Home page‎ > ‎

Research activities

First scientific topic: Human-Robot interaction


The main goal of this activity was the study of human-robot interaction in an industrial environment. I analyzed the state of the art literature about behavioral models, especially for behavior recognition and motion prediction. I developed a method for the definition of cluster centroids based on the p-median localization problem that can be used in conjunction with the Earth Mover’s Distance (EMD) for clustering and classification problems. We applied it for trajectory clustering, that is the partitioning of a set of observed trajectories into subsets, called clusters, so that observations in the same cluster are similar in some sense. We proposed an algorithm for on-line classification with incremental portions of trajectories and we expanded our approach in order to take different features besides spatial features into account, like  the direction chosen to cover the trajectory, the orientation or the posture of the target, and to represent 3-D motion.  We tested our methodology on a real benchmark data set, the CAVIAR data set.

The results are presented in the papers:

§  F. Boem, F. A. Pellegrino,  G. Fenu, and T. Parisini, Trajectory clustering by means of Earth Mover's Distance. In Proc. 18th IFAC World Congress, pp.4741-4746, 2011.

§  F. Boem, F. A. Pellegrino,  G. Fenu, and T. Parisini, Multi-feature trajectory clustering using Earth Movers Distance. In Proc. 7th  IEEE Conference on Automation Science and Engineering, pp.310-315, 2011.

and in a poster presented during the SIDRA congress Automatica.it 2011. I presented these works personally.
Second scientific topic: Distributed Fault Diagnosis

After the analysis of the state of the art about fault diagnosis, distributed and networked systems, consensus and distributed estimate, we developed a continuous-time distributed fault detection and isolation methodology for nonlinear uncertain possibly large-scale dynamical systems and then we analyzed the case of partial state measurements for discrete and continuous time systems. The system being monitored is modeled as the interconnection of several subsystems and a divide et impera approach allowing overlapping decomposition is used. The local diagnostic decision is made on the basis of the knowledge of the local subsystem dynamic model and of an adaptive approximation of the interconnection with neighboring subsystems. In the last year I proposed a delay compensation strategy, allowing managing delays and packet losses in the communication network between the Local Fault Diagnosers, and I developed a novel consensus-based estimator with time-varying weights, permitting to improve detectability skills in the case of variables shared among more than one subsystem. In the consensus protocol, at each step each agent uses only the information given by the communication link and the agent which are more reliable at that time. The convergence of the proposed estimator is demonstrated without any assumption on the communication network topology and analytical conditions for detectability are derived, showing that the novel consensus-based estimator improves detectability performance and that the fault detection time is reduced in all cases, with the presence of delays and without delays.

The results are presented in the papers:

§  F. Boem, R. M. G. Ferrari, T. Parisini, and M. M. Polycarpou,  A Distributed Fault Detection Methodology for a Class of Large-scale Uncertain Input-output Discrete-Time Nonlinear Systems. In Proc. 50th IEEE Conf. on Decision and Control and European Control Conference, Orlando, Florida, pp.897-902,  2011.

§  F. Boem, R. M. G. Ferrari, and T. Parisini , Distributed Fault Detection and Isolation of Continuous-Time Nonlinear Systems, European J. of Control, (ISSN:0947-3580), pp.603- 620, Vol. 17, 2011

§  F. Boem, R. M. G. Ferrari, T. Parisini, and M. M. Polycarpou, Distributed Fault Diagnosis for Input-output Discrete-Time Nonlinear Systems: New Results on Fault Isolation. Technical report.

§  F. Boem, R. M. G. Ferrari, T. Parisini, and M. M. Polycarpou, Distributed Fault Diagnosis for Input-Output Continuous-Time Nonlinear Systems. In Proc. Safeprocess Conference, Mexico City, Mexico, pp.1089- 1094,  2012.

§  F. Boem, R. M. G. Ferrari, T. Parisini, and M. M. Polycarpou, Distributed Fault Detection for Uncertain Nonlinear Systems: a Network Delay Compensation Strategy. American Control Conference, Washington, 2013 (submitted to).

I presented the first work personally at the CDC-ECC 2011 conference. At the moment,  we are analyzing the case of no perfect clock synchronization between sensors and local diagnosers  and the case of multi-rate systems.


I cooperate with Danieli Automation for the application of the proposed fault diagnosis architecture on a real industrial case.
Third scientific topic: Distributed Estimation

We proposed a novel distributed estimator for tracking an unknown noisy time-varying signal in a sensor network,  where the filter coefficients are updated locally in order to minimize both the variance and the mean of the estimation error by means of a Pareto optimization problem. We then investigated the effectiveness of the proposed estimation approach for diagnosis purposes with sensor networks. The results are presented in the following publications:

§  F. Boem, Y. Xu, C. Fischione, and T. Parisini, A Distributed Estimation Method for Sensor Networks Based on Pareto Optimization. In Proc. 51st IEEE Conf. on Decision and Control, Maui, Hawaii, 2012.

§  F. Boem, Y. Xu, C. Fischione, and T. Parisini,Distributed Fault Detection using Sensor Networks and Pareto Estimation, European Control Conference, Zurich, 2013 (submitted to).

and I presented this scientific activity during the SIDRA congress Automatica.it 2012.