This article investigates the inuence of digital technologies on youth participation in urban governance within the Global South, with a specic focus on Saint-Louis, Senegal. Rapid urbanization in developing countries has intensied the need for inclusive governance, especially as youth constitute a signicant portion of the urban population. This study examines how access to digital tools such as smartphones, internet connectivity, and social media shapes youth engagement in local decision-making processes. A mixed-methods approach was employed, involving a sample of 549 youth, of whom 55% were under the age of 20. Quantitative surveys with youth were complemented by qualitative interviews with key local governance actors and spatial mapping of digital infrastructure. The ndings reveal substantial disparities in digital access and planning across neighborhoods. While 90% of the Northern districts and Guet Ndar exhibit high levels of digital infrastructure and urban planning, areas like Bango (65%) and Pikine (50%) remain predominantly informal and under-resourced. Notably, digital connectivity in Guet Ndar is overwhelmingly mobile-based (83%), highlighting the centrality of mobile data as the primary access point. However, local government initiatives to enhance digital inclusion remain limited, with only 4% of respondents identifying concrete efforts by municipal authorities. The study underscores the urgent need to address infrastructural gaps and to foster participatory governance frameworks that enable meaningful youth inclusion. It advocates for targeted policy interventions to create a more equitable digital environment, enhancing civic engagement and governance outcomes in rapidly urbanizing contexts. By situating the case of Saint-Louis within broader Sub-Saharan dynamics, the article offers insights into how digital inclusion can act as a lever for sustainable urban governance in the Global South.

We estimate the magnetic field in the jets of the recently discovered 7 Mpc long Porphyrion system. We used nondetection of the system in gamma-rays to derive a lower bound on the co-moving magnetic field strength at the level of 10 nG (comoving). This value is consistent with recent estimates of magnetic fields in the filaments of the large-scale structure.We discuss the possibility that instead of being the extreme case of a radio jet formation scenario, Porphyrion actually traces a very high-energy -ray beam emitted by an active galactic nucleus. In such a model, jets do not need to spread into the voids of the large-scale structure to appear straight on a very large distance range, and several anomalies of the standard radio jet scenarios can be solved at once.

Biogas-fed solid oxide fuel cell (SOFC) technology based on Ni-GDC electrolyte-supported cell (ESC) offers a sustainable solution for European farms’ heat and power demands with superior efficiency. However, the understanding of the degradation and recovery behavior of the Ni-GDC based ESC under processed biogas environment is limited. This study investigates the responses of various electrochemical processes to the degradation under two cases: (1) cold recirculation before reformer (CB) and hot recirculation before reformer (HB) with no risk of carbon deposition. Chronopotentiometry and current–voltage (IV) results indicated that CB case had the highest degradation rate of -11.6 mV/kh, followed by the HB case with -6.4 mV/kh and the dry H2 case with -0.42 mV/kh. With the electrochemical impedance spectroscopy (EIS) measurements and distribution of relaxation time (DRT) method, it was found that CB and HB case demonstrated much lower gas conversion resistance than the dry H2 case, while two peaks representing the electrode gas diffusion process integrated with O2− surface exchange merged into one peak in HB and dry H2 case but split in CB case. Complex nonlinear least squares (CNLS) fit was used to quantify the resistance degradation under CB and HB environments, which depended mainly on the ohmic resistance and two electrode charge transfer resistances. The recovery process under dry H2 environment showed that the total resistance increased after 134 h, validating the degradation inertia under biogas environment, while the time constants of RQ elements and the peak shifting indicated that the electrochemical performance was partially recovering to the initial state. This study provides insights into the long-term performance degradation of Ni-GDC based ESC under processed biogas environment and guidelines to the operating conditions with a lower degradation rate.

In spatial timber assemblies, a multitude of factors can lead to errors that hinder the implementation of complete automation including but not limited to inaccuracies of the robotic setup, deformations of the structures during assembly, or the natural dimensional variability of wood. Any unplanned event, construction detail, or material variability that was not embedded ahead of time in the CAD environment can cause failure. To mitigate these challenges, this paper expands this conventional one-directional design-tofabrication pipeline and proposes interactive digital fabrication workflows where humans and robots work synergistically, blending the adaptability of human craft with the precision of robotic technology. The method is validated with two prototypes comprising linear timber elements and 3d printed connections that showcase adaptability in the fabrication setup and allow for design changes to happen concurrently with fabrication. In this paradigm, human operators are not mere extensions of the robotic system but rather central to dynamic problemsolving and instrumental in making immediate adjustments.

Artificial metalloenzymes (ArMs) combine the broad reactivity of transition metals with the exquisite selectivity of enzymes, expanding the scope of biocatalysis beyond natural enzymatic functions. The biotin-streptavidin (Sav) system offers a versatile scaffold for integrating synthetic metal cofactors into protein frameworks with high precision. This thesis explores the design of ArMs with multiple catalytic functions, advancing the development of multifunctional biocatalysts.

Chapter 1 introduces ArMs and highlights the Sav technology as a robust platform for embedding abiotic catalytic centers into proteins. ArMs benefit from the precision of protein engineering and the catalytic versatility of transition metals, and have found broad applications in asymmetric transformations. This sets the foundation for developing Sav-based ArMs for enantioselective and multi-step catalysis.

Chapter 2 describes the creation of the first base-metal artificial transfer hydrogenase (ATHase) using a biotinylated Mn(I) complex. While previous ATHases relied on precious metals, this work shows that Mn(I) can also function effectively in ArMs. Through chemo-genetic optimization, we engineered Mn-based ATHases that demonstrate high activity, broad substrate scope, and good functional group tolerance, underlining their potential for sustainable and cost-effective catalysis.

Chapter 3 tackles the challenge of creating ArMs with dual catalytic functions. We developed a system that incorporates two distinct biotinylated cofactors into Sav: one photoactive and one metal-based. This design enables tandem enantioselective transformations, including a photoinduced Câ H activation followed by asymmetric catalysis. The work highlights how ArMs can facilitate multi-step, integrated reaction cascades.

Chapter 4 advances dual-cofactor ArMs by focusing on their spatial arrangement within adjacent binding sites of Sav. By combining a biotinylated nickel complex with a peptide-derived amine catalyst, we achieved synergistic catalysis in an enantioselective Michael addition. This demonstrates how precise cofactor positioning can improve both activity and selectivity, offering new strategies for the design of cooperative catalytic systems.

In summary, this thesis presents significant progress in ArM design via the biotin-streptavidin platform. By integrating diverse and complementary cofactors, we created artificial enzymes with enhanced reactivity, selectivity, and substrate scope. The introduction of dual cofactor systems enables tandem and synergistic catalysis, showcasing new directions for engineering multifunctional ArMs with promising applications in green chemistry, synthetic biology, and sustainable synthesis.

Using locally available, unprocessed stones as construction material offers a promising solution to meet the modern construction industry's sustainability and low-carbon footprint requirements. Stone masonry construction typically depends on the expertise of skilled craftsmen, who arrange stones on-site based on their experience and intuition. This reliance on expert knowledge represents a significant barrier to the broader adoption of stone as a construction material. Also, the resulting structures often exhibit unique and irregular textures, introducing high levels of uncertainty in structural analysis and complicating their design, verification, control, and management.

To increase construction efficiency and modeling accuracy using digital technologies, this thesis developed computational methods for planning microstructure and performing microscale structural analysis. These methods have been applied in an experiment to demonstrate a digital construction pipeline for building mortar-joint stone masonry walls.

The microstructure planning algorithms were developed based on traditional masonry practices, arranging stones for 2D dry-joint masonry walls and for 3D mortar-joint masonry walls. For dry-joint masonry walls, the algorithms also consider placement stability and the structural performance of the wall. As validation, a multi-leaf wall with a stone layout generated by the algorithm was physically constructed using a robotic arm and tested under simple compression loading. The results demonstrate that the algorithms can efficiently plan stone arrangements, producing virtual and physical walls that are comparable to those built by skilled masons in terms of filling, horizontality, interlocking, and resistance.

To predict and analyze the performance of stone masonry walls while accounting for the influence of microstructure, I developed a rigid block modeling approach that explicitly represents both stones and mortar. Limit analysis and nonlinear static analysis were formulated as optimization problems, incorporating inequality constraints to model cracking, crushing, and shear failure. The numerical simulations were validated against data from various experiments in the literature and the test on the robotically constructed wall, predicting failure mechanisms and force displacement curves that closely aligned with the experimental results.

The research presented in this thesis forms key components of the digital construction pipeline using irregular stones. It includes solutions ranging from geometric digital twinning to structural analysis and from virtual design to physical construction, contributing to sustainable and reliable building practices with unprocessed materials.

Recent years have witnessed a growing integration of automatic control into a wide range of cyber-physical systems, including electrical grids and autonomous transportation networks. These systems operate in uncertain and dynamic environments, exhibit complex behaviors arising from the intricate interactions between multiple decision-makers, and are subject to both safety and real-time computational constraints. Their automation must, therefore, address the challenges related to these three characteristics. In particular, this involves developing novel control and learning solutions that ensure reliable operation despite unmodeled uncertainty, leverage the availability of large volumes of data to improve closed-loop performance, and maintain computational efficiency.

This thesis contributes to addressing the aforementioned challenges in modern control systems. Departing from classical probabilistic or adversarial uncertainty models and moving towards designing adaptive controllers that adjust to the true disturbance realizations, in the first part of this thesis we study optimal control from the perspective of regret minimization. Focusing on finite-horizon control, we first present a convex optimization approach for synthesizing both the noncausal optimal policy in hindsight and a causal policy that tracks it as closely as possible. We then consider a model predictive control scheme based on the repeated computation of regret-optimal controllers, and establish performance, safety, and stability guarantees for the resulting closed-loop system.

To address the lack of accurate mathematical models describing the dynamics of modern control systems and the uncertainty they are affected by, in the second part of this thesis we focus on data-driven robust control methods. For the case where the dynamics are unknown, we derive suboptimality bounds for safely learning constrained linear quadratic Gaussian regulators from noisy data. Our analysis shows that the suboptimality of the proposed method converges to zero approximately as a linear function of the mismatch between the nominal and the true dynamics, provided that these are sufficiently close. For the case where the probability distribution of the uncertainty is unknown, we present new duality results to reformulate Wasserstein distributionally robust optimal control problems with empirical center distributions and possibly bounded uncertainty supports as semidefinite programs.

Shifting our attention to numerical methods for solving optimization problems, in the third part of this thesis we study iterative optimization algorithms from a dynamical system perspective. Our key contribution is a complete and unconstrained parametrization of algorithms that are convergent for smooth, possibly non-convex, objective functions. This is achieved by viewing convergent algorithms as the superposition of a gradient-descent step and a stable term that can be chosen freely. The resulting framework is directly compatible with automatic differentiation tools, enabling the use of machine learning techniques to design algorithms with guaranteed convergence and optimized performance.

Hydraulic fracturing is a key technique used to enhance subsurface reservoir permeability, with applications in oil and gas extraction, geothermal energy, and carbon sequestration. The closure phase of hydraulic fractures plays a critical role in determining reservoir properties such as the leak-off coefficient and in-situ stresses, which are essential for predicting fracture behavior and optimizing engineering design. Despite the widespread use of pressure-based methods to estimate fracture closure time and pressure, conventional techniques are limited by empirical assumptions and fail to account for critical factors such as poroelastic effects. Recently, analytical models such as the sunset solution have been proposed to provide a physics-based framework for characterizing closure, yet their experimental validation remains unexplored. This thesis presents a combined experimental and numerical investigation of fracture closure behavior. Laboratory-scale hydraulic fracturing experiments were conducted on cubic Molasse sandstone samples using a true triaxial apparatus. A key advancement was the integration of an eddy current (EC) probe to directly measure fracture opening at the injection point. This method was calibrated using high-resolution CT scans and thin-section analysis. Four hydraulic fracturing experiments were performed using different fluids with different viscosities to capture closure behavior across multiple injection and shut-in cycles. Two experiments were conducted in the viscosity-storage-dominated regime and two in the viscosity-leak-off-dominated regime. The results showed that fractures never fully close, with a residual opening persisting even after unloading, challenging the assumption of full closure in traditional models. The fracture stiffness, derived from simultaneous pressure and opening measurements, provided a more precise closure indicator than conventional pressure-based methods. In low-viscosity tests, a pronounced poroelastic back-stress effect was observed, increasing closure pressure in successive cycles. Moreover, the time evolution of fracture opening and pressure near closure exhibited asymptotic behavior consistent with the sunset solution. Micro-scale analysis of the core samples was also conducted to investigate fracture morphology and contact mechanics. High-resolution CT scans were used to reconstruct the upper and lower fracture surfaces, preserving their integrity without physical separation. The analysis revealed an increasing concentration of contact and bridge points near the fracture tip, indicative of a fracture process zone. Surface roughness characterization confirmed intergranular propagation, with a low Hurst exponent (~0.4). Residual opening profiles were anisotropic, showing a long-range gradient along the propagation direction. To complement the experimental results, numerical simulations were performed to assess two modeling approaches for fracture closure. A fixed-grid model, incorporating a minimum residual opening constraint, was compared with a moving-mesh algorithm that accounts for near-tip asymptotics. Both models captured fracture propagation and arrest effectively. While the moving-mesh model better resolved the transition to closure, the fixed-grid approach successfully reproduced the sunset solution with sufficient spatial refinement, even under asymmetric conditions induced by material heterogeneities.

Developing agents that can reliably act on our behalf is central to artificial intelligence (AI). These agents must seamlessly interact with tools, like search engines and databases, and collaborate. In this thesis, we study the abstractions, methods, and infrastructure needed to enable and support the development of AI agents in the era of large language models (LLMs). The contributions of the thesis are divided into four parts.

Part 1 examines goal-oriented collaboration between two components, at least one of which is LLM-based. For an LLM-based component to interact successfully with others, it must adhere to specified interfaces, especially when interacting with traditional software-based components exposed through an API, and steer the collaboration toward high-utility outcomes. We show that LLM decoding algorithms serve as an efficient strategy to accomplish both objectives without modifying the underlying model.

Part 2 focuses on scenarios where the underlying model's capabilities are insufficient for effective collaboration, and the training signal necessary for improving the model is not readily available. To address this challenge, we introduce the principle of exploiting asymmetry for synthetic data generation and demonstrate how it can be applied to generate useful data even for tasks that LLMs cannot solve directly. We highlight the generality of this approach by drawing connections to seminal work on self-improvement for LLMs.

Part 3 addresses the collaboration among multiple AI systems, tools, and humans. We propose an abstraction that, in concert with the accompanying library, provides a theoretical and practical infrastructure with a modular and concurrency-friendly design, which enables the modeling, implementation, and systematic study of arbitrarily complex structured interactions. To demonstrate the potential of the framework and the accompanying library, we use them to systematically investigate the benefits of complex interactions for solving competitive coding problems.

Part 4 proposes a novel perspective called semantic decoding that allows us to systematically study the design space of structured interactions. We conclude this part by discussing the research opportunities and questions emerging from the semantic decoding perspective, enabled by the foundation laid in Parts 1, 2, and 3.