Craig and I have officially started our company, Activimetrics LLC. Our goal is to use the company as a platform to promote activity-based modeling approaches, but our target market is not as narrow as we first thought. My thinking about what we can do with our skills and experience has been broadened as a direct result of responding to the Markets for Good Data Interoperability Challenge.
Activity analysis in transportation is focused on looking at what people do, rather than how they get around. Traditional transportation analysis is trip based: there are x thousands of trips from Zone A to Zone B in the morning peak, and so on. While each trip has a purpose of sorts, there is no way to quantitatively analyze something like:
Suzy Q leaves her house and hits the espresso bar for her daily latte, then drives to work. She eats her lunch at her desk every day, and then leaves early to pick her daughters from school at 3:45pm. There are after school activities for both girls every day but Thursday. Her spouse drops the girls off in the morning, and then goes to work, returning home by 7pm every evening.
Activity analysis requires far more data than trip based modeling, and the modeling work can be a black hole of ever increasing complexity. On the other hand, almost everybody is carrying around an iPhone or an Android device, and is probably already collecting sufficient data for the most detailed activity based model imaginable. So we started our company to try to leverage smart devices and activity-based modeling techniques to work towards a better transportation system.
As a daily bicycle commuter, my personal bias is to work towards promoting active modes of transportation, like bicycling, walking, skating and skateboarding. Aside from the well known health and environmental benefits, I personally prefer walkable cities over drivable ones. But as a transportation engineer, I am well aware that the car is incredibly awesome. For very little expense and for zero effort, a car can propel a driver, passengers, and a large load miles and miles at very high speeds. The economic benefits of this fact should never be discounted. Replacing a car-biased system with a walking/biking one carries with it the danger of increasing the net travel costs of a large portion of the population.
So we have to be careful how we go about this. We have to look at data and carefully evaluate the alternatives. We need to be able to evaluate situations where active modes of travel are a strong net benefit, versus situations where the costs outweigh the benefits, or where the car will remain the dominant solution regardless of how much is invested in walkable infrastructure.
Which brings us back to data and data sources, and to the original topic of this post, data interoperability. We’ve been looking at cellphones as the ultimate data collection devices, but cellphones don’t collect aggregate information very easily. For that we have loop detectors embedded in freeway pavement—Caltrans calls their network of loops the vehicle detection system (VDS), and they have a website that allows one to download years of data for thousands of loop detector stations.
Loop detector systems cover only the urban parts of freeways. For everything else, there is a data set called the Highway Performance Monitoring System (HPMS), a data set collected all over the US as a requirement for receiving federal aid to build roads. This data set ostensibly covers every road in the US that the average person might drive on. It excludes private roads, and Bureau of Public Lands roads, but otherwise, every bit of road is built with federal aid and so must be accounted for in the HPMS data.
The problem is that the HPMS system in California is not geocoded, and is very difficult to work with directly. We took a look at merging the data with OpenStreetMap, but ran into difficulties almost immediately. In California’s HPMS data, a record for a road will identify the city and county where it is located, the road’s name and “from” and “to” links. But there are crazy cases in every city that seem designed to confuse. For example, is North West Street in Fresno supposed to be Northwest Street, or does it mean N West Street? If one compares it to the Census or to OpenStreetMap, what happens if one codes it as Northwest, while the other codes it as N West, but both are referring to the same thing? If you take a gander at Google maps, you’ll find that there is a N West Ave that Google also calls NW Ave and a NW E Street!
So these are the thoughts brewing in our start-up hive mind. I wrote this up, and had some great ideas for a concluding section, but then I hopped on my bike, rode about two miles, crashed, and broke my arm.
Now, 4 weeks later, my arm still hurts, and I have entirely forgotten my big finish.
But the issues are still relevant, and still brewing in our start-up discussions. And this web log has been sorely neglected.