Flexible Computing Systems For AI Acceleration At The Extreme Edge Of The IoT

Embedding intelligence in extreme edge devices allows distilling raw data acquired from sensors into actionable information, directly on IoT end-nodes. This computing paradigm, in which end-nodes no longer depend entirely on the Cloud, offers undeniable benefits, driving a large research area (TinyML) to deploy leading Machine Learning (ML) algorithms on micro-controller class of devices. To fit the limited memory storage capability of these tiny platforms, full-precision Deep Neural Networks (DNNs) are compressed by representing their data down to byte and sub-byte formats, in the integer domain. However, the current generation of micro-controller systems can barely cope with the computing requirements of QNNs. This thesis tackles the challenge from many perspectives, presenting solutions both at software and hardware levels, exploiting parallelism, heterogeneity and software programmability to guarantee high flexibility and high energy-performance proportionality. The first contribution, PULP-NN, is an optimized software computing library for QNN inference on parallel ultra-low-power (PULP) clusters of RISC-V processors, showing one order of magnitude improvements in performance and energy efficiency, compared to current State-of-the-Art (SoA) STM32 micro-controller systems (MCUs) based on ARM Cortex-M cores. The second contribution is XpulpNN, a set of RISC-V domain specific instruction set architecture (ISA) extensions to deal with sub-byte integer arithmetic computation. The solution, including the ISA extensions and the micro-architecture to support them, achieves energy efficiency comparable with dedicated DNN accelerators and surpasses the efficiency of SoA ARM Cortex-M based MCUs, such as the low-end STM32M4 and the high-end STM32H7 devices, by up to three orders of magnitude. To overcome the Von Neumann bottleneck while guaranteeing the highest flexibility, the final contribution integrates an Analog In-Memory Computing accelerator into the PULP cluster, creating a fully programmable heterogeneous fabric that demonstrates end-to-end inference capabilities of SoA MobileNetV2 models, showing two orders of magnitude performance improvements over current SoA analog/digital solutions.

Surveillance risks in IoT applied to smart cities

Nowadays, cities deal with unprecedented pollution and overpopulation problems, and Internet of Things (IoT) technologies are supporting them in facing these issues and becoming increasingly smart. IoT sensors embedded in public infrastructure can provide granular data on the urban environment, and help public authorities to make their cities more sustainable and efficient. Nonetheless, this pervasive data collection also raises high surveillance risks, jeopardizing privacy and data protection rights. Against this backdrop, this thesis addresses how IoT surveillance technologies can be implemented in a legally compliant and ethically acceptable fashion in smart cities. An interdisciplinary approach is embraced to investigate this question, combining doctrinal legal research (on privacy, data protection, criminal procedure) with insights from philosophy, governance, and urban studies. The fundamental normative argument of this work is that surveillance constitutes a necessary feature of modern information societies. Nonetheless, as the complexity of surveillance phenomena increases, there emerges a need to develop more fine-attuned proportionality assessments to ensure a legitimate implementation of monitoring technologies. This research tackles this gap from different perspectives, analyzing the EU data protection legislation and the United States and European case law on privacy expectations and surveillance. Specifically, a coherent multi-factor test assessing privacy expectations in public IoT environments and a surveillance taxonomy are proposed to inform proportionality assessments of surveillance initiatives in smart cities. These insights are also applied to four use cases: facial recognition technologies, drones, environmental policing, and smart nudging. Lastly, the investigation examines competing data governance models in the digital domain and the smart city, reviewing the EU upcoming data governance framework. It is argued that, despite the stated policy goals, the balance of interests may often favor corporate strategies in data sharing, to the detriment of common good uses of data in the urban context.

Fondamento e limiti costituzionali del potere punitivo nel sistema di legalità integrata

La ricerca affronta la questione della punizione nella prospettiva del diritto costituzionale nazionale integrata con quella del diritto europeo dei diritti dell’uomo. Nella Parte I è sostenuta la tesi secondo cui la trasformazione della Costituzione penale avviata sotto l’influsso della giurisprudenza CEDU rappresenta complessivamente un avanzamento nel processo di costituzionalizzazione del potere punitivo. Questa conclusione è supportata attraverso un confronto della filosofia costituzionale classica sulla punizione con i diversi approcci interpretativi alla Costituzione penale sviluppati durante il XX secolo (approcci tradizionale, costituzionalistico ed EDU). Nella Parte II è invece sostenuta la tesi secondo cui, nonostante gli effetti positivi dell’armonizzazione sovranazionale, lo statuto costituzionale della punizione dovrebbe comunque rimanere formalmente autonomo dal diritto EDU. Non solo, infatti, nessun paradigma dei rapporti interordinamentali finora sviluppato può giustificarne un’integrazione totale, ma essa rischierebbe anche di diminuire la normatività dell’aspetto sociale della Costituzione penale, già ipocostituzionalizzato rispetto a quello liberale. Nella Conclusione sono quindi sviluppati gli elementi fondamentali di un approccio interpretativo alternativo alla Costituzione penale che risponda meglio di quelli esistenti alle esigenze sia di garantire la massima costituzionalizzazione della punizione sia di facilitare l’integrazione sovranazionale. In base a un simile approccio costituzionalmente fondato, sostanzialista, rights-based e inclusivo di tutte le ideologie costituenti, la Costituzione potrebbe essere letta nel senso di prevedere un modello di disciplina unitario per tutte le forme di esercizio del potere punitivo (salvo quello disciplinare, distinguibile sotto l’aspetto istituzionale) caratterizzato da: una riserva di legge a intensità variabile; uno scrutinio stretto della Corte sulla giustificabilità costituzionale della pena; l’estensione dell’ambito di applicazione dei principi di colpevolezza e rieducazione; un pieno sviluppo degli aspetti di garanzia collettiva dei classici principi costituzionalpenalistici (obblighi di tutela penale e garanzia dell’effettiva collocazione della pena in capo al soggetto colpevole), nonché derivabili dall’art. 3 Cost. (proporzionalità della pena alle condizioni materiali del soggetto punito).

Interfaces between national and EU law: time limits in cross-border civil proceedings and their impact on the free circulation of judgments

The thesis aims at exploring possible legal solutions to remove the obstacles to the free circulation of judgments in the civil justice area that arise from the remarkably diverging national rules on procedural time limits. As shown by the case-law of the CJEU, time limits have recently come under closer scrutiny. The interplay between national and EU law illustrates that time limits raise significant deficiencies connected with the right to a fair trial under Art. 6 ECHR and Art. 47 CFR – e.g. the effective recovery of claims, effective judicial protection, effective cross-border enforcement of judgments – which negatively impact EU cross-border civil litigation. In order to overcome some of the weaknesses of the current legal framework governing the cross-border enforcement of judgments and strengthen the parties’ fundamental procedural rights the PhD thesis intends to determine whether and, to what extent time limits can be harmonised at EU level. EU action on time limits would indeed favour the speed, efficiency and proportionality of cross-border proceedings without sacrificing the fairness of the judicial process and the equality of the parties