Wednesday, August 13, 2014

Why did Detroit Commuters get Caught in a Flood? (It wasn't because Rain.)

The initial response to the rainstorm in Southeast Michigan on Monday 8/11/2014 is that it was a freak occurrence—unpredictable and unavoidable. In some ways this is true.

When you get a lot of rain, things flood. In floodplains, this is unavoidable. This is why FEMA publishes floodplain maps that show areas predicted to be underwater  ever hundred years or so. But Detroit is not in a floodplain. Flooding problems in Detroit were not caused by a storm surge or rising rivers. If there was no city here, there would not have been a flood. The reason areas in Detroit flood is because we either dig basements which very easily fill with water, or build streets that don’t let water drain. As in most older cities, Detroit’s stormwater infrastructure is a disaster and it’s completely expected that a lot of rain will flood the surface streets. So, yeah, OK then, that will happen.

But let’s focus on the Interstates.

Interstates are designed to very demanding standards. The federal government requires high standards on Interstates because it is acknowledged that they are strategically and economically essential to operate under all foreseeable conditions. The rainstorm on Monday August 11, 2014 was not a freak occurrence. It appears to be within the limits of a 100-year design storm that the Interstate should be built to withstand. Unless I’m misunderstanding something, it looks like MDOT should be designing for 3-inches of rain in a 2-hour period, and 5.5-inches of rain in a 24-hour period. From what I can tell from spot-checking weather station data, the 8/11 storm did not exceed that.

The fact that drivers were forced to abandon their cars on flooding Metro Detroit Freeways means that something went wrong somewhere.

I-696/I-75 Interchange

If I am wrong and Monday’s event did exceed the design storm, MDOT and MSP should have been able to predict flooding and closed affected highways before they flooded, or were at least prepared for the possibility on short notice (i.e., had crews ready to react if flooding was observed). It may seem over-demanding of MDOT and MSP to have this capability, but they invested in a state of the art Traffic Operations Center that is supposed to do just these kinds of things (e.g., monitoring traffic and weather conditions to anticipate and deploy mitigating strategies). They may never had considered that Interstates would flood, since it hasn’t been a problem in the past; but it is actually their job.

MDOT/MSP Southeast Michigan Traffic Operations Center (SEMTOC)

But again, it looks like the storm was within design parameters, meaning the freeways never should have flooded in the first place. One possibility is that Metro Detroit’s Interstates were not designed to withstand a 100-year storm. This does not seem likely.

The more likely possibility is that the Interstates did not function as designed. These things are usually built pretty well. But sometimes they fall into disrepair. Concrete crumbles, drain basins get silted in, etc. There are prescribed maintenance practices, but maintenance is usually underfunded. In the case of Detroit’s Interstates, many of which are built in below grade (in big troughs) the drainage system is heavily reliant on pump stations. My guess is that for some reason, pump stations were not in working order.

Whatever the cause, the fact that motorists had to abandon their cars to rising waters on Monday means that a mistake was made somewhere. This is a good thing, because we will get more rainstorms like this. The fact that it was a mistake means that we can find it and fix it. That process needs to begin as soon as the mud gets scraped off the roads.

Thanks, @Jake_Dudek

For crazy pictures, search Twitter for #detroitflood. Yeah, OK then.

Monday, February 3, 2014

Did NHTSA say anything new on a connected vehicle mandate?

The National Highway and Traffic Safety Administration (NHTSA) has been studying the benefits of connected vehicles for decades. This culminated in the Connected Vehicle Testbed Model Deployment in Ann Arbor, where a fleet of connected vehicles hit city streets in 2012. NHTSA's end-goal has long been to standardize and mandate a particular type of connected vehicle technology, known as direct short-range communication (DSRC). The Ann Arbor deployment was designed to acquire data to support such a regulatory decision.

On February 3, 2013, NHTSA via formal press an intent to "move forward with vehicle-to-vehicle communication technology for light vehicles." In the center of boilerplate press language about the benefits of V2V, this seems to be the meat of the announcement:
NHTSA is currently finalizing its analysis of the data gathered as part of its year-long pilot program and will publish a research report on V2V communication technology for public comment in the coming weeks. The report will include analysis of the Department's research findings in several key areas including technical feasibility, privacy and security, and preliminary estimates on costs and safety benefits. NHTSA will then begin working on a regulatory proposal that would require V2V devices in new vehicles in a future year, consistent with applicable legal requirements, Executive Orders, and guidance.
Yeah, OK then... So NHTSA is studying V2V technology. We knew that already.

This press release suggests that a the formal regulatory process has not yet even been initiated. NHTSA has not even yet begun work on a regulatory proposal? They didn't mention posting anything in the Federal Register. It seems like nothing has changed since yesterday. Am I reading this wrong? What, exactly, did we learn from this press release?

NHTSA announcement is here:

Wednesday, January 22, 2014

Barriers to Successful Implementation of a National DSRC Connected Vehicle Network

The following is an extended abstract that I have submitted as a draft technical paper for the ITS World Congress 2014.


The primary goals of the U.S. DOT connected vehicle research are to improve surface transportation safety, mobility, and environmental performance in the United States. As of this draft technical submission submitted January 22, 2014, the ITS community awaits a National Highway Traffic Safety Administration (NHTSA) decision regarding direct short-range communication (DSRC) connected vehicles in the United States. Many stakeholders anticipate that NHTSA will announce an intent to mandate that DSRC equipment be installed in all new vehicles at a future date. However, a critical assessment of this strategy suggests that there potential complications that must be considered. This paper provides an overview of procedural, political, and technical barriers that impose significant difficulty in pursuing a DSRC connected vehicle mandate.

U.S. DOT connected vehicle research can be traced back to the early 1990s, when the Department commissioned the development of a strategic plan for Intelligent Vehicle Highway Systems (IVHS) in the United States. The resulting report was prepared by IVHS America—an official congressional advisory committee—and was published in May of 1992. The strategic plan outlined a 20-year program by which a partnership of federal, state, and private agencies would work together to design and deploy a national IVHS program for increased safety, improved mobility, and environmental quality. Such goals proceeded through the U.S. DOT's Automated Highway System (AHS) demonstration, the subsequent Vehicle Infrastructure Integration (VII) program, and lives on today in the strategic research program managed by the ITS Joint Program Office within the Research and Innovative Technology Administration (RITA) of the U.S. DOT. These federal initiatives have successfully supported the standardization, technological advancement, validation, and deployment of numerous ITS systems across the U.S.

Most existing ITS deployments in the U.S. today are managed by local or state transit agencies, and often operate independent of adjacent or overlapping ITS deployments. Many ITS industry stakeholders believe that the greatest potential for returns to investment on ITS installations will require that ITS systems are integrated into a national network of connected vehicles and infrastructural components. The U.S. DOT has adopted such a view in its Statement of Principles for a Connected Vehicle Environment which reads, "System implementation must be national in scale and extensible across North America." The system is envisioned as a centrally managed national network provided as a public service using open nonproprietary communication and performance standards. The U.S. DOT has found that DSRC is the only known viable technology for safety critical applications provided within such a system. While the Department's approach to pursuing the goals of the connected vehicle program is continually evolving, many ITS and safety advocates have called on NHTSA to pursue a mandate requiring DSRC standard equipment be installed in all new vehicles sold in the United States.

A critical analysis of a DSRC mandate scenario indicates that such an approach may be fraught with difficulty. The most immediate barrier to a DSRC mandate concerns limitations on the U.S. DOT and NHTSA as regulatory agencies. Pursuant to federal requirements, regulatory agencies are subject to a critical review process to ensure that regulations are designed to achieve benefits that justify any costs imposed. Advocates of the U.S. DOT connected vehicle program argue that the cost of a mandate is likely to be marginal when compared to the potential savings in fatalities, injuries, property damage, lost economic activity, etc. NHTSA estimates that motor vehicle crashes in the United States have a negative economic impact of about $230 billion each year (including thousands of serious injuries and lives lost) and that up to 80% of these crashes could be prevented with connected vehicle technology. However, the results of a formal cost-benefit analysis are far from certain. This is partially a result of established accounting principles that emphasize near-term costs and benefits; a national DSRC network is likely to require significant upfront cost with the bulk of benefits accruing years or decades later. Long-term impacts of a DSRC mandate would be subject to a discount rate that will significantly depreciate the value of projected benefits. Further, many claims about the safety gains of a DSRC mandate may not stand up to critical review. Even under favorable assumptions, the installation of DSRC equipment alone is unlikely to have much safety impact unless coupled with automated safety systems. Many automated safety systems rely on sensor technology and can operate independent of DSRC support. Such systems already appear to be contributing to a downward trend of vehicle crashes in the U.S. The safety gains of DSRC would likely have to be compared to baseline assumptions of safety increases provided by alternative technologies, further reducing the estimated marginal benefits provided by a DSRC mandate. In addition to the likelihood that the benefits of a DSRC mandate would be less than proponents expect, it is difficult to envision NHTSA being capable of estimating the costs of such a mandate with much precision—as some very basic elements network design and management remain unknown.

If NHTSA is able to show that the benefits of a DSRC mandate outweigh the costs, the successful adoption of such regulation may require extensive political capital. The current political climate on the United States is such that government involvement that appears reasonable to some may appear onerous—even nefarious—to others. Any proposed regulations regarding a DSRC mandate in light vehicles will likely attract a great deal of public attention. Special interest groups and media sensationalists will compete to control the narrative. Potential adversaries of a DSRC mandate include civil liberties organizations and advocates of limited government. One possible talking point of DSRC adversaries is that the Government wants to put a tracking device in everybody's car. And to be fair, this is partially true. Placing aside the technical difficulties of ensuring that DSRC equipment is secure and motorists are not tracked, the rhetorical difficulty of convincing the public that privacy will be maintained will be significant. The U.S. DOT and NHTSA would be well advised to consider such broader implications when considering regulatory mandates.

Beyond barriers relating to proposing and adopting a DSRC mandate, the potential difficulties for implementing a mandate are considerable. Current conceptions of the potential DSRC connected vehicle network heavily emphasize security and privacy, which adds complexity to the system. Specific security standards remain in development, but are expected to include a public key infrastructure (PKI)—essentially encoding each DSRC transmission with an authentication ID code that identifies the message as coming from a legitimate device. In order to protect the privacy of drivers, it is expected that each vehicle will update its ID code periodically so that the code cannot be used to track the location of the vehicle for a significant time or distance. The maintenance of the PKI system, as well as the general operation of the network, will require complex engineering and intensive IT support. The cost of deploying and running the system is very difficult to estimate at this stage in development but is likely to be substantial. The deployment and operation of such a network may amount to an IT project at a magnitude and complexity that is without precedent. Further, the scope NHTSA's authority does not likely include the operation of such a system, implying that novel public-private management structures may have to be designed to deploy and operate the network. This management dynamic may be very complex and plans for a supporting institutional structure remain very preliminary.

This paper provides a critical analysis of the prevailing direction of U.S. DOT connected vehicle efforts, focusing on a potential DSRC mandate. It is found that NHTSA intent to propose a mandate would involve a variety of procedural, political, and technical barriers. The culmination of the various potential barriers to a DSRC mandate for connected vehicles suggests that alternative regulatory approaches should be evaluated in earnest

Friday, December 20, 2013

Google's Purchase of Boston Dynamics Could Fast-Track Automated Driving

Google has been working on artificial intelligence (AI) and machine-learning algorithms for many years. By acquiring Boston Dynamics, Google has not only purchased terrifying proto-Terminator robots, but they have gained a plethora of AI that has been specifically developed for robotics.

The broadened catalog of AI algorithms will likely benefit Boston Dynamics legacy programs, but has huge implications for one of Google's most hyped projects: the Google Autonomous Car. The Google Car has already logged millions of miles in traffic situations and appears to be quite reliable on highway driving. This is impressive, but not unique. Many traditional automakers have already commercialized automated driving systems capable of limited self-driving (or partial-automation) in a highway environment. Automated an entire trip is more challenging by an order of magnitude. It's relatively easy to program a car not to crash, but it's not at all clear how to program it to properly navigate congested surface streets around unpredictable traffic, pedestrians, obstructions, etc. What Google and others have realized, is that it is much easier not to program these vehicles, but to train them.

Google and others have become remarkably good at the mechanical aspect of automated driving. Driving is only partially mechanical; it is largely social. We don't realize how complex of a task driving is because we have been evolving algorithms for millions of years that allow us to navigate both our physical and social environments. You can easily program an automated vehicle to obey the rules of the road and not hit things. But unless the vehicle can interact in our social world in a relatively human way, it will become confused and freeze up while trying to navigate in our world.

The conventional wisdom is that Google is using its fleet of automated vehicles to test its automated driving system. This is certainly true, to an extent, but it's not the most important aspect of what Google engineers are doing. Whenever the Google car encounters a unique or unusual situation that the software is not ready for, the test driver must temporarily take over the dynamic driving task. This is important for safety reasons to test in public, of course. But this is important for program development, because the car is watching and learning from the human driver. The aggregated experience of the fleet of Google cars is used to update and refine the controlling algorithms. Software programmers are surely involved in this process, but they don't have to program each unique traffic maneuver from scratch. They have they data regarding what the cars sensors saw in a unique situation, and can see how and why the algorithm became confused, causing the driver to take over. They can use this information, in addition to what the driver did, to increase the capability of the software. Boston Dynamics very likely has a lot of technology that will improve the ability of the Google automated driving system to learn from experience.

A year or so ago, I believed Google was basically playing with toys here. Or maybe the technology would be ready decades from now. Every time I've revised my view on this, it has involved increasing my respect for the Google Car program and subtracting from the time I would predict the technology will be ready. Yeah, OK then... My current prediction: I believe Google will have a consumer-ready driverless (NHTSA level 4 automation) product or service before 2018.


Tuesday, December 17, 2013

Michigan Considering Ban on Automated Vehicles

As I previously reported, Michigan's state legislature has been working to make Michigan the 4th state to pass a law concerning automated vehicles. The bill has gone through some revision and now passed through congress as SB 0169 of 2013 and is ready to be signed by the governor. While the intent is to highlight Michigan's potential as a high tech state, a significant effect of this law would be to make Michigan the first state to explicitly ban automated vehicles for non-testing purposes. This may sound reasonable for those who think we're still several years away from road-worthy driverless cars. However, the way the bill defines automated vehicles, some of today's cars may fall under the law.

Here is the S.B 169 language; emphasis is mine.
SEC 2B (1) "Automated motor vehicle" means a motor vehicle on which automated technology has been installed, either by a manufacturer of automated technology or an upfitter that enables the motor vehicle to be operated without any control or monitoring by a human operator. Automated motor vehicle does not include a motor vehicle enabled with 1 or more active safety systems or operator assistance systems, including, but not limited to, a system to provide electronic blind spot assistance, crash avoidance, emergency braking, parking assistance, adaptive cruise control, lane-keeping assistance, lane departure warning, or traffic jam and queuing assistance, unless 1 or more of these technologies alone or in combination with other systems enable the vehicle on which the technology is installed to operate without any control or monitoring by an operator.

(2) "Automated technology" means technology installed on a motor vehicle that has the capability to assist, make decisions for, or replace an operator.

(3) "Automatic mode" means the mode of operating an automated motor vehicle when automated technology is engaged to enable the motor vehicle to operate without any control or monitoring by an operator.

Sec. 663. Except [for testing purposes], a person shall not operate an automated motor vehicle upon a highway or street in automatic mode

So SB 169 explicitly bans operation of a vehicle in automatic mode. In other words, outside of testing scenarios, one cannot use features of a vehicle that can "operate without any control or monitoring by a human operator."

The language of the bill is sloppy enough that this could be interpreted such that no driverless car could ever be banned. Alternately, a few 2014 model-year vehicles could be illegal to operate. This is because the legislature tied crucial definitions to words like control and monitoring, but did not define these terms.

On one extreme, this bill may not ban any automated vehicle ever. Even a completely driverless car with no human inside would ever be said to be out of control unless it is malfunctioning. It is under the control of the software and software designer, and most likely could be remotely controlled, operated, and monitored in a variety of ways that would satisfy such a legal requirement.

On the other hand, we may adopt a very strict definition of control would require a driver to have hands on the wheel and/or feet on the brakes, or at least ready to do so in a moment's notice. Legacy vehicle codes use the language "actual physical control" to describe what it means to operate. By dropping that language, it seems SB 169 recognizes that control may not require physical contact between a human person and a car, but seems to suggest that legislators assumed a type of control where a human operator could instantaneously assume actual physical control at any time.

It is also critical that in SB 169 automatic mode occurs if the human driver gives up control OR monitoring, but not necessarily control AND monitoring. In other words, if you watch your car control itself, you are in automatic mode. This is the biggest problem, because there are at least a few 2014 vehicles that would suddenly be illegal to drive (or monitor the driving of) if SB 169 becomes law.

Here are three vehicles that could be banned from Michigan under this law:

1. The Acura RLX

For about $12,000 over the base model, the 2014 Acura RLX can be purchased with the Advance Package. This slew of driver assistance, convenience, and comfort features includes advanced adaptive cruise control with low-speed following, collision mitigating breaking, and lane keep assistance. The combination of these features in highway driving would allow highway driving with minimal if any monitoring by a human operator.

2. The Mercedes Benz S-Class
Front view 2014 Mercedes-Benz S-Class

Mercedes offers the 2014 S Class with an optional $2,800 Driver Assistance Package that allows the car to be placed in [what SB 169 would call] automatic mode in stop & go highway traffic. The automatic mode can also be used at high speeds, though only for 15 seconds at a time.

3. The Infiniti Q50

Infiniti's 2014 Q 50 can be purchased with an optional $3,200 Technology Package that includes full speed intelligent cruise control, active lane control, and forward emergency braking. Infiniti does not label this automated driving, but it is hard not to interpret this combination of features as providing an automatic mode, or what NHTSA would define as Level 3 automation, limited self-driving. A New York Times reviewer observed, "I found myself driving the Infiniti on surprisingly long highway stretches without touching the accelerator, brake pedal or steering wheel."

From would I can tell, it wouldn't be illegal to OWN these vehicles in Michigan, but it would be illegal to use these optional features. Maybe Ford or GM should point out that they are planning to bring similar technologies to market within the next few years.