Robert Cornelius Hampshire

Inventory rebalancing and vehicle routing in bike sharing systems

Bike sharing systems have been installed in many cities around the world and are increasing in popularity. A major operational cost driver in these systems is rebalancing the bikes over time such that the appropriate number of bikes and open docks are available to users. We combine two aspects that have previously been handled separately in the literature: determining service level requirements at each bike sharing station, and designing (near-)optimal vehicle routes to rebalance the inventory. Since finding provably optimal solutions is practically intractable, we propose a new cluster-first route-second heuristic, in which a polynomial-size Clustering Problem simultaneously considers the service level feasibility and approximate routing costs. Extensive computational results on real-world data from Hubway (Boston, MA) and Capital Bikeshare (Washington, DC) are provided, which show that our heuristic outperforms a pure mixed-integer programming formulation and a constraint programming approach.

Drivers overtaking bicyclists—An examination using naturalistic driving data

This paper demonstrates a unique and promising approach to study driver-bicyclist interactions from a drivers perspective by using in-vehicle sensory data from naturalistic driving studies. A total of 4789 events of drivers overtaking bicyclists were extracted from an existing naturalistic driving study in Michigan, United States. The vehicle lateral placement at the time of passing bicyclists was used as a surrogate safety measure. A number of factors were examined, including the lane marking type, the presence of a bike lane or paved shoulder, the presence of traffic, lane width, and driver distraction. Some notable findings include that (1) when a bike lane or paved shoulder was present, a dashed non-center line (i.e., a dashed line separating two lanes in the same direction) was associated with significantly less vehicle lane-crossing and closer distance to the bike lane/shoulder compared to a solid centerline; (2) an alarming 7.8% of the overtaking occurred when the drivers were distracted within five seconds prior to passing bicyclists. From a bicyclists perspective, that translates to one overtaken by a distracted driver for every thirteen times they are overtaken. In addition, drivers manipulating a cell phone were associated with significantly less vehicle lane-crossing when overtaking bicyclists. The results of this work could be potentially used by traffic engineers, policymakers and legislators to support the designs of better road infrastructures, education programs, policies, and traffic laws that aim to improve the safety of all road users. The quantitative results could also be potentially used as a baseline to develop and benchmark automated vehicle technologies on how to interact with bicyclists on the road.

Analysis of parking search behavior with video from naturalistic driving

The number of cars searching for parking, also known as “cruising,” is a risk factor linked to increased pollution and congestion and decreased road safety. Although the detrimental effects of cruising are known, the actual amount of cruising is unknown. A novel video data set of naturalistic driving is shown to provide reliable estimates of cruising behavior. The distribution of search start times, search distances, and search times is characterized. Cruising behavior variation between 109 different drivers is also reported on in the study, located in southeast Michigan. It was found that 30% of the drivers generated more than 70% of the meters cruised. This finding suggests that the search strategies of a few drivers disproportionately affect the many. These results facilitate the estimation of the number of vehicles searching for parking and the amount of pollution generated by cruising drivers. Researchers may also use these results to develop more realistic models of parking search and parking interventions. The results, based on video data, have implications for settings in which video data are not available. The results from this study can serve as input to a model that classifies GPS traces as cruising or not cruising.

The Mode Most Traveled: Transportation Infrastructure Implications and Policy Responses

Using the United States Census Public Use Microdata Sample (PUMS) dataset, we documented the severity of the disparity in commuting pattern across the contiguous US. The analysis was complemented by a more granular analysis with the Greater Pittsburgh area as the geographic area of focus. In addition to the locational variation in travel mode obtained using population estimates derived from the PUMS dataset, the dataset was utilized for a discrete choice model that generated detailed commuting profiles for the region’s workforce, showing statistically significant differences not only by socio-economic attributes but more importantly, by commuters’ place of abode. Policy levers that could address travel mode shift are discussed primarily with regards to changing population and its impact on transportation resources and the onset of fully autonomous vehicle in transportation networking companies’ space—a subject of key topical interest given the choice of the city as the test bed for Uber’s driverless ride sourcing services.

An Evaluation of Information Sharing Parking Guidance Policies Using a Bayesian Approach

Real-time parking occupancy information is critical for a parking management system to facilitate drivers to park more efficiently. Recent advances in connected and automated vehicle technologies enable sensor-equipped cars (probe cars) to detect and broadcast available parking spaces when driving through parking lots. In this paper, we evaluate the impact of market penetration of probe cars on the system performance, and investigate different parking guidance policies to improve the data acquisition process. We adopt a simulation-based approach to impose four policies on an offstreet parking lot influencing the behavior of probe cars to park in assigned parking spaces. This in turn effects the scanning route and the parking space occupancy estimations. The last policy we propose is a near-optimal guidance strategy that maximizes the information gain of posteriors. The results suggest that an efficient information gathering policy can compensate for low penetration of connected and automated vehicles. We also highlight the policy trade-off that occur while attempting to maximize information gain through explorations and improve assignment accuracy through exploitations. Our results can assist urban policy makers in designing and managing smart parking systems.