Ricerc@Sapienza

In optimal control problems, once the model, the boundary conditions and the constraints are fixed, the result depends obviously on the choice of the cost index and, in particular, on the weights assumed for the variables considered. The weights take into account the different mutual influence of the elements included in the cost index; therefore different choices yield to different control strategies.

This paper presents a decentralized Model Predictive Control (MPC) for Plug-in Electric Vehicles (PEVs) charging, in presence of both network and drivers' requirements. The open loop optimal control problem at the basis of MPC is modeled as a consesus with regularization optimization problem and solved by means of the decentralized Alternating Direction Method of Multipliers (ADMM).

The paper addresses the problem of optimal
control design in presence of singular solutions. For this case, a
procedure for avoiding the integration of the costate dynamics
is proposed, giving the conditions under which the costate
can be directly computed, under controllability condition for
the dynamics, and presenting an approach for extending this
property by a dynamic extension. The procedure is here
described for a single input systems and for the case in which
the first step of the iterative procedure is sufficient to get the

The problem of minimizing the transfer time along a given Cartesian path for redundant robots can be approached in two steps, by separating the generation of a joint path associated to the Cartesian path from the exact minimization of motion time under kinematic/dynamic bounds along the obtained parameterized joint path. In this framework, multiple suboptimal solutions can be found, depending on how redundancy is locally resolved in the joint space within the first step.

In this paper, the performance appropriateness of population-based metaheuristics for immunotherapy protocols is investigated on a comparative basis while the goal is to stimulate the immune system to defend against cancer. For this purpose, genetic algorithm and particle swarm optimization are employed and compared with modern method of Pontryagin's minimum principle (PMP).