February 22, 2016

Paper Review: “Using An Activity-Based Model To Explore Possible Impacts Of Automated Vehicles”

Parastoo Jabbari

In a paper published last year, staff from the Puget Sound Regional Council (PSRC) examined the impacts of automated vehicles (AVs) on travel behavior using their activity-based travel model for the Seattle region. The authors concluded “improvements in roadway capacity and in the quality of the driving trip may lead to large increases in vehicle-miles traveled, while a shift to per-mile usage charges may counteract that trend”. To predict the impacts of AVs, they also recommended updating travel models to account for shared-ride, taxi modes and the influence of multitasking.

The authors created four scenarios to model the potential effects of AVs on demand through modification of capacity, perceived travel time, parking cost, and operating cost. They applied PSRC’s activity-based travel model, SoundCast, to simulate the effects. Since the original model was not developed based on AVs, modifying some of model’s assumptions was necessary.

1. Their first scenario assumed that AVs use the existing facilities, and that automation yields a 30% increase in all freeway and major arterial capacities.
2. The second scenario assumed a lower perceived cost of travel time for AVs compared to conventional cars, along with the capacity increase from the first scenario. This scenario assumed that important trips would happens in AVs, and the trip-based value of time would be reduced by 65%.
3. In addition to the previous assumptions, the third scenario also assumed that all cars are self-driving and parking costs are reduced by half.
4. The final scenario envisioned the situation in which all autos are AVs and AV ownership is no longer necessary. It assumed that users are responsible for the total cost of driving such as vehicle and infrastructure maintenance, road construction and etc. This cost was estimated at $1.65 per mile.

The most essential point of this study is that it clearly demonstrates the sensitivity of demand to AVs’ impacts. The results of travel model show that reduction in perceived travel time and parking costs will generate great additional demand which will counteract the effects of increased capacity. Quantifying the amount of changes caused by AVs, even though it is based on assumptions, depicts how severe the impacts of AVs on demand can be. It is also reassuring that the authors’ detailed analysis using a disaggregate modeling approach produced results that are generally consistent with other analyses based on aggregate data. There are many other effects of automation that can lead to additional demand such as perception of multitasking. This clearly establishes the need for a comprehensive study to address the many possible factors that link automation to travel demand.

One limitation of this study is the validity of basing travelers’ usage of AVs on observations of behavior in conventional vehicles. The assumptions around capacity, perceived time and parking cost changes are based on current users’ behavior, not possible future options. As the authors state, “people are going to behave based on the options available to them and on the constraints they face in their daily lives”. For example, for parking costs, it is assumed that since the number of options are increased, the costs would be lesser. In reality, it is possible that the attitudes of users toward parking may change completely: they might prefer to send the car home or to pick up groceries or similar activities.

As a conclusion, the approach of the paper and methods is generally sound. Since there is little data about AVs right now, using existing models and adjusting them is the best approach available. The limitations and implications of the research were explained convincingly. Over time, we hope that the authors and others can continue to refine their assumptions around the model’s parameters as new knowledge accumulates.