Section: New Results

Global optimization for online resource allocation

Participant : Jean-Marc Lasgouttes.

As part of the Mobility 2.0 FP7 project, we have considered the possibility to allocate charging stations to Full Electric Vehicle (FEV) users in a way that, instead of merely minimizing their travel time, tries to improve the travel time for the whole community.

The aim of the global optimization algorithm is to pursue the minimization of the mean squared travel time encountered by each user. Our setting can be seen as a resource allocation problem, known as the “Transportation Problem” in Operational Research literature. It is solvable using several algorithms, among which the simplex algorithm. Unfortunately, these algorithms are not usable here for two reasons:

• The allocation of slots to the users is done online, when the user does a request. It is not possible to wait until all the users are known before doing the allocation;

• The complexity of these algorithms is very high, especially since, due to the effect of range limitations, each request has different characteristics.

We therefore present here a simplified approach, which should be fast enough to scale for large systems. The principle of the algorithm is to penalize the cost for the user with an approximation of the extra cost incurred to future users who compete for the same resource (a charging or parking slot). Since the implications can be intricate, we only consider a first order effect.

Our work in the Mobility 2.0 project has been briefly described in [37] .