Course: CS7032: Agents, AI & Games
Lecturer: Saturnino Luz
Demonstrator: Fasih Haider
Lecture notes and slidesNote: the slides, notes, and practicals (lab sheets) below have been compiled into a single reader, which you may find more convenient. This reader will be updated as the course progresses, so make sure you download it from time to time so that you have the most recent version. Make sure you clean the cache of your web browser when before downloading.
- Combining TD and approximation architectures: Neuro Dynamic Programming
- Temporal difference methods
- Basic strategies for solving the Bellman Optimality Equations
- Defining the Reinforcement Learning problem: Markov Decision Processes
- Reinforcement Learning: Evaluative Feedback
- Machine Learning in games: introduction and 2 case studies
- Ant Algorithms for optimisation tasks
- Reactive Agents & Simulation
- Deductive agents & Utility
- Agents, definition and formal architectures
- Course Overview
- Evaluating Evaluative feedback and
Describing a Markov Decision Process
- Source code in R.
- TSP path finding with an ACO algorithm
- Source code and resources for TSA setting (gzipped tar file) used in this lab.
- Alternative TSA source code distribution packaged as an Eclipse project (thanks to Oscar Cassetti)
- TSPLIB: A TSP Benchmark collection
- Multi-agent simulation environments
- Source code for simulations used in this lab.
- Exploring and formalising Robocode, Robolog ... This week you will install and get acquainted with the Robocode environment. The excercises can be done and handed in during next week's lab.
Course projectThis year's project will consist in devising an AI approach to game playing which will be tested in a major Game AI competition. The competition will be chosen based on a survey to be carried out by the class, and presented after the Study Week.
- Preparation phase: Conduct a survey
of game AI competitions (to be presented on 25th Nov).
- See current assignemts of game competitions to students/teams. (Assignments made on a "first-come, first-served" basis.)
- Project specification: Building an agent architecture (controller or controllers), or module for one of the suggested AI Games competitions.
- General AI and Agents:
- Russel, S. and P. Norvig, Artificial Intelligence: a Modern Approach. 2nd Edition (preferably). Prentice Hall, 2002.
- M. Wooldridge, An Introduction to MultiAgent Systems. John Wiley & Sons, 2002. ISBN 0 47149691X.
- Reinforcement learning (aka neuro-dynamic programming)
- Bertzekas, D. and Tsitsiklis, J. Neuro-dynamic programming Athena Scientific, 1996.
- Sutton, R. and A. Barto An Introduction to Reinforcement Learning. MIT Press. 1998.
- Other references (papers etc):