Ricerc@Sapienza

The analysis of oscillating systems controlled by mechatronic devices, relies classically on differential equations, and the problem is frequently attacked in the frequency domain, for linear systems based on more conventional controls, or in the time space-state formulation to include also nonlinearities. In this paper we are faced with a system that exhibits memory effects. These are borne because of the presence of added mass and damping that are due to the rigid body motion coupling with the surrounding water.

Feedback control techniques for vibrations suppression are widely investigated in the technical literature. However, the most powerful method in this context, the optimal control theory, despites its generality and flexibility related to the user defined cost function, meets serious engineering limitations. This is due to the lack of a feedback law, because of the intrinsic formulation of the Pontryagin optimality criterion. This paper proposes a novel method of control that, still maintaining the structure of the optimal control technique, provides a feedback control.

Vertical landing is becoming popular in the last fifteen years, a technology known under the acronym VTVL, Vertical Takeoff and Vertical Landing [1,2]. The interest in such landing technology is dictated by possible cost reductions [3,4], that impose spaceship’s recycling. The rockets are not generally de- signed to perform landing operations, rather their design is aimed at takeoff operations, guaranteeing a very high forward acceleration to gain the velocity needed to escape the gravitational force.

The autonomous vehicle is one of the greatest challenges in modern vehicle design. This paper proposes a new method of control named FLOP, Feedback Local Optimality Principle, recently proposed by the authors. The method, starting from the Pontryagin's theory, introduces a new optimality principle that minimizes a sequence of individual functionals with the chance of a direct feedback control.

The autonomous vehicle is one of the greatest challenges in modern vehicle design. This paper proposes a new technique to define the optimal trajectory in a feedback form for an autonomous car, moving on a track. The algorithm defines the trajectory taking into the account the vehicle dynamic instead of kinematic constraints, leading to a more robust path. The technique is also used to control the vehicle in feedback, providing the optimal maneuvers to track the defined path.

A novel indirect variational optimal control theory is proposed for integral-differential equations of motion. This algorithm is applied to an engineering application: the control of a an underawater autonomous vehicle's 2-Dof lifting surface. The variational approach proposed is an extension of the classical Potryagin optimal control theory which normally refers to differential equations. The control has been extended by developing a novel integral MPC technique. Numerical results show good performace of the optimal control proposed compared with the standard LQR method.

The autonomous vehicle is one of the greatest challenges in modern vehicle design. This paper proposes a new method of control named FLOP, Feedback Local Optimality Principle, recently proposed by the authors. The method, starting from the Pontryagin's theory, introduces a new optimality principle that minimizes a sequence of individual functionals with the chance of a direct feedback control. The theory is applied to the steering and traction control of a two-wheeled vehicle, showing the ability of tracking a given trajectory and obstacles avoidance in a rather complex environment.

In this paper the problem of optimizing the output regulation of a weakly dual redundant plant with multiple actuators is addressed from a control theoretic viewpoint. When a system is underactuated, only subsets of the outputs can be independently controlled, while the remaining ones are constrained. With a specific focus on the asymptotic output tracking problem for MIMO systems with periodic references, we investigate the connection between the overall optimal input and the individually optimal controllers that lead to a perfect tracking of each output component.

In this paper the problem of optimizing the regulation of a weakly dual redundant plant with multiple actuators is addressed. When the system is underactuated, only a subset of the outputs can be independently controlled. In this regard, the main objective of the paper pertains the optimization of the steady-state performance of the plant. A connection between the overall optimal input and the inputs that provide a perfect reference tracking of the full controllable square subsystems is established.

In recent years Reinforcement Learning (RL) has achieved remarkable results. Nonetheless RL algorithms prove to be unsuccessful in robotics applications where constraints satisfaction is involved, e.g. for safety. In this work we propose a control algorithm that allows to enforce constraints over a learned control policy. Hence we combine Nonlinear Model Predictive Control (NMPC) with control-state trajectories generated from the learned policy at each time step. We prove the effectiveness of our method on the Pendubot, a challenging underactuated robot.