Distributed, Private, and Derandomized Allocation Algorithm for EV Charging
For researchers in distributed resource allocation, this provides an incremental improvement by derandomizing an existing algorithm and extending its applicability to new utility function types.
This work derandomizes a stochastic AIMD algorithm for distributed resource allocation, extending it to handle concave, non-monotone, and sigmoidal utility functions. Simulations for EV charging show improved efficiency over the stochastic version.
Efficient resource allocation is challenging when privacy of users is important. Distributed approaches have recently been used extensively to find a solution for such problems. In this work, the efficiency of distributed AIMD algorithm for allocation of subsidized goods is studied. First, a suitable utility function is assigned to each user describing the amount of satisfaction that it has from allocated resource. Then the resource allocation is defined as a total utilitarianism problem that is an optimization problem of sum of users utility functions subjected to capacity constraint. Recently, a stochastic state-dependent variant of AIMD algorithm is used for allocation of common goods among users with strictly increasing and concave utility functions. Here, the stochastic AIMD algorithm is derandomized and its efficiency is compared with the stochastic version. Moreover, the algorithm is improved to allocate subsidized goods to users with concave and non-monotone utility functions as well as users with Sigmoidal utility functions. To illustrate the effectiveness of the proposed solutions, simulation results is presented for a public renewable-energy powered charging station in which the electric vehicles (EV) compete to be recharged.