Loading [MathJax]/extensions/TeX/upgreek.js
Quantification on the efficiency gain of automated ridesharing services | IEEE Conference Publication | IEEE Xplore

Quantification on the efficiency gain of automated ridesharing services


Abstract:

Ridesharing services often require constant rebalancing vehicle supply in order to meet passenger demand across the transportation network. We compare the cost of rebalan...Show More

Abstract:

Ridesharing services often require constant rebalancing vehicle supply in order to meet passenger demand across the transportation network. We compare the cost of rebalancing between two different methods for controlling the vehicle flows: (1) direct control, which models on-demand dispatchable vehicles such as autonomous vehicles and (2) indirect control based on price differences, which models human drivers as in the current ridesharing scheme. We propose a metric that quantifies the efficiency gain of automated ridesharing (i.e., direct control) based on the maximum difference between the rebalancing cost of two methods. The benefit of the proposed metric is that it is independent of the actual demand and only relies on properties of the transportation network. We present a set of numerical tools for computing the metric. For a general graph, the metric can be computed using an efficient local search method called the difference-of-convex algorithm (DCA). Numerical experiments on a practical graph (constructed from a pricing map for the Washington, DC area) show that the DCA often converges within a few iterations. For fully connected and symmetric graphs, the metric can be computed from an equivalent convex program. Moreover, the convex program adopts a simple closed-form optimal solution.
Date of Conference: 24-26 May 2017
Date Added to IEEE Xplore: 03 July 2017
ISBN Information:
Electronic ISSN: 2378-5861
Conference Location: Seattle, WA, USA

References

References is not available for this document.