aidanscannell / phd-thesis

Bayesian Learning for Control in Multimodal Dynamical Systems | written in Org-mode

PhD Thesis: Bayesian Learning for Control in Multimodal Dynamical Systems

This repository contains the org and LaTeX source for my PhD thesis.

Publications & Code

A lot of the work in my thesis is yet to be published. Nevertheless, an initial version of the multimodal dynamics model and one of the mode remaining trajectory optimisation algorithms are published in:

Trajectory Optimisation in Learned Multimodal Dynamical Systems via Latent-ODE Collocation Aidan Scannell, Carl Henrik Ek, Arthur Richards

The work in each of the content chapters is roughly split into the following code bases:

Identifiable Mixtures of Sparse Variational Gaussian Process Experts Aidan Scannell, Carl Henrik Ek, Arthur Richards

Mode Remaining Trajectory Optimisation Aidan Scannell, Carl Henrik Ek, Arthur Richards

Mode Remaining Exploration for Model-Based Reinforcement Learning Aidan Scannell, Carl Henrik Ek, Arthur Richards

Instructions to Build PDF

Generate phd-thesis.pdf from phd-thesis.org using

emacs --batch -l init.el phd-thesis.org -f org-latex-export-to-pdf --kill

This uses the Emacs LaTeX exporter so I provide a minimal Emacs configuration in init.el and export to pdf in batch mode.

The Dockerfile creates a working environment which can be built with

docker build -t emacs-image .

Cite

@phdthesis{scannell22,
    title = {Bayesian Learning for Control in Multimodal Dynamical Systems},
    author = {Aidan Scannell},
    school = {University of Bristol},
    year = {2022}}