Find below the list of projects which I am responsible.


Engineering Applications of Machine Learning

Machine learning has had in the recent past a considerable amount of research work both from academia and industry. The goal of the present project is to build upon solutions from the computer science community and bridge the gap from their conception to real-world engineering applications. Particular interest is devoted to (i) monitoring of equipment using heterogeneous field data; (ii) advanced control and identification of dynamical systems; (iii) predictive modeling.

Advanced Nonlinear Control, Estimation and Identification

The goal of the project is to devise novel, optimal, and computationally efficient solutions for advanced control, estimation, and identification. Theoretical aspects, as well as practical computational implementations, are sought. In particular, receding-horizon and intelligent approaches are investigated, towards novel nonlinear methods that take into account the limitations of the systems explicitly and are, at the same time, computationally efficient to run at high-frequency rates.


Efficient Real-time Muscle Deflection Tracking based on Myokinetic Sensors and Artificial Neural Networks for Prosthetic Control

Myokinetic sensors have been very recently proposed and studied for the purpose of measuring the intent of amputees in order to provide information for actuating the end effectors of prosthetic devices. It is however very difficult to establish nonlinear mappings which are able to perform soft-sensing from the measured magnetic information of myokinetic sensors and the setpoints for target tracking in the prosthetic devices. In the present project the goals are (i) to build data-driven soft-sensors which are able to translate the myokinetic sensor information to usable data for driving the prosthetic device, as current techniques are based on very complex models which in turn are not suitable for neither real-time implementation or embedded solutions; (ii) to design novel hardware implementations of soft-sensors for myokinetic transducers used in prosthetic control based on artificial neural networks. We build up on previous works of the participants towards achieving (i) and (ii) in a synergistic international collaboration effort. We expect to obtain precision and energy-aware optimized solutions which will enable better and fine-grained prosthetic control, providing unique features in the field of myokinetic sensors, related embedded solutions and fundamental machine learning methods..

CNPq Universal 430395/2018-3; ERC – CONFAP – CNPq 400119/2019-6.