EPFL - Ecole Polytechnique Fédérale de Lausanne

Open Opportunities

Automating Reward Function Design Using LLMs for Robotic Disassembly of EV Batteries EPFL - Ecole Polytechnique Fédérale de Lausanne Automatic Control Lab This project explores the integration of large language models (LLMs) into a reinforcement learning (RL) pipeline to automate reward function design for the robotic disassembly of end-of-life electric vehicle (EV) batteries. By iteratively refining reward functions using LLMs, the approach aims to enhance training efficiency, accelerate learning, and improve the precision and safety of complex disassembly tasks. Utilizing Nvidia Isaac Sim for simulation and transferring skills to real-world robots, the research seeks to reduce human intervention in reward engineering, providing scalable solutions for advanced robotic manipulation in battery recycling and beyond. Intelligent Robotics, Mechanical Engineering, Systems Theory and Control, Text Processing, Virtual Reality and Related Simulation Master Thesis

Optimizing HVAC Commissioning: A Deep Learning Approach for Parameter Prediction and Error Reduction EPFL - Ecole Polytechnique Fédérale de Lausanne ENAC - Civil Engineering Section The project is a collaboration between EPFL IMOS Lab and Belimo exploring Deep Learning algorithms for commissioning parameters of Heating, Ventilation, and Air Conditioning systems Engineering and Technology Master Thesis

Physics-Inspired Graph Neural Networks for Modeling of Building Thermal Dynamics EPFL - Ecole Polytechnique Fédérale de Lausanne ENAC - Civil Engineering Section This project is a collaboration between the EPFL Intelligent Maintenance and Operations Lab and the EMPA Urban Energy Systems Lab exploring Building Thermal Modeling by fusing Graph Neural Network models with Reduced Order Models to optimize building energy system operation. Engineering and Technology Master Thesis, Semester Project

Fusing Physics-based and Machine Learning Models for Building Energy Modeling EPFL - Ecole Polytechnique Fédérale de Lausanne ENAC - Civil Engineering Section This project is a collaboration between the EPFL Intelligent Maintenance and Operations Lab and the EMPA Urban Energy Systems Lab exploring building thermal modeling that combines data-driven and physical modeling to optimize building energy system operation. Engineering and Technology Bachelor Thesis, Master Thesis, Semester Project

Motion planning for avian-inspired morphing drones in cluttered environments EPFL - Ecole Polytechnique Fédérale de Lausanne Institute of Mechanical Engineering At the Laboratory of Intelligent Systems, we study autonomous flight with avian-inspired morphing drones. By actively varying the wingspan and leveraging avian-inspired actuation of wings and tail, morphing-winged drones can fly with higher agility and pass through narrower gaps between obstacles than their fixed-wing counterparts. This can save crucial mission time when flying in indoor environments or cluttered forests. To best leverage the abilities of a morphing-drone to find a shortest path through a field of obstacles of varying size and density, a three-dimensional local motion planning strategy must be developed. Flight Control Systems, Flight Dynamics, Mechanical Engineering Master Thesis

Graph Neural Networks for Cascading Failure Analysis in Power Grids EPFL - Ecole Polytechnique Fédérale de Lausanne ENAC - Civil Engineering Section This project aims to investigate the use of GNNs for cascading failure analysis in power grids. The primary goal is to evaluate various GNN architectures and, if possible, develop new models tailored to the task. Artificial Intelligence and Signal and Image Processing Master Thesis, Semester Project

Frictional Instabilities in Topologically Interlocked Structures EPFL - Ecole Polytechnique Fédérale de Lausanne ENAC - Civil Engineering Section Topologically interlocked structures (TIS) represent a new class of innovative designs inspired by the mechanics of puzzles. Constructed from individual building blocks that interlock without the use of adhesives, these structures exhibit remarkable mechanical properties, relying solely on contact and frictional forces for their integrity. Experimental observations have revealed sudden failures and sharp load drops in TIS, indicating that frictional slip instabilities play a significant role in their structural response. This project aims to explore the influence of stick-slip frictional instabilities and interfacial heterogeneity on the failure mechanisms of TIS. Using the level-set discrete element modeling framework, the student will investigate the dynamic behavior of these systems under various conditions. The project offers an opportunity to delve into the unique mechanical behavior of TIS and gain experience with modern computational tools in structural engineering. Civil Engineering, Materials Engineering Master Thesis, Semester Project

Multiphysics Modeling of Fault Gouge EPFL - Ecole Polytechnique Fédérale de Lausanne ENAC - Civil Engineering Section The fault gouge, a layer of cohesionless material formed by fragmentation of parent rock, plays a key role in the macroscopic frictional behavior of faults, including their stability and energy release. This material exhibits complex behavior influenced by mechanical deformation, thermal effects and pore fluid flow. In this project, we utilize a combination of discrete and continuum simulations to investigate gouge rheology. In particular, the student will explore the effect of material heterogeneity and grain-scale characteristics on the macroscopic behavior, including the influence of particle fracture. Additionally, phenomena arising from hydromechanical and thermomechanical coupling will be studied. The findings from the project aim to provide new insight into earthquake mechanics. Civil Engineering, Earth Sciences, Materials Engineering, Physics Master Thesis

Physics-informed Deep Neural Operators for Solid Mechanics EPFL - Ecole Polytechnique Fédérale de Lausanne ENAC - Civil Engineering Section Recently, physics-informed neural operators (PINOs) have been introduced as a new approach for solving complex problems in engineering, by combining data with knowledge of the underlying governing equations. The concept is an extension of previously successful purely data-driven deep neural operators. In this project, the student will explore the application of PINOs on solid mechanics problems, with the goal of simulating the behavior of materials under various loading conditions. Applications will be considered in the context of geotechnical or structural engineering. The generalization capabilities of the method will be evaluated, and its accuracy will be compared to conventional numerical solutions. The findings aim to advance computational tools for engineering design and analysis, bridging the gap between traditional numerical methods and scientific machine learning. Aerospace Engineering, Civil Engineering, Mechanical and Industrial Engineering Master Thesis, Semester Project

Segregation in Granular Flows EPFL - Ecole Polytechnique Fédérale de Lausanne ENAC - Civil Engineering Section Segregation in granular flows is an important phenomenon influencing various natural and industrial processes, from landslides to pharmaceutical manufacturing. This project investigates the role of particle shape in segregation dynamics within granular flows. Using numerical simulations, the student will analyze how particles of varying shapes and sizes segregate under shear and gravity-driven conditions, by analyzing the forces developed between particles. The final goal is to provide insights into the continuum modeling of segregation phenomena in different scenarios. Civil Engineering, Earth Sciences, Physics Master Thesis, Semester Project