|
As transportation is increasingly electrified, understanding travel and charging choices of electric vehicle drivers and users of shared electric mobility systems is crucial for service planning and operations. Ultimately, quantitative models of users’ behavior enable the estimation of local and regional demand impinging on transportation and electricity networks and the development of approaches to manage the demand so that such networks can operate efficiently. Developing realistic models of individual behavior to use as “users’ digital twins” in planning and operations is challenging enough, due to the inherent uncertainties of how individuals make everyday choices and their heterogeneity in approaching decisions. Even more challenging is to attempt to anticipate how users might respond to new technologies and changes in infrastructure or services. In this case choice data from systems’ operations or “revealed preference” users’ surveys may not exist, be available only from early adopters, or relate to the system before the change. In this case, often the only option available is to generate new data using choice experiments. In the first part of this seminar, Dr. Daina will present pioneering work adopting these techniques to model electric vehicles driver's charging choices under smart charging. While in the near term e-mobility is and will be dominated by human drivers, on demand e-mobility services delivered by fleets of autonomous vehicles are on the horizon. Therefore, in the second part of this talk, he will present recent work on reinforcement learning approaches to smart charging management of shared autonomous electric vehicle fleets.
|
