Ongoing research addresses the issues related to adaptive information processing systems and intelligent embedded systems and, in particular, “Active and Passive Wireless Sensor Networks and embedded intelligent systems”, “Adaptive Systems”, “Application-level methodologies for the analysis, synthesis and diagnosis of Information Processing Systems” and “Quality Analysis and Composite Systems Design”.

Research activities are carried out both at academic and industrial level. More in detail,

Active and Passive Wireless Sensor Networks and Intelligent Embedded Systems

The research addresses methodological and application-related aspects of that class of embedded systems known in the literature as Wireless Sensor Networks both passive (RFId based) and active. Active wireless sensor networks are subject of deep analysis at different HW and SW levels. More specifically, aspects related to energy harvesting (through adaptive embedded systems able to maximise the energy transfer between small solar panels/Peltier cells and rechargeable batteries/supercapacitors), energy management (by means of energy-aware local routing protocols and unit management, adaptive sampling, dynamic data accuracy acquisition) and integration of hybrid wired/wireless monitoring systems are envisaged. Particular attention is devoted to credible applications in harsh environments. A sophisticated automatic, adaptive, sustainable and reliable wireless monitoring system for marine environment has been deployed in Queensland, AUS, November 2007. Current applications refer to the design and implementation of intelligent embedded systems for rockfall collapse forecasting with traditional and microacoustic emissions with data/command communication, storage and aggregation in a control room (deployments: S.Martino Mountain April 2010; Torrioni di Rialba July 2010) and landslide monitoring (Torrioni di Rialba deployment: July 2011).

Adaptive Systems

The research focuses on theoretical, implementation and application related aspects of computational intelligence-based systems with a specific focus on classifiers. Relationships among accuracy, confidence, robustness and computational complexity of hierarchical classifiers both during training and operational phases have been studied. Results and developed methodologies allow the scholar for understanding underlying structural and functional properties as well as for solving the performance/constraints trade-off. Analysis and synthesis tools have been designed and developed. Most of methodologies are rather general and, as such, can be easily extended to other computational paradigms. Currently, the attention is on the development of advanced non - stationarity detection indexes, the design of adaptive classifiers reacting just in time to the environmental changes (e.g., due to ageing, faults), the relationship between k and n in k-NN classifiers and estimating model performance through virtual LOO & k-fold CV.

Application-level methodologies for the analysis, synthesis and diagnosis of Information Processing Systems

The ongoing research addresses application level properties of the computational flow associated with an embedded system and its relationships with low level design aspects. The developed methodologies and theories for analysis, synthesis and diagnosis, based on statistical and Randomised Algorithms approaches, allow us for fully characterising the nature of the computation with an acceptable complexity. Such information can be used to measure the robustness of the application (analysis phase), provide design guidelines (synthesis phase) and identify the test injection points and test vectors in analog devices (diagnosis phase). A theory on probably approximated correct computation, i.e., theoretical framework for embedded systems in an uncertainty-affected environment has been provided.

Quality Analysis and Composite Systems Design

The research is focusing on theoretical and application-oriented aspects of Composite Embedded Systems (systems encompassing neural networks and more traditional information processing techniques). Among the most relevant applications we identify the quality analysis for industrial process, classification, image and signal processing, prediction, identification and control ones. The goal of the research is to formalise and develop methodologies for the automated selection and configuration of optimal embedded systems, where the optimality concept is tailored to the particular application needs and constraints (e.g., accuracy, real time, algorithm complexity).