Sunday, June 12, 2016

The Road to Driverless Cars -- Safety Systems

One unique characteristic of autonomous vehicles is their potential to be safer than the mode they replace.

Cars today are safer than ever, but their use still results in over 35,000 deaths a year in the U.S. alone. Safety systems are now becoming available with the potential to mitigate the results of driver error, and avoid some accidents entirely. Advanced safety systems - lane departure warnings, collisions alerts, even direct accident avoidance - are finding their way into an increasing number of luxury and even mid-price vehicles.

At this point most of these features are optional and rely on consumers to value them highly enough to pay more.This is reminiscent of the of the early stages of the process that saw the adoption of  passive safety systems such as airbags and ABS.

Consumer interest may lead to active safety features being adopted, perhaps more quickly than ABS and airbags, which had minimal adoption rates early in their introduction (even though they were technologically mature by the 70's). The willingness of consumers to buy, and manufacturers to provide safety features, has increased  over the last couple of decades as safety has become an important consideration when buying a vehicle.

The Insurance Industry was ultimately the driving force behind the near universal uptake of passive safety systems. Reluctance by manufacturers was only overcome by legislation mandating the deployment of the safety features, usually after intense lobbying by the insurance industry. The manufacturers main complaint, that such features would raise costs unbearably, was mitigated as large production runs reduced unit costs.

There is the possibility that active safety systems will be implemented more quickly than passive restraints were. The speed at which more active safety technologies are implemented, especially those that are capable of assuming complete control of the vehicle, is impossible to predict. But it is important to keep in mind that airbags and ant-lock brakes provided only marginal reductions in accident and injury rates. Active safety systems, on the other hand, have the potential to substantially reduce the number of casualties on the road. If that potential can be documented in the limited number of deployed systems the pressure to expand their use is likely to be significant. 

A truly active safety system, one that could steer and brake a vehicle in a dangerous situation, is likely to have an outsized impact attitudes towards such systems in general. If functional it would provide a constant stream of well documented examples of lives and limbs saved. A video or LIDAR track of the first pedestrian saved by a swerving (temporarily) computer controlled car, with a temporarily incapacitated driver behind the wheel, would be.... priceless.

Because there is on average a fatal accident one accident every 2,000,000 miles driven it is more likely than not we will see a an active safety system deal with a life or death situation after about one and a half million miles. This means that with a relatively small number of well equipped cars on the road (say around 50,000) we should see two or three such incidents a year. Even this small a number, because they will be so well documented, could have an outsized impact on public perception as well as policy.

Monday, March 21, 2016

Capacity Case Study

Huch wood would a woodchuck chuck?  -or-  What is the real potential of a driverless system?

Lets start the devil we know. The present "system" is fairly well studied and, for all its diversity, exhibits fairly consistant behaviors across all its forms.  Different roads, weather conditions, and vehicles notwithstanding there is a common factor-- the driver. Drivers can be counted on to take a roadway of almost any design and reduce its capacity.   Give the driver a potholed, too narrow, and winding roadway and there will be more horns than progress. Create a broad, well lit multi-lane freeway and every commuter in the region will converge on it and create a parking lot. 

The driver, not the road, creates a fairly firm bottom line. The average capacity for a lane of traffic at highway speeds is about 1,200 vehicles per hour. Above this number and slowdowns and stoppages become inevitable.  At lower speeds the number increases marginally to a max of 1500 vehicles at 45 mph, but who wants to drive 45?  The bottom line is that it is not a large number (about one car every three seconds) and that it is inescapable, largely because it is rooted in the average driving abilities of a large and heterogenous population.

A driverless system, on the other hand, is all about the design of the road and the quality of the vehicles.   Its true capacity is intimately involved in the limits of technologies we haven't even fully explored yet. Theoretically this limit might be cars moving nose-to-tail down a single narrow lane at 150 mph (35,000 vehicles per hour, give or take). The practical limit is certainly lower, but by how much is above all an engineering question.

Case Study

Trying to determine the capacity of a driverless system therefore depends on creating a plausible picture of what technical capacities might be developed in the not too far distant future (our theoretical woodchuck). The case study that follows describes a mature technology, but not the theoretical limit of what can be achieved under controlled conditions. It imagines that two of Seattle’s most heavily utilized east-west arterials are converted to allow limited utilization be driverless vehicles.  It reserves the center lane for intermittent  (one minute out of five) use by automated traffic.

       The Pulse
Every five minutes or so a "pulse" of automated vehicles would pass down this center lane (westbound on 45th, eastbound on 50th).  Special signaling will alert traffic to vacate the center lane (and complete left turns) well in advance of automated traffic. Traffic at intersections along the route would be stopped in all directions and in a sequence allowing uninterrupted passage of vehicles at a steady speed of @ 35 mph along a route stretching from the UW to Ballard neighborhood in the west. The lane would be reserved for automated traffic for 45 seconds (including a 10 second warning period) every five minutes. Vehicles would travel in groups of three tightly spaced vehicles with three vehicle lengths between groups. Aggregation and disaggregation of vehicle groups would take place opportunistically as there would be no physical barrier between traditional and automated lanes. The lanes themselves, especially at intersections, would be heavily invested with sensors and the capacity to communicate directly with vehicles.


● 45 second passing times with 10 second warning period
● Automated traffic on 5 minute intervals
● 3 vehicle in each platoon, minimal spacing between vehicles (@16”)
● Average Speed 35 mph
● following distance between vehicle groups-- 3 vehicle lengths (@55')
● total platoon length = 3 vehicles (each 18') with following distance = 110ft

● Distance covered at 35 mph (@51 feet per second) = ½ platoon/sec = 1.5 vehicles/sec
● Total Number of vehicles in transit during available transit period at 80% efficiency = 45sec x
1.5 vehicles/sec x .8 = @54 vehicle capacity per transit period
● Total hourly capacity of Lane = 12 transit periods x 54 vehicles= 648 per hour 
● 35 mph Automated Lane 24 hr capacity = 15,500 (one way)
● Actual 45th street traffic (one way)=11,000 (max @ 900/hour)

It’s a startling result. A single lane of automated one-way traffic on an arterial such as
45th street, operating for less than one minute intervals every five minutes, could nearly double the one-way capacity of that arterial. A full time automated system would have about five times the capacity.

Extending the model to an interstate with its much higher speeds yields even more impressive results, especially if we keep following distances similar to the original model.

● 55 mph Automated lane hourly load capacity = 6,400
● Typical Freeway lane, hourly load capacity= @1,200
● 55 mph Automated lane 24 hr capacity (uninterrupted)= 153,000
● Typical freeway one-way 24 hr traffic (I-5 in downtown Seattle) = 150,000

Wit uninterrupted service a single automated lane (at 80% capacity) can carry the
equivalent of five lanes of traffic, essentially the entire freeway. 


The potential capacity of a driverless system is what usually fires the excitement of
someone who wrestles with the glaring inefficiencies of the modern car based transportation
system. Much of the time this excitement leads people to look at designing fully engineered
infrastructure, things like PRT’s. While it's fun to imagine the creation of a world tailored to
our transportation needs, history suggests a different path of change. The widespread use
of driverless cars would suggest a fundamental shift in our society and culture, and changes
that big must be driven by more than an engineered ideal. That is not to say that such an ideal
will not someday be reached. The reliability and efficiency of many systems, such as high
speed rail or the aviation industry in general, could hardly have been contemplated by even
the most enthusiastic visionaries of the 19th century. But the process of change is inevitably

From my office I overlook the ship canal bridge, an important component of I-5 as it passes through Seattle. Over 50 years old, the bridge carries vehicles whose basic design is older still. A driver from 1930 could more quickly learn to operate a modern car than he or she could learn to use a cellphone. The sheer scale of this structure, and the massive transportation system of which it is a small part, seems utterly resistant to significant modification of any kind. This is the story of the last half century. Will the next 50 years to be that much different?

Possibly. Changes on this scale do happen. History give us some idea of the dynamics that drive dramatic and rapid changes. It gives us examples of the forces that impede progress for long periods of time, and then drive it forward at speed when change finally does take place. That sounds like a big enough subject for another blog post eh?

45th Street at Latona, looking west.

Saturday, January 23, 2016

Resiliency - Sustainable Transportation and Driverless Vehicles

Sustainability is usually associated with manageable problems - how to make our everyday activities less of a burden on the environment, and on each other.

But there are other problems. Big ones. Each time we watch a city (or region) respond to an uncommon but inevitable events like the East Coast's snowmageddon we wonder about tomorrow and, here in the Northwest,  the Big One.

What to expect? Well, the script is usually pretty much the same;
  1. Encourage people to prepare before the event takes place. This often involves sheltering in place as road net will inevitably jam as large numbers of people attempt to use it.
  2. Shut down the system as the event begins (the events of #1 may have already taken care of this)
  3. Wait
  4. Use emergency vehicles to get to those in desperate need.
  5. Bring in the heavy equipment to clear the road of the vehicles left behind during phase one.

On a side note; The above process also describes pretty well what happens when it snows in Seattle.

Like a well oiled machine

So the existing system sets a pretty low bar for effectiveness, but it does have some advantages
  • simplicity: each vehicle is dependent only on gas and a driver to be mobile
  • resiliency: the roads are wide and tough and there are a lot of them (about 20% of most urban landscapes are paved), even sidewalks will do in a pinch.
  • flexibility: just about any sized vehicle can use any road

What would driverless cars add to the mix? A a substantial number of autonomous (or auto-capable) cars on a road that is largely unchanged from todays infrastructure. The second is a fully autonomous system with an infrastructure largely optimized for smaller, more numerous vehicles and embedding a intelligent element in the road itself.

The first option would likely respond to natural disaster in the same way as the present system with a few limited exceptions. Driverless vehicles would be able (to the extent the roads haven't already been jammed) to move about during the event and develop a fuller, real time picture of road conditions etc. 
This would help during the rescue phase and even allow some non-drivers to escape if conditions allow. During the rescue phase some people could be moved in driverless vehicles freeing up manpower for other tasks. In the aftermath, depending on the efficiency of driverless technology, mobility might be restored more quickly by allowing only driverless vehicles into the area thereby avoiding bottlenecks etc. Otherwise, not much change from the present system....bring in the helicopters.

The second option would likely look much different. Most important would be the potential for a well designed system to move large numbers of people very quickly. Ideally this would mean more numerous small evacuations at critical points rather than general evacuations over large areas. If the system were well designed and had an extensive network it would be possible to route resources around almost any disruption ( a bit like the internet).

Of course, if it is not well designed it runs the risk of a total system failure, possibly at the disastrous moment when thousands are on the move and far from shelter. Bring in the helicopters.

On balance it appears to provide at least the potential for improving our response to natural disasters. At present the bar is set pretty low, too low, and new tools should be welcomed.