AUVSI's Driverless Car Summit 2013 kicked off with discussion on how to make automated driving a reality.
The summit, held for the second year in a row at the MotorCity hotel in Detroit, has always had the goal of making driverless cars hit the streets by 2022, but with a recent release from the National Highway Traffic Safety Administration on its commitment to increasing autonomy in vehicles to achieve greater safety, this year's conference hit a pragmatic tone early in the presentations.
The morning keynote speaker, Dr. Bryan Reimer of MIT's AgeLab, discussed how the industry must make cohesive steps forward to progress from the current level of automation in vehicles - outlined by NHTSA as Level 0 and Level 1 automation - to Levels 3 and 4, currently active on Google's fleet of driverless vehicles.
"How do we really begin to think to chart a pathway to deploy self-driving autonomous vehicles or even semi-autonomous vehicles so we can begin to successfully realize what we're suggesting here?" he said.
Reimer said there are benefits to automated systems that improve safety, but with the average car around 10 years old, the industry needs to make sure the technology doesn't need a high penetration level to be relevant.
Another difficulty on the path forward, he said, is the more people rely on autonomy, the worse they get at performing tasks. So if in the future, a human has to take over momentarily in a self-driving vehicle, that person's level of driver expertise will be lower than it is today, an issue he said the Federal Aviation Administration is currently addressing with its autopilot-reliant pilot pool.
"Automation in the cockpit has caused many accidents over time when the pilot is no longer capable of managing the situation," he said.
Reimer took on a skeptical tone when discussing Google's efforts, telling the Michigan audience that he'd be wary of the Internet search engine company's testing procedures. He cautioned that one Google accident, even if it weren't the car's fault, would turn the media and legislation against the industry, and the automakers would pay the price.
The company recently announced it would not adopt NHTSA's recommendations dealing with automation in vehicles.
Bob Denaro, a speaker that works as a Department of Transportation Intelligent Transportation Systems consultant, countered Reimer's opinion.
"I don't think they should be restricted from operating on open roads," he said, regarding Google. However, he questioned if real-world driving scenarios are as valuable of a learning experience as a test facility.
"I think the answer to that question really depends on what level of automation you're talking about," he said. "If you're doing general cruise control on an open road, you might argue you'll get sufficient date from a controlled test compared to that."
Denaro said ultimately that urban driving would be the most difficult scenario to test.
Denaro's speech focused on the research gaps in driverless car testing now.
He keyed in on a test scenario he thought could be aided by smarter vehicles. If you are in a car, following behind two other vehicles, and the first car brakes and the second car swerves instead of brakes to avoid an accident, the limited vision from the third car would make avoiding an accident difficult. However, if there was a vehicle-to-vehicle communication device in the mix, the third car would know the first car was braking without the driver having to see the car's brake lights.
"I think more sensors is better as long as we can figure out the cost implications of that," he said. He says he favors cars with interior sensors to detect the state of the driver, stability sensors that deal with the road situation, and near-field sensors like cameras, lidar and even inter-vehicle communication systems.
"We know that we have self parking cars. … To me that makes a lot of sense for those that have to parallel park a lot," he said. "Traffic jam assist, absolutely - that's brilliant in my opinion."
Denaro also addressed human factors. He said he owns two cars, an SUV with a rear camera and a Mazda Miata. And often when he's backing up in the Miata, he forgets to look behind him.
Michigan Gov. Rick Snyder, who addressed the audience via a live Google Hangout session, discussed these kinds of societal issues, saying he believes the technology isn't as difficult to conquer.
"To be blunt, I think that's the easiest of the issues," he said, regarding driverless technology. "I see it really being, one, as a culture issue."
Synder, whose Twitter handle is @onetoughnerd, said he's taken a spin in a driverless vehicle before, but wouldn't purchase a self-driving car right now since their space on roadways is limited to testing.
Michigan currently has pending self-driving legislation that Snyder says he'd like to see pass sooner in the technology curve, because in his experience, people are more reasonable in early stages of technology adoption.
"We hope to continue to be a leader in automotive technology, so this is important to us," he said, adding that he'd also like to work with other states, like California, and the federal government to best leverage this technology's development.
Automakers and Suppliers' Perspectives
Today's Driverless Car Summit featured sessions from both automakers and suppliers that followed two common themes: the goal to have accident-free vehicles and the steps they need to take to ease in self-driving technology.
GM's Dr. Jerry Salinger discussed the company's Super Cruise model, which he says is "a step along the continuum to automated driving." The company has undergone testing of the concept with adaptive cruise control, and ACC with and without "countermeasures" Salinger never elaborated on. GM studied the amount of time drivers at its Milford Proving Grounds would glance away from the road. With just ACC, almost no one looked away for more than two seconds, but without any countermeasures that number ballooned to up to 12 seconds, an undesirable level at this stage of automation.
"Not surpring … we saw that with these automated driving systems … there was a much broader range of tasks drivers engaged in, and they were a much riskier set of tasks," he said.
Speakers from Volvo and Ricardo discussed the SARTRE platooning concept testing that finished up within the last year.
"By 2020 no one will be killed or injured in a Volvo Car," said Erik Coeligh, stating the company's ultimate goal with its driverless testing.
This project also had a focus on how a human was interacting with the car, like when it would be safe to hand over control.
Platooning also has other challenges, he said. It's a chicken-or-the-egg scenario of how you'd create a system with lead drivers and cars that follow them, since both are reliant on the other existing first. He also showed a photo of the first follower car that trailed a large truck, whose radiator was completely caked with kicked up rocks from the leader vehicle.
Nissan's Maarten Sierhuis bridged the gap between human and machine interaction with his unique qualifications. Formerly a NASA employee, Sierhuis worked on how machines and humans should work as a team, before transitioning over to the automotive world for Nissan's new autonomous test facility in Silicon Valley, Calif.
"I strongly believe that if we want to move to autonomous vehicles we have to look at the car and driver as a team, kind of how we look at the equestrian and the horse," he said.
To get to a perfect balance, he said a lot of testing needs to be done to see how and when people want to use cars autonomously so manufacturers can build vehicles that will collaborate with humans to fit these needs. And this type of testing should be done quickly, so the path to automation doesn't become bogged down in years of researching just one scenario.
"It means not wasting time for years and years to figure out what is going on in a lab before we're trying this in the real world."