Monday, December 17, 2018

Self-Driving Truck Loses Its Remote Connection, But Not Its Shot at Milestone Achievement

Starsky Robotics
Article thanks to Pete Bigelow and caranddriver.com. Links provided:
March, 2018  From his vantage point in a car a few dozen feet in front of a 20,000-pound big rig barreling along a Florida highway without anyone behind its wheel, Stefan Seltz-Axmacher felt like everything was going according to plan.
As the founder of self-driving-truck startup Starsky Robotics, he had spent more than two years building to this moment, when a truck with no human sitting in the front seat, no safety driver waiting in the sleeper berth, no human anywhere on board, would trundle along a public road. Now, watching the first two miles of the groundbreaking journey from a lead car riding in front of the tractor-trailer, he confessed that it all felt routine. Maybe even, like the reality of so much of current autonomous-vehicle testing, boring.
And that’s precisely when the truck inexplicably slowed and stopped in the middle of the road.
“I’m thinking, ‘This is not planned,’ ” Seltz-Axmacher said. “And so we get out and turn off the engine and start investigating.”
Starsky Robotics engineers tested and developed the autonomous-driving system, which has been installed in the Freightliner Cascadia for more than a year, and had spent a full week conducting dry runs on the stretch of County Road 833 in southern Florida, just north of the Everglades. They had plotted every conceivable contingency and every imaginable edge-case scenario.
In doing so, they illustrated one of the more vexing aspects of automated driving not just for themselves, but for every company that is developing self-driving technology: For all the tens of thousands of scenarios gathered through millions of miles of testing, it’s the unforeseen outlier that looms as the one of the most complicated challenges in deploying a self-driving vehicle.
On this highly anticipated mid-February afternoon, during the most important test to date, that outlier was the improbable circumstance that the building housing the company’s remote operations team in Plantation, Florida, lost power. When the truck lost its signal from the teleoperations center, dozens of miles away from the test road, it immediately went into a safe mode and came to a gradual stop in its lane of travel.
Far from what was expected, but for Seltz-Axmacher and his team, it was reassuring to see the truck proved it could handle the unexpected.
“Of all the different flaws that could have happened, all those things we tested and expected, we never tested shutting down the power to the building,” he said. “But by having this safety architecture in place, you are able to be confident that if a failure happens, even weird failures that have never happened to us before, we will catch them. If we catch them, we will come to a stop.”
Taking Driverless to New Levels
Once the issue was identified and power restored, the company continued the fully driverless journey and completed five more miles of testing along the intended route. One day later, Starsky Robotics completed the entire seven-mile stretch along County Road 833 without a human aboard, and without incident.
Among companies developing self-driving-truck technology, it’s believed that this was the first fully driverless test on a public road with no human safety driver aboard. In October 2016, Otto, a self-driving truck subsidiary of Uber, conducted a 120-mile test along Interstate 25 between Fort Collins and Colorado Springs, Colorado, but it had a human safety driver monitoring the test from the bunk. Since then, Otto has been folded into a unit of the company now known as Uber Freight.
Both tests were highly structured, in that local officials and law enforcement had shut down the roads temporarily or worked to separate the test trucks from regular traffic. But for Seltz-Axmacher, last month’s test, conducted at a maximum speed of 35 miles per hour, was another milepost in the race to deploy autonomous trucks.
“It’s a huge deal for us, and a really big sign of how serious we are,” he said. “The other teams have not done that. We can test all day long with someone behind the wheel and everyone could just focus on testing reliability and growing a feature set of what their system can do. They can make each thing reliable enough that it doesn’t fail often—but if it does, it can be controlled by the person behind the wheel.
“But the problem we’re trying to solve is the person in the truck. It’s time to do unmanned testing.”
Eleven months ago, Starsky Robotics started using self-driving trucks to haul commercial freight, and it fine-tuned its system in truck yards, hauling everything from 5000 pounds of milk crates to 40,000 pounds of tile. Last fall, a truck hauled bottled water to Florida residents affected by Hurricane Irma. It traveled a 68-mile stretch of Interstate 75 between Fort Myers and Miami under control of its self-driving system without requiring any intervention from its human safety driver.
Starsky’s steady progression has brought the notice of venture-capital firms. On Thursday, the company announced that Shasta Ventures has led a funding round that has collectively raised $16.5 million. Other investors include Y Combinator, Trucks Venture Capital, Fifty Years, and 9Point Ventures.
Over the remainder of 2018, Seltz-Axmacher said, the company intends to up the frequency of its unmanned testing runs, which will include freight hauls at some point. Florida is the current location for testing, but Starsky likely will expand to other states by year’s end.
Remote Connections Grow in Importance
Although based in the Bay Area, the company cannot test in its home state because California prohibits automated testing of vehicles that weigh 10,001 pounds or more. But the California Department of Motor Vehicles and others may nonetheless want to pay close attention to the company and its experience with teleoperations, a slice of autonomous operations that’s increasing in importance.
In late February, the California DMV finalized revised statewide regulations that govern autonomous testing which permit the testing of fully driverless vehicles with no safety drivers aboard, as long as the cars can be controlled through remote operations.
In Starsky’s case, cameras installed in the cab and around the truck give remote operators a bird’s-eye view around the truck. These operators can take control and remotely guide the truck through situations that its self-driving systems can’t comprehend on their own. For example, one of the company’s trucks approached an intersection where a fire truck was parked near the shoulder of the road and a firefighter was standing in the middle of the road directing traffic. Although a safety driver was aboard the test truck, remote operators guided the truck through the intersection.
“If a vehicle in front of the truck slams on its brakes, that’s a safety-critical decision that’s handled by the truck itself,” Seltz-Axmacher said. “Decisions where teleoperations are involved take five to 15 seconds to make; they’re not safety critical. It’s ‘I’m stuck behind a slow-moving vehicle. Should I pass or not?’ ” Those kinds of decisions are easy to make in the office.”
And when the office unexpectedly closes down, the trucks make the safest decision of all:  simply stop.


Wednesday, December 12, 2018

Railroads have spent a lot of money the last 10 years, and customers have barely benefitted: FTR speaker

Article thanks to John Kingston and freightwaves.com. Links provided:
After spending plenty of money on investment in the last 10-plus years, the nation's class 1 railroads are basically standing still.

That was the sobering view presented by John Schmitter, a former railroad executive who now heads his own consultancy, KEP LLC. Addressing the 2018 FTR Transportation Conference in Indiana, his subject's title--"How Technology will Change Rail Economics"--reviewed how technology has made those changes, and yet has provided little shift in overall network velocity and essentially no increase in capacity since 2005.

All of that came to a head last year, Schmitter noted, with the rapid deterioration of rail service, a downward shift severe enough to have the Surface Transportation Board put the class 1 railroads under review.

With the failure to invest in more capacity, Schmitter said, "there's not much system resiliency or flexibility on most railroads. Rapid unexpected volume increases have and will result in congestion and service deterioration that can last months." And that's what happened in 2017, running into 2018.

If there's a culprit in all this, Schmitter said, it's the investment community. "If the railroads invested in too much capacity, Wall Street would kill them," Schmitter said. The lack of resiliency in the system, Schmitter said, "is a management a decision."

"A few railroads invested more than half their operating cash flow back into the systems," according to Schmitter, "but if it gets out of line and Wall Street finds they overinvested in capacity, they'll find some new management."

And when spending is undertaken to make needed fixes, it's too late. "The investments that are being made now are being done to handle the volume that showed up in 2017," he said.

Citing the weather for the 2017-2018 is misplaced, Schmitter said. "The reason was rapid unexpected increases in volumes, because the system by design is capacity constrained."

Schmitter asked the meeting's attendees in the room how many of them were rail customers, and how many had suffered disruptions during the recent disturbance. Of the hands that went up the first time, most of them stayed up when the second question was asked. But when Schmitter asked how many had changed carriers or moved that business over to trucks, virtually all hands went down. "So there's no penalty," Schmitter said.

Schmitter noted that the class 1 railroad competitive market basically boiled down to--though he did not mention the companies by name--BNSF vs. Union Pacific (NYSE: UNP) in the West, CSX (NYSE: CSX) vs. Norfolk Southern (NYSE: NSC) in the east, with Canadian Pacific (NYSE: CP) and Canadian National (NYSE: CNI) in Canada with their systems extending into the U.S. (He did not mention Kansas City Southern (NYSE: KSC), but they view themselves as the "NAFTA railroad" into Mexico.) With that minimal level of rail competition, most locations have only one option on the rails, with a duopoly at best. Asked if that situation inspires companies to innovate, Schmitter said he questioned "that it's any one's goal to be better than the other guy. It's not really a competitive marketplace like you would experience in trucking."

The lack of capacity is stopping railroads from aggressively grabbing freight from a driver- and capacity-squeezed trucking market. "They haven't been able to take advantage of all the problems with trucking, and technology is not going to make it happen," Schmitter said. "This is a management issue, a strategic issue and a cultural issue." The goal, he said, needs to change company operations to ensure that "the shipments will be there when they are promised."

The reduction in capacity combined with the rise in demand for that capacity means that on multiple occasions in recent years, traffic on the rails has approached capacity. "Capacity on class 1 railroads is now closely matched to projected volume," Schmitter said. Echoing what he heard a rail CEO say about the capacity crunch, Schmitter exclaimed: "It's five minutes to midnight."

The lack of significant progress since 2005 is in stark contrast to the gains made between 1980 and that year. The start of that stretch, in 1980, is significant because it marks the year of the Staggers Act, which deregulated the U.S. rail industry. Not all of the gains in efficiency were because of technology, Schmitter said; some of it was just simply the abandonment of unnecessary capacity. But during that 25-year stretch, the productivity of employees and the network rose significantly, with volume up 84% while capacity has had no significant increase. The average freight per train is up 63%, according to Schmitter.

But now, Schmitter was mostly pessimistic: "I think things will get more efficient and safer, and there will be a good impact on the (operating ratio). But I don't see technology providing more capacity or to do much more for service."

Sunday, December 2, 2018

Semis involved in fatal collision shouldn't have been on road: Driver trainer

Article thanks to Jason Warick and cbc.ca. Links provided:

Saskatchewan semi drivers are required to pass a road test, but hundreds have passed without any training

Originally Published Nov. 26, 2018  The three semi-trailers connected to the death of a Rosetown firefighter should not have been on the road that morning, says one of Saskatchewan's most experienced driving trainers.
"What were those trucks doing on the highway?" said Reg Lewis. "I don't care whose load was on those trucks. I would never have left Rosetown."
Lewis, a certified semi driving instructor for the past 22 years, was in Rosetown Wednesday morning. He was supposed to take a student for his semi road test. But Lewis postponed the test after seeing the thick fog and icy conditions.
"You couldn't see a block down the street in town. I never left until 1:00," he said.
According to RCMP, at 9:00 a.m. Wednesday, two semi-trailers collided on Highway 4 roughly 20 kilometres north of Rosetown. One semi was attempting to come on to the highway when it was struck by the other.
No one was seriously injured, but Rosetown volunteer firefighter Darrell James Morrison was killed after arriving at the scene to help. He was struck by a third semi and died shortly after being transported to Rosetown hospital.
Lewis, whose own parents were killed in a collision with a semi, said mandatory, extensive training is the key.
Saskatchewan semi drivers are required to pass a road test, but hundreds have passed without any training. Lewis said mandatory training time on the road with a certified instructor will save lives.
"I don't believe any of these trucks is doing anything on purpose. I do believe that it's inexperience and not thinking far enough ahead," he said.
Lewis noted that another Saskatchewan man was killed in a semi collision Wednesday morning just one hour earlier near Wakaw.
A semi hauling crusher dust collided with a car just outside Wakaw. The semi driver was uninjured but the man in the car was declared dead at the scene. His name has not been released.
RCMP said Friday there is no new information on either fatality and both investigations are ongoing.
Lewis and others have been vocal advocates for mandatory semi driver training since 16 people aboard the Humboldt Broncos team bus were killed in a collision with a semi April 6. They say mandatory training would save lives.
Ontario is the only province with mandatory training, but Alberta and other provinces have set time frames to make training mandatory. Saskatchewan officials announced a mandatory training course in the weeks following the Broncos crash, but then reversed their position the same day. They've said things need to improve and they're looking at all options.