Advanced Computing in the Age of AI | Friday, March 29, 2024

Autonomous Vehicle Future Drives Closer to the Present 

<img style="float: left;" src="http://media2.hpcwire.com/dmr/DARPAAV.jpg" alt="" width="95" height="58" border="0" />Less than a decade after the U.S. Defense Advanced Research Project Agency (DARPA) gave the greenlight on autonomous vehicle research, Carnegie Mellon University (CMU) has driven self-driving cars closer to reality, demonstrating that it’s not a question of if driverless cars will grace our highways and backroads, but when.

Less than a decade after the U.S. Defense Advanced Research Project Agency (DARPA) gave the greenlight on autonomous vehicle research, Carnegie Mellon University (CMU) has driven self-driving cars closer to reality. With U.S. Rep. Bill Shuster (R-Pa.) and Barry Schoch, the state’s transportation department chief aboard, an autonomous Cadillac made a trip from Cranberry, Pa., to the Pittsburgh International Airport all on its own, demonstrating that it’s not a question of if driverless cars will grace our highways and backroads, but when.

In addition to eliminating the 93 percent of accidents caused by human error, and cutting down drastically by rush-hour gridlock by making more efficient use of crowded streets and highways, a societal transition away from individually-driven cars toward a collective of autonomous vehicles would cut down on the stressful, often draining task of driving to and from work every day.

“Now is the time for us all to be looking at vehicles on the road the same way we look at smartphones, laptops and tablets—as pieces of a much bigger, richer network,” says Bill Ford Jr., executive chairman of Ford.

But as inevitable as a highway full of self-driving cars now seems, it does not come without nearly a decade’s worth of trial and error.

In 2004, for instance, DARPA’s first open autonomous vehicle competition was set over a 142-mile course in California’s Mojave Desert. But of the 15 cars entered, not a single one finished, with the most successful only traveling seven miles before getting caught on a rock.

But the following year progress became clear, with four of 23 vehicles completing the competition course that featured 100 sharp turns and a home stretch through a meandering mountain pass.

Two years after that, a field of 11 cars made their way through a 60-mile course, where 30 professional drivers behind safety cages replicated city traffic for the vehicles to navigate through. Although the top speed across the field was a mere 14 mph, one entry, Oshosh Trucks’ 30,000-pound Terramax had its plug pulled to keep it from running into a building.

“To me DARPA deserves a huge amount of credit,” says Raj Rajkumar, the leader of CMU’s autonomous vehicle program. “The 2007 Urban Challenge was a watershed moment because it proved that autonomous driving is not science fiction anymore. It is only a question of ‘when,’ not ‘if.’”

And when you consider that these technologies now guide NASA’s Mars Curiosity rover, it’s not hard to see where Rajkumar is coming from.

On the home front Nissan CEO Carlos Ghosn predicts that driverless cars will be in showrooms by 2020, while Google cofounder Sergey Brin expects to see his team fielding them within five years.

But in the meantime, the greatest changes consumers are seeing aren’t in prototypes, but in the cars they already see on the road. In addition to cars that can parallel park themselves or warn drivers when they sense that the car is drifting into another lane, some vehicles are integrating a smarter form of cruise control where the car is instructed to maintain distance between in and the car in front of it, rather than simply maintaining speed. The trucking industry could be the first to benefit from the tool, creating a caravan of trucks with a single active driver at the front of the line.

According to John Maddox of the University of Michigan’s Transportation Research Institute, each of these efforts are significant because we’ve already seen the safety benefits of measures such as airbags and structural enhancements reach a plateau. While they cut driving fatalities from 45,000 in 1990 to 32,367 in 2011, that number hasn’t changed in recent years, which could give these new high-end features a competitive advantage if they are introduced to a broader range of vehicle models.

So for the time being, it appears that despite the success of autonomous vehicle testing programs, change won’t be too dramatic, which falls in line with the thinking of John Paul MacDuffie, director of the Program on Vehicle Mobility and Innovation at the University of Pennsylvania’s Mack Institute for Innovation Management.

“I think you could have a lot of people commuting without having to have their hands on the wheel, or their foot on the gas, but I think they’ll still have to be attentive, if somewhat less so, and with a lot of warnings,” says MacDuffie. “But to actually get to the stage where people can completely take their attention away is further away, technology-wise, and acceptance-wise.

“How much are we ready to turn over to an algorithm, even if on balance, it’s safer than humans in many conditions? Those are harder questions to get resolved.”

But even if this gradual shift meant ten percent of cars on the road were driverless, the Eno Center for Transportation estimates that this could have an impact to the tune of 211,000 fewer crashes, 1,100 fewer deaths, and $22.7 billion saved.

EnterpriseAI