2012/07/18

Driverless journalism

Added on 2012.09.13
     Pro: Kiss Your Bus Goodbye

People are starting to note a lot of positive implications for the self-driving vehicle (SDV) that is resonating on either side of the car-anti-car divide. In the WSJ, Brooking’s Clifford Winston wrote: Paving the Way for Driverless Cars: Instead of focusing on an enormously expensive high-speed rail system, government should promote modern highway design for cars of the future. (2012.07.17)

While Winston is half right that "a much better technological solution [than high speed trains] is on the horizon", the comparison is off-base since the sweet-spot trip for heavy rail vs that for the SDV are widely separated. Kind of like saying a baby stroller is equivalent to a mountain bike, since both move children. I wish he had written: “a much better technological solution [than today’s car] is on the horizon". But he was more using the SDV story as an excuse to diminish heavy rail and promote private investment in transport infrastructure. (While appreciating a role for rail, I whole-heartedly support private investment.)

Winston is also half right to claim: No worries about rush hour, vacation congestion, bad drivers, speed traps and accidents." Urban rush hour trauma and congestion would be reduced. The cessation of bad driving (on whose definition?) will take a long period of attrition, but one interesting idea might be to suspend driver licenses for “bad drivers” for a year constraining such drivers to a SDV for that year. Perhaps onerous to some it would be life-changing and even life-saving for others. The same with “no accidents”. There would be fewer—many fewer, eventually—but zero would be unlikely. When I saw “speed traps” listed among the bad bits such as "rush hour", "congestion", "bad drivers" and "accidents", I first recalled the Sesame Street jingle: "One-of-these-things-don't-belong-together…", then I wondered if that provides us with a hint of an opportunistic attribute of Winston’s own driving?

I think Winston’s concern that “one-third of the nation's highways are still in poor or mediocre condition” is both exaggerated and in no way a showstopper with respect to the SDV. The SDV’s lead designer, Sebastian Thrun has admitted that the work is not done—specifically listing the challenges of driving in snow, construction zones and “avoiding a mattress on the roadway”. I am certain he or someone will solve the pothole problem. Thrun started on this in 2005 and has accomplished a lot in eight years (actually the early vision of the SDV goes back to the 1939 World’s Fair, and was electric, no less). Thrun has been clear that the “car will be ready when the car is ready”, so he and Google (and several other competitors) will not rush-to-release a technology that would be unsafe or trip on potholes—nor would they be allowed to do so. Nonetheless, the SDV will suffer from the same problem as does commercial air carriers—while far safer than human operated cars, they will be held to a far higher standard. And Thrun knows that, too.

Winston’s proposal to build a whole new infrastructure is alarming. Any scheme that puts two different roadbed compression strengths in adjacent lanes is especially misguided as a simple lane departure of a heavy vehicle could cause tremendous damage. If his intention is that such lanes be physically grade separated, then flexibility would be greatly reduced and the tiny portion of the network that would then avail to SDVs would greatly restrict their movement. If we can’t get people to buy EVs with restricted ranges why would they buy SDVs with restricted routes? Sounds like the bus to me. And to imagine that because of this separation “driverless cars … would not have to distinguish between cars and trucks” is a terrifying idea. Would you agree to be whisked along in a robotic car that could not distinguish among vehicle sizes, one of the simplest of robotic vision feats? Count me out.
The illustration accompanying Clifford Winston's article is misleading.
This shows a car on a guided runway, not an autonomous vehicle.
The SDV will operate on the existing infrastructure and an internal map can keep if off any parts of the network that would put vehicle or rider(s) at risk. Even the signal timing problems can be addressed partly by fixing some critical parts and adapting the vehicles to others. Consider that as the SDV fleet grows, learning algorithms that use massive quantities of trip data (such as do those used by INRIX to map country-wide congestion in near-real time) can get very smart. They could easily prioritize which of the signal timings need most urgent fixing.

The frontier benefits of the SDV will accrue during 2022-2042 as special, restricted applications such as replacing mostly-empty and oversized urban buses, expensive and poorly driven taxis and shared cars. Here is where I would like to see Winston’s call for private funding focused: urban fleets of self-driving jitneys to replace every form of motorized shared vehicle (bus, taxi, street car, shared car, vanpool) from the front door of your home or work right up to the light-rail and heavy-rail transit station and vice versa. Replace them all. Then by 2045, maybe the US Congress will be able to pass another Surface Transportation Reauthorization Bill in plenty of time to eulogize the last of the personally-operated SOVs and fix the last of the traffic signals in time to remove them all, because they will no longer be needed

Not to lose sight of the key value of Winston’s message, however, he is 100% right that more would be achieved per SDV dollar than per heavy rail dollar, although both are needed.

2 comments:

Matt Young said...

I have speed, low service times the main driver for robots on the road.

Speed, when robots start carrying 50 plus passengers safely at 120MPH, then service times drop all across the board and the demand will sweep through the BRT industry.

At ranges greater than 50 miles, cars simply cannot compete with high speed robotic buses, especially since the robotic buses retain the driver and the steering wheel to navigate the last mile (which rail cannot do).

Transportation experts need to look at queueing analysis, when trip services time drop with high speed there is a huge multiplier up and down the economic chain.

Paul Godsmark said...

Today (25 Sept 2012) at a press conference for California to sign the Autonomous Vehicle Bill into law Sergey Brin of Google made a conservative estimate that autonomous cars (I call them autonomes) will be on sale in less than 5 years.

It is my expectation that the taxi industry will disrupt immediately, and that the more entrepeneurial private autonome owners will quickly realize that they can hire out their vehciels for a small profit rather than leave them standing 95% of the time - thus giving birth to the Transport as a Service (TaaS) model. This competition with the taxi companies will rapidly result in a very low cost efficient door-to-door service that will disrupt bus services. This could also impact on LRT and other public transit systems.