Final Year Project Instructions

Checklist for the projects

You can find a Checklist for Final Year Projects, which includes things I expect you to check before handing in the written from. If you are working with me, you will need to write in LaTeX, and here's a good place to start.

Implementation of a Natural Deduction proof generator for propositional logic, and hopefully eventually for First-order Logic to automatically generate exercises for Logic for CS. The likely approach for this one would be to use search (perhaps heuristic) together with some kind of proof assistant to check the generated proofs (e.g. JAPE) This software could gauge the difficulty of proofs based on the number of hypotheses, complexity of the deduction rules, or length of the proof -- Keywords: Logic, Search
Implementation of an HGN planner using a formalism like GoDel, ideally with Heuristics for HGN Planning, and cost-optimality guarantees -- Keywords: Automated Planning, Hierarchical Planning
Implementation of an HTN planner in Python capable of planning for both the SHOP/JSHOP2 formalism and the more recent HDDL formalism used in the HTN IPC. The implementation can either be a compilation technique, like our HyperTensioN planner, or remain completely internal implementation based on our PDDL parser -- Keywords: Automated Planning, HTN
- Implementation of new HTN heuristics within an existing HTN Planner -- Keywords: Automated Planning, Hierarchical Planning

Train reinforcement-learning based agents to play games in environment labs such as:
- OpenAI Universe Keywords: Reinforcement Learning, Deep Learning, Game AI
- DeepMind Lab Keywords: Reinforcement Learning, Deep Learning, Game AI -- Keywords: Reinforcement Learning, Deep Learning, AI for Computer Games
Implementation of a heuristic planner for incomplete PDDL domains in Python -- Keywords: Automated Planning
Implementation of a fragment of the PRS interpreter using Python -- Keywords: Multiagent Systems
Implementation of a MultiAgent Planner using a Graphplan base code -- Keywords: Automated Planning, Multiagent Systems
Implementation of HTN-based planning into AgentSpeak -- Keywords: Automated Planning, Multiagent Systems
Implementation of the CANPLAN semantics using an AgentSpeak-like syntax - -- Keywords: Multiagent Systems

Ideas for final year projects that I'd be keen to advise

Implement a new AI system for the Battle for Wesnoth game, they seem to have a tutorial for customising the AI (hopefully better than the one from a few years ago).
- Using reinforcement learning -- Keywords: Machine Learning, Game AI
- Connecting it to an AgentSpeak(L) interpreter -- Keywords: Agents, Game AI
- Using some kind of automated planner -- Keywords: Automated Planning, Game AI
Implement a level generator for Super Mario type games backed by a planning system (to design achievable levels) in a similar way to generating narratives automatically -- Keywords: Automated Planning, Game AI
Implement an HTN planner (from scratch, in Java or Python) that accepts these formalisms: -- Keywords: Automated Planning
- JSHOP2
- SHOP2
- PDDL Tasks - this is optional but pretty cool
Competitively try to play VizDoom -- Keywords: Machine Learning, Game AI

Implementation of a machine learning filter for arXiv daily feed based on affinity with past publications. This idea was inspired by the Toronto paper matching system used by conferences to select papers. This is rather similar to the existing ArXiV-Sanity -- Keywords: Deep Learning, NLP - Débora Pires

Implementation of a text summarization technique using deep recurrent neural networks and sentence to vector embeddings - Maurício Steinert
Implementation of a planning-driven game playing agent that interprets the state by learning a neural network-based transition function - Raphael Baldi
Implementation an AgentSpeak(L) interpreter in Python - André Leonhardt
Implementation of mini BDI interpreter in Python
Implementation of Adversarial HTN planning for RTS games - Matheus Redecker
Implementation of Montecarlo based game search for Settlers of Catan - Gabriel Rubin e Bruno Paz
Implementation of a bot for Spelunkbots - Martin Móre e William Martins
Implementation of mini BDI interpreter in LUA - Guillermo Borges
Application of Reinforcement Learning to a bomberman implementation - Leonardo Amado
Implementation of artifact-based components for robot control - Thales
Implementation of a search-based planner - Maurício Magnaguagno
Implementation of optimization techniques to the JavaGP version of Graphplan - Ramon Pereira e Fernando Giroletti