Open Master's Thesis Positions

This project is set to limited visibility by its publisher. To see the project description you need to log in at SiROP. Please follow these instructions: - Click link "Open this project..." below. - Log in to SiROP using your university login or create an account to see the details. If your affiliation is not created automatically, please follow these instructions: http://bit.ly/sirop-affiliate

This project is set to limited visibility by its publisher. To see the project description you need to log in at SiROP. Please follow these instructions: - Click link "Open this project..." below. - Log in to SiROP using your university login or create an account to see the details. If your affiliation is not created automatically, please follow these instructions: http://bit.ly/sirop-affiliate

In recent years, mobility has become a cen- tral issue in many countries and some alarming statistics show a growing need for change. For example, some studies assess that the cost of congestion for the EU is no less than 267 bil- lion euros per year. This affects not only the citizens’ well-being but also the environment, since traffic jams heavily increase the emission of CO2 [1]. Rather than increasing road ca- pacity [2], policymakers and researchers are increasingly looking at control strategies that make better use of the infrastructure already in place. The literature on traffic control is vast. Classical approaches include ramp metering, e.g., ALINEA [3], variable speed limits [4], traf- fic signal control, e.g., fixed-time plans [5] and max-pressure [6], and perimeter (gating) con- trol based on the macroscopic fundamental di- agram [7, 8]; see also [9]. More recently, rein- forcement learning and other machine learning methods have shown promising results, espe- cially for signal control [10]. Despite this abun- dance, there is no common ground to compare them. Each work relies on its own networks, de- mand, simulator settings, and metrics, so that results reported in different papers are rarely com- parable. Existing benchmarks [10, 11] mostly focus on a single problem, typically learning-based signal control, and do not allow one to compare different families of controllers on the same net- works. A key reason behind this gap is that setting up realistic closed-loop simulations is still difficult and time consuming. Microscopic simulators such as SUMO [12] are powerful, but controlling them through their programming interface (TraCI) requires specialized expertise, and running and com- paring large batches of experiments is mostly done by hand. To overcome this issue, our group de- veloped Unjam, a “traffic gym” that plays for traffic control a role similar to the one Gymnasium [13] plays for reinforcement learning. Unjam provides a unified interface on top of SUMO in which (i) a controller is defined as a simple Python function, (ii) arbitrary sensors and actuators of a network are connected through a modular pipeline (sensors, pre-processing, controller, actuator mapping, actuators), also via a no-code visual interface, (iii) multiple experiments with different seeds and sce- narios run in parallel, and (iv) experiments are fully reproducible and evaluated with built-in metrics. Unjam is not yet publicly available and will be provided to the student. In this project, we want to exploit Unjam to build a comprehensive benchmark for traffic control. First, the student will identify the most relevant control problems and algorithms, and a set of rep- resentative networks. These include classical networks from the literature, e.g., the ring road [14] and synthetic grids, real urban networks, e.g., the city of Zurich (or part of it) or existing calibrated scenarios such as LuST [15], and highway stretches as a special case where classical highway con- trollers can be tested. All networks are assumed to be already calibrated or to come with realistic origin-destination demand. Then, the selected controllers, both classical (still widely used in prac- tice and in the literature) and learning-based, will be implemented in Unjam and tested on all the networks under several scenarios. The final outcome is a comparable scale of performance across controllers, problems, and networks, and a demonstration of how entire families of controllers can be systematically tested.

The goals of the project are as follows: 1. Learn about traffic control and the Unjam framework; 2. Select the relevant control problems (ramp metering, variable speed limits, traffic signal con- trol, perimeter control, and possibly others), the algorithms to compare, and the networks on which to test them; 3. Implement a mixture of classical and learning-based (e.g., reinforcement learning) controllers in Unjam; 4. Define a set of scenarios and performance metrics (e.g., total time spent, delay, throughput, queue lengths, emissions) and a common evaluation protocol; 5. Run an extensive comparison of the controllers across all networks and scenarios, and char- acterize their performance on a common scale.

ccenedese@ethz.ch, arimoldi@ethz.ch

This project is set to limited visibility by its publisher. To see the project description you need to log in at SiROP. Please follow these instructions: - Click link "Open this project..." below. - Log in to SiROP using your university login or create an account to see the details. If your affiliation is not created automatically, please follow these instructions: http://bit.ly/sirop-affiliate

A power system can be stable while exhibiting weakly damped oscillations. As converter controls interact with dynamic network lines, ensuring that oscillations decay sufficiently quickly becomes an important design objective. Assessing damping through eigenvalue analysis requires an assembled system model and generally needs to be repeated when the network configuration or operating conditions change. A common dissipativity multiplier transforms local device and line admittances into matrix inequalities that can be optimized by semidefinite programming (SDP). This project will investigate extending the underlying no-crossing argument to boundaries of prescribed pole regions. A sector around the negative real axis specifies a minimum damping ratio, whereas a shifted vertical boundary specifies a minimum exponential decay rate. These requirements capture different aspects of oscillation performance. The central question is which local conditions and interconnection assumptions make these damping guarantees compositional. For an admitted component class, we would synthesize a shared multiplier; each converter or line could then check compliance independently using its own terminal model. The study will examine whether useful bounds can be independent of network size and topology, and identify restrictions where needed. Initial development will use analytical or state-space models, since finite imaginary-axis impedance scans alone cannot rigorously determine responses on the required complex-plane contours.

* Define damping specifications, admissible component models, and operating conditions, using a minimum decay-rate requirement. * Derive sufficient local conditions that exclude pole crossings satisfying the target specification. Address component poles, hidden modes, the reference-angle zero mode, and contour behaviour near the origin and infinity. * Adapt the Chebyshev multiplier parameterization and adaptive interval splitting to the selected contours. Implement the search in MATLAB/Python and develop checks between sampling points with justified error bounds, distinguishing a rigorous certificate from numerical evidence. * Evaluate the certificates on a small converter network and the modified IEEE benchmark. Compare guaranteed and actual modal damping across droop settings, converter placements, and selected operating points, using eigenvalue analysis and small-disturbance simulations in MATLAB/Simulink. Report conservatism, computation time, and certified parameter regions.

Dr Andrey Gorbunov, agorbunov@ethz.ch

This research will be carried out in parallel with a researcher at the Robotic Fabrication Lab (RFL) at the D-ARCH department, where the 3D printing of prototypes can be conducted. The student will build on an existing parametric workflow (Rhino + Grasshopper) to enable the integration of thermal and structural simulations with ANSYS software. Selected iterations may be 3D-printed at 1:1 scale, with performance validated through physical experiments including compression tests (structural) and a hotbox setup (thermal).

The main aim of this project is to develop a workflow for simulating the thermal and structural properties of 3D-printed ceramic components. Results of a parametric workflow should feed into the simulation software to establish relationships between set parameters and the performance of the components. Expected outcomes include: • An established workflow for enabling parametric design and simulation • Series of design iterations with thermal and structural simulation results • Possible experimental results with physical prototypes Successful results will be published as part of a scientific journal.

Please send your CV to lim@arch.ethz.ch. This project can be undertaken as an open project, a semester project or a master's thesis, and the scope will be adjusted accordingly. Feel free to send an email if you have any questions. More details will be given about the project during the initial discussions.

The secure operation of the Swiss transmission grid requires the voltage at every node to stay within tight limits. Since voltages are closely linked to reactive power, Swissgrid relies on the cooperation of connected power plants and DSOs, which support the voltage by adjusting the reactive power they exchange with the grid. In Swissgrid's voltage support program, Swissgrid computes voltage references for all nodes from an optimization (OPFV) and settles the reactive energy of each participating unit: reactive power that moves the voltage towards its reference is remunerated, whereas reactive power that pushes it away is penalized. Within these rules, each participant decides autonomously how much reactive power to provide. Operational data show that participants respond to these incentives in very different ways. Some power plants closely track the references, whereas some DSOs co-optimize several of their connection points and shift reactive power between them to increase their remuneration or to avoid penalties. Such strategic behavior determines which voltages are actually obtained for a given set of references, but it is not correctly represented when the references are computed. Moreover, the reactive power of one participant changes the voltages, and hence the payments and decisions, of its neighbors. Voltage support is therefore a game among the participants, coupled through the physics of the grid. Previous projects at the IfA have formalized the program as a Stackelberg game and used inverse optimization to reverse-engineer the objectives and constraints of individual participants from Swissgrid's data. What is still missing is a model that brings the participants together with the transmission grid and predicts their collective response to a given set of voltage references, i.e., a digital twin of the voltage support program. This master thesis aims to build such a digital twin. For given voltage references, the twin shall reproduce the reactive power decisions of the participants and the resulting change of the grid voltages, taking into account both the strategic behavior of the participants and their coupling through the transmission grid. The OPFV itself is not part of the twin, and voltage references are an input. The thesis builds upon a previous master thesis, extends its inverse optimization technique to all participants, and is carried out in close collaboration with Swissgrid, which provides operational data of all participating units at a 15-minute resolution as well as a model of the Swiss transmission grid.

The thesis addresses two research questions. The first, Who plays, and how?, asks you to identify the strategic players from data and to model how each player decides on its reactive power, for single units as well as for company portfolios in which a company co-optimizes several of its connection points. The second, Does the game reproduce the grid?, asks you to couple the player models through the transmission grid and to assess whether the outcome of the resulting game agrees with the measured reactive power and voltages, including their response to changes of the voltage references. The expected results are: - a data-calibrated model of the players, covering single units and company portfolios, with a small number of behavioral types; - a digital twin, implemented in Python, that maps a given set of voltage references to the reactive power of all participants and to the grid voltages; - a validation on historical Swissgrid data that shows where the twin is accurate and where it is not, including a comparison with simple benchmarks such as the participant behavior currently assumed in the OPFV; - insights into the game itself, for example whether its outcome is unique. In the longer term, such a twin will allow Swissgrid to check voltage references before they are dispatched, to test improved versions of the OPFV, and to evaluate new incentive schemes, which are core goals of the MAESTRO project.

Please email your resume/CV (including relevant projects) and transcript of records in PDF format to Zhisen Jiang (zhijiang@ethz.ch), Jonas Matt (jmatt@ethz.ch), Saverio Bolognani (bsaverio@ethz.ch)

This project is set to limited visibility by its publisher. To see the project description you need to log in at SiROP. Please follow these instructions: - Click link "Open this project..." below. - Log in to SiROP using your university login or create an account to see the details. If your affiliation is not created automatically, please follow these instructions: http://bit.ly/sirop-affiliate

**INTRODUCTION** Thanks to advances in urban climatology and biometeorological techniques, some detailed microclimatic analyses and outdoor comfort assessments have been conducted in the last decade. Therefore, qualitative methods (e.g. interviews) that provide an explicit combination of thermal and spatial information have been developed to link outdoor environmental and spatial information about people’s perceptions. Based on these findings, new insights can be gained for the design of comfortable outdoor urban spaces. Nevertheless, frequently design/renovation plans of urban areas like university campuses lack the consideration of key factors such as outdoor microclimate, people, and urban features, which determine the vibrancy of the place, sustainability, outdoor comfort level as well as the local socio-economy (Figure 1) **METHODOLOGY AND TOOLS** The methodology proposed for this thesis is based on supervisors’ previous research and teaching experience (Figure 2). Hence, once identified the critical spots of the campus via interviews, surveys, simulations and measurements, the design decisions to improve the outdoor comfort in different domains have three levels of impact: element, urban, and human. This holistic methodology based on scientific methods allows to the designer justify the retrofit decisions. The methodology is based on mixed-method approach (combination of qualitative and quantitative research methods): on-site interviews, online surveys, simulations (e.g. LadyBug tools, ClimateStudio, Rhinoceros), and on-site measurements of the environmental factors. A proper processing of all the resulted data will allow a clear identification not only of the critical spots of the campus but also campus users’ opinion.

When designing outdoor spaces, the selection of suitable urban furniture, materials of opaque objects/architectural elements in relation with the urban area typology and outdoor environment factors is key to ensure the vibrancy of the place, sustainability and outdoor comfort. Hence, the main aim of the research project is to improve outdoor comfort of the university Campus ETH Hönggerberg. Specifically, the objectives of this thesis are as follows: • To map critical spots of the Campus in terms of both, campus users’ opinion and different outdoor environmental factors such as solar radiation, air temperature, relative humidity, wind speed/direction, illuminance levels, surfaces temperature, soundscapes, air quality, and noise levels. • To propose urban retrofit solutions to improve human comfort while doing specific outdoor activities in the identified critical spots of the campus.

Please contact us if you are interested or/and have questions about the project: Dr. Abel Sepúlveda Luque (sepulvedaluque@arch.ethz.ch Dr. Lars O Grobe (grobe@arch.ethz.ch)

**INTRODUCTION:** The majority of the 441.5 million individuals worldwide who are visually impaired have low vision, with 23.8 million people suffering from low vision in Europe alone. Low vision is a long-term visual impairment that cannot be fully corrected by glasses or contact lens. Individuals with low vision have reduced visual acuity (sharpness of vision) or field of view compared to normal vision but not to the extent of complete blindness. This condition poses challenges in everyday functioning and navigating architectural spaces, particularly those with steps, corners, or level differences, leading to safety concerns. Current architectural and lighting practices for designing visually accessible spaces rely on total illumination on surfaces, overlooking the complex interaction between light and surface material properties, as well as the contrast sensitivity and reduced acuity experienced by individuals with low vision. With the growing interest in architectural and lighting design practices to create accessible and inclusive spaces, a suite of tools was developed under the project DeVAS (Designing Visually Accessible Spaces) by, leveraging the backward ray tracing algorithm Radiance. **METHODOLOGY AND TOOLS:** The DeVAS suite of tools are standalone programs that are nor friendly for architects and designers to use. Moreover, they are not integrated into the widely used design workflows facilitated by the Rhino 3D software [5] along with its visual scripting plugin Grasshopper [5]. Hence, members of the research group, Architecture and Intelligent Living (AIL), at Karlsruhe Institute of Technology (KIT) developed a Grasshopper plug-in for Rhino 8 called DeVASforRhino. This tool enables the calculation of a Hazard Visibility Score based on the percentage of visible edges (HVS metric) within a region of interest (ROI) from a Radiance-rendered image (Figure 2).

When designing indoor spaces, the selection of suitable fenestration systems, materials of opaque objects/architectural elements is key to ensure the visibility of the indoor scene. Hence, the main aim of the research project is to study the effect of indoor materials and artificial lighting systems on HVS considering different visual impairment severity levels. The case study is a hospital building located in Zürich. That is translated in different objectives: • To investigate the impact of indoor materials on HVS for different visual impairment severity levels; • To investigate the impact of artificial lighting systems on HVS for different visual impairment severity levels.

Please contact us if you are interested or/and have questions about the project: Dr. Abel Sepúlveda Luque (sepulvedaluque@arch.ethz.ch) Dr. Lars O. Grobe (grobe@arch.ethz.ch)

This project is set to limited visibility by its publisher. To see the project description you need to log in at SiROP. Please follow these instructions: - Click link "Open this project..." below. - Log in to SiROP using your university login or create an account to see the details. If your affiliation is not created automatically, please follow these instructions: http://bit.ly/sirop-affiliate

This research is a parallel investigation in collaboration with a researcher at the Robotic Fabrication Lab (RFL) at the D-ARCH department, where the paste iterations can be subjected to the large-scale printing and sintering experiments during development. The student will carry out the following tasks: • Characterisation of clay composition • Formulation and iterative testing of candidate organic pastes • Rheological characterisation to assess printability • Extrusion/casting and drying experiments to characterise shrinkage of the developed paste The candidate should meet the following criteria: • Interest in additive manufacturing technologies • Ability to independently plan and carry out hands-on material experiments • Knowledge of material characterisation techniques • Background in soft materials or a related field is an advantage

The main aim of this project is to develop a printable paste with an organic composition that is compatible with clay and reduces the overall shrinkage of the printed component during drying and sintering. The paste should be developed according to the following criteria: • Sufficient viscosity for printability and stable extrusion at 2 - 8 mm • Sufficient stability for layering • Low shrinkage during the drying process • Fully and safely combustible composition

Please send your CV to lim@arch.ethz.ch. This project can be undertaken as an open project, a semester project or a master's thesis, and the scope will be adjusted accordingly. Feel free to send an email if you have any questions. More details will be given about the project during the initial discussions.

In this project, we aim to learn stable control policies by exploiting information that trajectory optimization already provides but usually discards. Under standard assumptions, the optimal cost of the trajectory optimization problem, viewed as a function of the initial state, acts as a Lyapunov function: it decreases along the trajectories of the optimal controller. Moreover, its sensitivity with respect to the initial state can be obtained at almost no extra cost from the Lagrange multipliers at the solution, and it indicates which directions of motion decrease the value. Rather than training the policy to merely reproduce the expert's actions, this approach penalizes violations of the descent condition for the expert's value function (or an approximation thereof), encouraging Lyapunov stability as part of the learning objective. We therefore structure the method in two stages operating at different levels of computational cost: 1. **Offline expert and value-function learning.** We solve trajectory optimization problems offline from many initial states, recording the optimal inputs, the optimal cost, and its derivatives with respect to the initial state. From these samples, a smooth global approximation of the value function is constructed (e.g., by blending local first-order models or by fitting a gradient-supervised neural network), so that the Lyapunov candidate can be evaluated and differentiated at states that were never sampled. 2. **Online learned policy.** A parameterized policy, either a neural network or a differentiable MPC with learnable cost weights, terminal ingredients, and constraint-tightening margins, is trained to run in real time. Its parameters are adjusted through a gradient-based procedure that combines expert imitation with a penalty for violations of the Lyapunov descent condition along the system dynamics. If applicable and theoretically possible, parameters of the learned policy may also be adjusted online. We will evaluate the proposed algorithm in simulation on benchmark nonlinear systems with state and input constraints. We will assess the approach's performance by comparing closed-loop stability, region of attraction, constraint satisfaction, and closed-loop cost against the expert controller and against policies trained by pure imitation, using the same policy class and model.

The central contribution of this project is to turn trajectory optimization data into stability-aware learned controllers: by penalizing violations of the descent condition for the expert's value function or an approximation thereof during training, we aim to obtain policies that retain most of the expert's performance while providing verifiable closed-loop stability, without requiring additional online computations for the descent objective when the policy parameters are fixed. - Implement an offline data-generation pipeline that solves the trajectory optimization problem to record optimal inputs and costs, and extracts the value function derivatives using the optimal Lagrange multipliers. - Develop a smooth approximation of the value function and a training scheme that combines imitation and Lyapunov descent, and apply it to neural-network or differentiable-MPC policies. Depending on the project scope, compare both policy classes. - Validate the learned policies in closed-loop simulation, comparing descent-based training against imitation-only baselines, and study how the accuracy of the value-function approximation affects the stability of the resulting closed loop.

Please email your resume/CV (including lists of relevant publications/projects) and transcript of records in PDF format to johnemus(at)ethz.ch and rzuliani(at)ethz.ch.

Industrial contact-rich manipulation tasks, such as polishing and assembly, are inherently geometric tasks where directional information—such as the insertion axis and local surface normals—is critical. Current policy learning methods [1] leverage vision as the main modality, which requires extensive data to learn this directional information. In contrast, point clouds are better suited for geometric reasoning and can be explicitly augmented with directional information. However, existing point cloud encoders [2] are rarely evaluated on industrial tasks, and they do not incorporate task-relevant directions [3]. This project focuses on developing a point cloud encoder for physical AI applications and combining it with other modalities from TACTO1 (see inline Figure) to achieve superior generalization across novel environments and target objects. Conducted in collaboration with Bota Systems AG, this project offers direct exposure to real-world physical AI applications within the modern robotics industry.

The goal of this project is to develop a point cloud encoder for learning force-aware manipulation policies with superior geometric generalization compared to RGB-only policies. The encoder processes spatial points augmented with pre-computed local and global directions, such as surface normals, principal curvatures, and the nominal insertion axis. The output of this encoder is then fused with latent embeddings from a wrist force/torque sensor and the end-effector pose. By encoding explicit geometric priors rather than relying on raw RGB pixels, the primary objective is to improve cross-environment generalization and policy transfer relative to standard vision-based architectures. The main goals of the project are as follows: - Integrating depth/point cloud streams into the TACTO setup and developing automatic extrinsic calibration routines for wrist and workspace cameras; - Developing a specialized point cloud encoder capable of processing directional vector inputs; - Combining the point cloud encoder’s output with latent force and proprioception embeddings to condition the diffusion policy; - Evaluating the system experimentally and benchmarking its cross-environment and cross- object transfer capabilities against standard vision and baseline point cloud encoders. This work combines elements from robotics, 3D computer vision, and machine learning, with a strong focus on real-world hardware implementation.

Please email your resume/CV (including relevant projects) and transcript of records in PDF format to Cem Bilaloglu (cem.bilaloglu@inspire.ch)

Industrial contact-rich manipulation tasks, such as polishing and insertion, involve complex transitions across multiple states—including free-space motion, impact, sliding, sticking, and separation. Reliably registering and anticipating these events is critical for executing proactive control actions. Different sensor modalities offer complementary advantages for cap- turing these state changes: while wrist force-torque sensors register interactions reactively after contact occurs, visual data and point clouds provide proactive geometric cues. For instance, as a robot approaches a workpiece, spatial data allows the system to anticipate the shift from free space to contact before forces peak. This project focuses on developing a learning-based framework for physical AI that leverages rich multi-modal sensor streams from TACTO1 (see inline Figure) to anticipate contact transitions ahead of time. Conducted in collaboration with Bota Systems AG, this project offers direct exposure to real-world physical AI applications within the modern robotics industry.

The goal of this project is to develop a multimodal transformer that recognizes early precursors to transitions between contact states, such as sticking, sliding, and separation, enabling proactive robotic control. The model will be trained and evaluated on multimodal data from both simulation and the real world. Inputs will include gravity-compensated external wrenches, RGB images, point clouds, and estimated robot motion. Training labels will be generated automatically through physics-informed weak supervision, combining first-principles contact mechanics with the existing force, motion, and interaction observers. These mechanics-based detectors provide causal estimates of the current contact state and therefore serve as physically grounded training labelers. The central research objective is to determine whether a multimodal transformer can exploit the richer multimodal sensory history to identify precursors to contact-state transitions before those transitions can be confirmed by the causal mechanics-based detectors used to generate its training labels. The resulting predictor should generalize across simulated and real-world interaction data and provide sufficiently early and reliable transition estimates to support proactive adaptation of the robot controller. The main goals of the project are as follows: - Collecting multi-modal real-world (TACTO) and simulated data streams (Isaac Lab); - Formulating contact-state labeling via first-principles contact mechanics and force-motion ob- servers; - Training a multimodal transformer to predict upcoming contact transitions; - Evaluating the system experimentally and benchmarking its anticipation latency against pure mechanics-based detectors. This work combines elements from machine learning, robotics, control theory, and contact mechanics with a strong focus on real-world hardware implementation.

Please email your resume/CV (including relevant projects) and transcript of records in PDF format to Cem Bilaloglu (cem.bilaloglu@inspire.ch).

This project is set to limited visibility by its publisher. To see the project description you need to log in at SiROP. Please follow these instructions: - Click link "Open this project..." below. - Log in to SiROP using your university login or create an account to see the details. If your affiliation is not created automatically, please follow these instructions: http://bit.ly/sirop-affiliate

At the core of HVAC operation is the process of heat exchange, where a controllable primary fluid exchanges heat with a secondary fluid. The objective is to control the outlet state of the secondary fluid in order to satisfy a set of indoor environmental requirements in terms of temperature, indoor air quality (CO2), and occupant comfort. While the thermodynamic principles describing heat exchange are well understood, their inherent complexity, non-linear behavior, and internal system constraints arising from the real-world operation render the control task notoriously difficult. Following ideas from recent literature the task of how to create data-driven parametrized models from data that ensure the satisfaction of a set of priors will be pursued. Extending the current literature, which mainly focuses on autonomous dynamics, studying the influence of the control action. Following the core idea of Meta-Learning, strategies for updating the parametric model based on the controller outcome will be analyzed and extended, ensuring the adapted model consistently preserves priors. The algorithms developed will be validated against high-fidelity simulation environment provided by an industrial-partner. Extension to real test-bench data is also possible.

The goal of the project is to develop online data-based identification and control algorithm which adapt to unknown scenarios and work in cross-domain solutions. The algorithms will be validate on detailed simulation environments coming from industrial partners and possibly on real test-bench data. The project can be divided in the following six main objectives: 1. Literature review on HVAC systems and thermodynamical priors; review on physics informed ML algorithms. 2. Define a parameter-based model ( N.N, Kernel-based, etc..) to learn the system dynamics. Formulate strategies for incorporating priors ( e.g., projection layers, physics-based structures such as PINNs, Hamiltonian-Systems, etc...) 3. Test the obtained model against a detailed simulation environment from industrial partner. 4. Formulate an optimal constrained control problem for temperature tracking ( possibly also additional references ) incorporating the data-based model. 5. Formulate a prior-based meta-learning update of the system's parameters based on the outcome of the controller. 6. Validation of the methods in a high-fidelity simulation environment. 7. (optional) Extend the results on test-bench data.

To apply send an email with your resume/CV, including lists of relevant publications and projects, and transcript of records in PDF format to Alessandro Cupo at calessandr@control.ee.ethz.ch and Aren Karapetyan at aren.karapetyan@belimo.ch

Please have a look at the attached document.

Please have a look at the attached document.

Please email your CV (including lists of projects and optionally any document helpful to evaluate your background) and current grade transcript in PDF to: schandraseka@ethz.ch. In addition, if you have any questions on the project, please do not hesitate to contact me.

The project will investigate how day-ahead simulation results can be translated into concrete demand-shifting recommendations for residents, and whether modest incentives induce a measurable behavioural response. Based on this framing, various research directions can be explored (e.g. developing a prototype that informs residents). Potential research questions are: ▪ Can day-ahead push notifications and incentives measurably shift the demand profile of residents? ▪ Which incentive structures are most effective at sustaining participation over time without relying on sensitive data? ▪ Which KPIs best capture current performances and future projections and how do they support activities of residents? You would be hosted and supported by the Digital Twin & Informatics group at UESL with direct access to the NEST living-lab demonstrator, its residents, the ehub research platform and existing simulation models. Smart watches and supporting infrastructure for a field study are available, alongside a collaborative, supporting, international research environment.

You will explore how day-ahead simulation results can be translated into demand-shifting recommendations that residents act on, drawing on building simulation, data analysis, and behavioural research to design your approach and evaluate whether incentives drive measurable change.

Christian Schmid christian.schmid@empa.ch