Deep reinforcement learning with constraints and demonstrations
Reinforcement Learning (RL) has been successfully applied to a number of problems, such as robotic control, task scheduling, and telecommunications.
IFPEN’s doctoral thesis positions for next fall are now available!
Check them out by choosing “open” in the filter menu below.
Additional thesis subjects will be posted in the coming weeks: to receive them automatically by email, subscribe to our notification system.
309 Subject(s) of these
Filter by domain
Reinforcement Learning (RL) has been successfully applied to a number of problems, such as robotic control, task scheduling, and telecommunications.
Electrification of vehicles and improved efficiency of internal combustion engines (ICE) are the main levers to reduce greenhouse gas emissions. The second priority of the French strategy for the deployment of low-carbon hydrogen is the development of renewable hydrogen for use in fuel cells or for combustion in a spark-ignition engine (SIE).
The performance of simulators has a direct impact both on the quality of the simulation results and on the study of a wide variety of scientific hypotheses.
The drop in production costs of distributed energy production systems and electrochemical storage systems coupled with the evolution of regulations make it possible to build local energy management operations.
Environmental and health concerns are now accelerating behavioral changes related to personal mobility in an unprecedented way.
The recent advent of electric vehicles is pushing car manufacturers to design more compact electric motors running at higher speeds, leading to higher local heat generation. Cooling is therefore crucial to preserve the efficiency and the reliability of the electrical machine.
Alumina-based catalyst supports are used for biomass conversion processes under development. They consist of multi-scale porous materials, obtained in most applications after calcination of boehmite powders (aluminum hydroxide precursor of alumina) and composed of dense alumina nano-platelets with complex arrangements.
A significant reduction of greenhouse gases emissions is needed to keep global warming at an acceptable level in the next decades. In particular, the decarbonization of the energy and transport sectors is necessary and the use of hydrogen is a plausible solution as its consumption, through combustion processes or in fuel cells, is carbon-free.
In this thesis, we are interested in the modeling of fracture propagation. Historical formulations have two types of drawbacks. In the case of local damage models, the limitations come from the fact that the results depend on the computational mesh.
Context. Nowadays, we are witnessing a rapid spread of multimodal mobility in our cities and a willingness on the part of communities to promote new mobility behaviors.
Many fluids are used in the energy, transport and environmental sectors, for various applications ranging from the production of renewable energy to the mobility of people. These fluids are often complex mixtures whose components are mainly made up of hydrocarbon species.
Following the rapid emergence of electric vehicles in the automotive market, manufacturers are seeking to regularly improve their product’s performance to stand out from the competition.