PhD research proposals

Cristiana Bolchini's research interests fall into two categories: dependable system design and design methodologies, and context-aware data design, integration and management. Interested PhD candidates can either discuss and propose topics in these reasearch lines, or refer to the following proposals.

Dependable embedded system design and design methodologies

Reliable Multi-Core Architectures: methodologies and tools
As the incidence and sensitivity of new technologies to both transient and permanent faults increase, the adoption of dependability-oriented techniques becomes more and more relevant, even in non-critical application scenarios. Therefore, even innovative, general-purpose architectures need suitable dependability-oriented design methodologies and tools. In this scenario, we focus our attention on multi-core architectures, thea expose inherent redundancy and (re)configuration features, typically expoited to improve performance. The PhD work will invetigate the definition of suitable hardware and software techniques to enable the on-line detection of transient and permanent faults, triggering a dynamic re-organization of the architecture to mitigate faults' effects and, if possible, to recover. A draft overview of the issues to be tackled is the following:
1. definition of an appropriate fualt model with respect to the execution model and architecture;
2. selection of existing hw/sw techniques and/or definition of new ones, exploiting the redundancy of the available resources;
3. definition of sw/hw reconfiguration strategies to enable correct computation and acceptable performance even in presence of permanent faults;
4. validation/evaluation of the achieved dependability level.

A design methodology for reliable systems on Multi-FPGA platforms
The aim of the PhD work is to build an embedded system implemented on a Multi-FPGA platform, with a tunable level of reliability against both transient and permanent faults for space applications. In particular, the research aims at designing a system where tasks are initially distributed over the network of FPGAs and, during the life of the system tasks can be "migrated" within the system to cope with the occurrence of faults. Based on the functionality being performed, a selective hardening of the system may be pursued, whenever faults may eventually cause an acceptable degradation of the manipulated data. We would like to exploit the dynamic reconfiguration properties of the platform to mitigate the effects of soft errors and permanent failure for allowing the overall system to continue working, eventually undergoing a graceful degradation. The open issues within this scenario are numerous; the PhD thesis would focus on:
1. Define strategies and mechanisms to allow the single tasks/subsystems to detect the occurrence of their own or others’ faults, and to trigger a reconfiguration of the overall system to cope with the critical situation.
2. Identify and define strategies for a selective, tunable hardening of the tasks, in relation with the functionality being performed, exploring different trade-offs between the level of accuracy of the computation being performed and the reliability overheads.
3. Define an overall system, exposing an adaptive reliability property, based on the two identified elements.
Fault injection strategies can be adopted to validate the proposed solutions.

Diagnostic testing strategies for complex digital devices
The aim of the PhD work is to identify, design and develop suitable strategies to enable diagnosis of complex devices, either off-line, during maintenance or on-line, when it is necessary to detect and diagnose the occurrence of faults to mitigate them as soon as possible (as for instance in critical systems). The PhD work will investigate approaches to define appropriate test sets, specifically suited for diagnosis rather than testing. In this perspective, the research will investigate:
1. the effectiveness, completeness of the usually adopted tests w.r.t. diagnosis;
2. definition of a strategy to create test sets optimized to diagnose the component affected by the fault;
3. improve the system design to improve fault diagnosiability.

Context-aware data design, integration and management

Context-aware data querying and integragration from heterogeneous data sources

More and more often interesting data is stored in several disparate sources, in different formats, eventually providing also overlapped contents. In order facilitate the user in accessing this data, it is necessary to provide a unified, integrated view of the overall data, and an infrastructure supporting queryy formulation and answering, and data fusion of the results. Furthermore, based on the actual context, the user's interests may change and differently affect both his/her queries and data deemed as important. The PhD research aims at defining a context-aware methodology and framework for integrating, query formulating and answering over heterogeneous data sources, by means of a unified representation.
Reference: prof. Elisa Quintarelli