Dynamic Ride-Sharing

The DRS algorithm implemented here is a heuristic to facilitate better use of empty seats in SAVs while limiting the delay experienced by each traveler in the SAV. The heuristic attempts to match incoming requests to available vehicles that are either idling or performing a pickup, dropoff, or repositioning trip in the direction of the incoming request’s destination. This directionality is quantified as the angle between the lines joining the current and proposed trips based on available coordinates. This angle is a succinct proxy for the extent of detours that may be allowed while maximizing pooled trips, and a threshold is provided as an input to the model. Additionally, each traveler’s approximate delay (based on the estimated initial routing time without detours) is measured throughout their trip to avoid new travelers from being added to the SAV when any traveler experiences a delay beyond the predefined absolute delay or the predefined percentage delay while en route. Both absolute and percentage delays are important since short trips are sensitive to percentage delays while longer trips are sensitive to absolute delays.

Prior code used a simple utility equation to assign pooled or empty TNC vehicles to a ride request based solely on the following parameters: ASC_not-pooled, β_pooled, and β_not-pooled. The utility equations multiplied the betas by travel time from the closest (empty or pooled) vehicle to the customer’s origin. Work is ongoing to restore this ability to choose a pooled or empty TNC vehicle instead of assuming all riders use a TNC vehicle. It is based on work by Lavieri and Bhat (2019). It includes the travel time but information on the traveler to attempt to reflect that some riders will likely avoid pooling.

Important

Please refer to the the following paper for more details on the WTP strategy, adapted from Table 9 (available here):

Gurumurthy, K.M., and Kockelman, K.M. 2022. Dynamic Ride-Sharing Impacts of Greater Trip Demand and Aggregation at Stops. Transportation Research Part A 160: 114-125. https://doi.org/10.1016/j.tra.2022.03.032

Pooling Model

The baseline assumption when setting the fleet model parameter DRS_FLAG = true is that 100% of riders are willing to share a ride with another stranger (subject to max detour constraints that are advertised in advance). Pooling models modify this assumption. There are several pooling models to choose from: - zonebased - WTP - Clemson university (CU) - UW

Pooling Model Resources

Important

Please reference the following paper for more details on the WTP strategy:

Gurumurthy, K.M. and Kockelman, K.M. (2018). Modeling Americans' Autonomous Vehicle Preferences: A Focus on Dynamic Ride-sharing,
Privacy, and Long-Distance Mode Choices. Technological Forecasting & Social Change 150. https://doi.org/10.1016/j.techfore.2019.119792

Please reference the following paper for more details on the zonebased strategy:

Dean, M.D., Gurumurthy, K.M., de Souza, F., Auld, J., Kockelman, K.M. (2022). Spatial Variation in Shared Ride-Hail Trips &
Factors Contributing To Sharing. Journal of Transport Geography 91, 102944. https://doi.org/10.1016/j.jtrangeo.2020.102944