Current students


Section: Systems and Control

Major Research topic:
A distributed optimization framework for predictive control of multi-agent systems affected by stochastic uncertainty.

The latest technological advances have enabled the introduction of engineering systems with enhanced capabilities in many fields, such as manufacturing, energy, and transport. Such development went hand in hand with an increase of complexity, that led to deep transformations and is now calling for novel efficient solutions.

Most of the systems involved are aggregates of interleaved components, that must be coordinated to achieve optimal performance while accounting for heterogeneous constraints. In the energy sector, for example, traditional consumers are taking an active role as prosumers as the use of renewable sources grows. In transport, instead, new challenges are arising due to the increasing number of electric vehicles that can both use and store energy (acting as batteries for balancing purposes). 

As systems grow in complexity, their design becomes more challenging. For this reason, recent works in distributed optimization focused on developing scalable algorithms to find (sub-optimal) solutions to problems with such multi-agent nature.  

Unfortunately, most of the approaches neglect the presence of the additional uncertainty typically affecting those systems. Those few that, instead, explicitly account for it, are likely to end up with a conservative solution unless they exploit feedback from the system operation to extract some additional knowledge on the uncertainty. In fact, such information could be used to refine the values of the decision variables and repetitively update the solution, according to a model predictive control paradigm with a receding-horizon strategy.  

To this end, our research will address the extension of distributed optimization to feedback control of large-scale multi-agent systems. More precisely, it will investigate the integration of Stochastic Model Predictive Control (SMPC) and stochastic filtering in the mentioned framework, possibly tackling issues arising from the computational complexity of the optimization problems and/or the need to preserve the privacy of local information.   

The developed approaches will be validated by extensive simulations on case studies in the energy domain, where multiple residential prosumers equipped with photovoltaic panel installations, batteries, and programmable loads are aggregated to offer balancing services to the grid.