GithubHelp home page GithubHelp logo

eleurent / phd-thesis Goto Github PK

View Code? Open in Web Editor NEW
13.0 3.0 4.0 16.85 MB

My PhD thesis. I defended on the 30th of October, 2020! See https://github.com/eleurent/phd-defense/

License: MIT License

TeX 99.71% Python 0.29%
phd-thesis phd-dissertation phd-thesis-latex

phd-thesis's Introduction

Safe and Efficient Reinforcement Learning for Behavioural Planning in Autonomous Driving

Social Attention Risk Awareness Robust Control

This repository holds the LaTeX code of my PhD Thesis.

Build LaTeX

Publications in this thesis

Robust-Adaptive Control of Linear Systems: beyond Quadratic Costs
Edouard Leurent, Denis Efimov, and Odalric-Ambrym Maillard
NeurIPS 2020
Paper Code
Monte-Carlo Graph Search: the Value of Merging Similar States
Edouard Leurent and Odalric-Ambrym Maillard
ACML 2020
Paper Code
Robust-Adaptive Interval Predictive Control for Linear Uncertain Systems
Edouard Leurent, Denis Efimov, and Odlaric-Ambrym Maillard
CDC 2020
Paper Code
Budgeted Reinforcement Learning in Continuous State Space
Nicolas Carrara,* Edouard Leurent,* Romain Laroche, Tanguy Urvoy, Odalric-Ambrym Maillard, and Olivier Pietquin
NeurIPS 2019
Paper Code
Social Attention for Autonomous Decision-Making in Dense Traffic
Edouard Leurent and Jean Mercat
ML4AD Workshop at NeurIPS 2019
Paper Code
Interval Prediction for Continuous-Time Systems with Parametric Uncertainties
Edouard Leurent, Denis Efimov, Tarek Raïssi, and Wilfrid Perruquetti
CDC 2019
Paper Code
Practical Open-Loop Optimistic Planning
Edouard Leurent and Odalric-Ambrym Maillard
ECML-PKDD 2019
Paper Code
* Equal contribution.

Defense

I defended on the 30th of October 2020, in front of the jury composed of:

  • Lucian Buşoniu
  • Jorge Villagra
  • Luce Brotcorne
  • Marc Deisenroth
  • Denis Efimov
  • Odalric-Ambrym Maillard

The slides 📰 that I used are also available online, and open-source.

How to build the PDF?

I recommend using latexmk to build this project.

latexmk -pdf PhD_thesis__Edouard_Leurent.tex

It is used e.g. in the latex-action of the automatic build workflow. LaTeX editors such as Texstudio can be configured to use latexmk as the default compiler.

Credits

The template is inspired from @Naereen/phd-thesis, also released under the MIT License.

phd-thesis's People

Contributors

eleurent avatar

Stargazers

 avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar

Watchers

 avatar  avatar  avatar

phd-thesis's Issues

Chap. 1 : Introduction

Sections principales:

  • Intro: Moral philosophy / ethics and the science of decision-making
  • Nuts and bolts of AD
  • Context and Scope: Quel est le problème, quels sont nos objectifs ? (décisions moyen terme) Où se situe la difficulté ? (incertitude, interactions) Pour cela, quelles sont nos armes ?
  • Contributions: comment avons-nous approché le problème ? Quels obstacles avons-nous franchi ?
  • Outline

Chap. 3: Problem Statement

Reprendre la checklist du Chap 2, en précisant les choix pris dans cette thèse (par opposition au SOTA)

  • State: positions et vitesses des objets, voies, potentiellement toute information.
  • Actions: on ne veut pas optimiser le contrôle bas-niveau (e.g. confort), mais des objectifs sémantiques court terme. Structure discrète.
  • Dynamique: modèle cinématique bicycle. Modèles de comportement pour les autres agents, issus de la simulation du traffic. Ils doivent pouvoir réagir au comportement du véhicule. Contrôleur retour d'état pour l'égo.
  • Récompense: la plus simple possible: velocity et collision. En particulier, on ne veut pas spécifier tous les aspects du comportement par la reward (e.g. distance de sécurité). Si possible, on ne donne que nos objectifs finaux et on souhaite voir le comportement adéquat émerger (la distance de sécurité est nécessaire et optimale à cause de l'incertitude sur le comportement du véhicule de devant, en particulier en cadre worst case vs en cadre average)
  • Implémentation: quelques mots sur highway-env, et renvoi en annexe.

Chap. 2: Literature Review

Tour d'horizon: comment les chercheurs ont-ils approché ce problème ?
Quels obstacles ont-ils rencontrés, et comment ont-ils essayé d'y répondre ?

Idée: un cadre général, et à chaque étape de modélisation un obstacle pratique: le réel est difficilement réductible à un MDP.

  • Motion planning
  • Imitation learning
  • Reinforcement learning
  • State and partial observability
  • Actions and temporal abstraction
  • Rewards and IRL
  • Dynamics, simulation and transfer sim2real
  • Optimisation criterion and safety

Il faudrait parler des approches contrôle / motion planning bas niveau. A quel endroit ?

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google ❤️ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.