehicles are safer than themselves as drivers, and no one has solved it ... [that]," he said. "We are not remotely close to the possibility of being truly autonomous in different conditions," he said.


The Governor's Highway Safety Association released a written report from former NHTSA official Dr. Jim Hedlund on Wednesday, summarizing the questions that car manufacturers, start-ups and original equipment manufacturers face before the cars they drive get on the road in large quantities.

He suggested that states seek to encourage autonomous "responsible" testing and deployment of vehicles while "protecting public safety" and that legislators "review all traffic laws" for changes that might be needed to adapt to those tests.

Read More:

Currently, 21 states: Alabama, Arkansas, California, Colorado, Connecticut, Florida, Georgia, Illinois, Indiana, Louisiana, Michigan, New York, North Carolina, North Dakota, Pennsylvania, South Carolina, Tennessee, Texas, Utah, Virginia, Vermont - and the District of Columbia has passed laws governing the deployment and testing of driverless cars. And the governors of 10 states: Arizona, Delaware, Hawaii, Idaho, Maine, Massachusetts, Minnesota, Ohio, Washington and Wisconsin, have issued executive orders related to them.

Unfortunately, the various pieces of legislation are not consistent with each other. As Brookings notes, state laws explain at least three different definitions for "vehicle operator": in Texas, it is the "physical person" who travels in the car, while in California telemarketing also falls within this definition.

The NHTSA has published guidelines for autonomous vehicles in September in the Vision for Safety 2.0 guideline: it traces its guidelines for 2016. It provides a model for lawmakers and officials of state highways to follow, but also clarifies that manufacturers Cars and self-directed startups do not need to wait for federal legislation to test or implement their systems.

Robert Brown, director of TuSimple's public affairs, believes it is a good start.

"A 50-state solution is the perfect and ideal solution for us, but almost all states are in favor of deployment," he said. "We are working closely with them, it was a collaborative effort".

But Baruah claims he does not go far enough. Supports a rigorous investigative system similar to the Federal Aviation Administration's flight licenses.

"The government could grant self-taught company certificates through a very rigorous testing procedure, which should demonstrate the safety of their systems."

The only viable alternative in the short term is self-regulation, he said.

"As long as all companies agree to comply with a range of industrial practices, it could be a reasonable alternative [to government intervention]," he said. "There is a social pressure to avoid the deployment of an insecure system, no one will buy their car if it perceives that it is going to hurt them, there is a lot to lose".

Most companies do not wait for Uncle Sam to make a move.

Waymo will pilot a driverless car service in Phoenix later this year and Cruise has announced plans to start a taxi service in San Francisco in 2019. In the meantime, startups like, NuTonomy and Optimus Ride have been deployed. driverless prototypes in Phoenix. Frisco, Texas and Boston.

"The security, reliability and experience that technology will allow are the cornerstone of trust," said Ford Autonomous Vehicles CEO Sherif Marakby in a letter to Transport Secretary Elaine Chao at the start of the quest. 'year. "The development of autonomous vehicles is not just about technology, but about the trust of our customers and those cities and companies that will eventually use it".

Considering the challenges that lie ahead for driverless cars, it is likely that this trust is far away.

Post a Comment