skip to primary navigationskip to content

Kara Kockelman: Abstract and background reading

                                                Title image Kara Kockelman

 

SHORT ABSTACT:

Large-scale modeling of the future of shared autonomous vehicles (SAVs) is necessary to appreciate various model assumptions and policy implications. Geographical constraints on an SAV fleet, operational modifications to accommodate trip request aggregation at stops, and the inclusion of parking search behavior in large regions are studied to identify potential benefits. Artificially increasing trip density within a region or at stops is expected to help improve fleet performance, but the question remains - by how much? Similarly, the bias in fleet metrics from not including parking behavior is also worth quantifying. The presentation will discuss the use of POLARIS and MATSim to quantify such policy effects for different regions and varying samples of trips simulated.

READ MORE:

With autonomous vehicles (AVs) still in the testing phase, researchers and planners rely on simulation to explore shared AV (SAV) fleet operations and system-design strategies. SAV operations with and without dynamic ride-sharing (DRS) or “pooling” options across five Chicago-area geofences are compared here, along with pickup and drop-off stop (PUDO) aggregation (to help merge riders) and curb-use restrictions on busy streets across the Bloomington, Illinois and Minneapolis-St Paul (MSP) regions. The geofences include the central business district of Chicago, the city’s formal boundaries, the region’s suburban core, its exurban core, and the entire 20-county region. 

Results demonstrate how service area limitations lower SAV response times, lower system-wide VMT across all modes, and ensure rather uniform response times over space. 

As expected, SAVs perform best in areas with high demand and with DRS active.  Inside Bloomington, various PUDO spacings and trip-demand densities were studied, also using the agent-based simulation model POLARIS. Results suggest that greater PUDO spacings and higher SAV-use levels increase SAV occupancies marginally, while lowering VMT notably, as compared to door-to-door SAV fleet operations without DRS or PUDOs. A quarter-mile PUDO spacing is recommended in downtown regions to keep walking trips short and demand relatively high. At 0.25 mi PUDO spacings (thoughtfully placed, using origin and destination clusters), travelers walked less than 2 min at either trip end, on average, while 0.5 mi spacings lead to about 3.5 min of walking. It is also important to prepare for queuing areas at PUDOs in settings of high trip densities, to limit curbside congestion. 

The MSP study used the MATSim model to track 2% and 5% of the 7-county region’s 9.5 million daily person-trips and 20% of the Twin Cities’ trips. Results suggest the average SAV in this region can serve at most 30 person-trips per day with less than 5-minute average wait time, thereby replacing about 10 household vehicles but generating 13% more vehicle-miles traveled (VMT). With dynamic ride-sharing (DRS), SAV fleetwide VMT fell almost 20%, on average, while empty VMT (eVMT) fell by 26%. While eVMT and wait times are relatively high (averaging 22.5% and 11.5 min) during peak times of day, they fall significantly in the PM peak if DRS is offered and actively used. Compared to idling-at-curb scenarios, parking-restricted scenarios (not allowing parked SAVs on the busiest streets) generated 8% more VMT across all four companion scenarios. Relying on 52 mi/gallon hybrid electric SAVs was estimated to lower travelers’ energy use by 21% and reduce tailpipe emissions by 30%, assuming no new or longer trips. A 106 mi/gallon equivalent battery-electric fleet does much better by lowering energy use by 64%.  For more details on these topics and highly related AV papers by Dr Kockelman’s research team, please visit https://www.caee.utexas.edu/prof/kockelman

 

BACKGROUND READING

The following three papers are helpful to understand this topic: 

(1) A System of Shared Autonomous Vehicles for Chicago: Understanding the Effect of Geofencing the Service (K Gurumurthy, J Auld and K Kockelman)

https://www.caee.utexas.edu/prof/kockelman/public_html/TRB20PolarisGeofencing.pdf

(2)  Shared Autonomous Vehicle Fleet Performance: Impacts of Parking Limitations and Trip Densities (H Yan, K Kockelman and K Gurumurthy)

https://www.caee.utexas.edu/prof/kockelman/public_html/TRB21MSPSAVs.pdf

(3) How Much Does Greater Trip Demand and Aggregation at Stops Improve Dynamic Ride-Sharing in Shared Autonomous Vehicle Systems? (K Gurumurthy and K Kockelman)

https://www.caee.utexas.edu/prof/kockelman/public_html/TRB21savstopaggregationbloomington.pdf