Syllabuses - PG

CS823 - Reasoning for Intelligent Agents

TIMETABLETEACHING MATERIAL
Credits20
Level5
SemesterSemester 1
AvailabilityMandatory
PrerequisitesN/A
Learning Activities Breakdown
Lectures: 20 hours | Practical/Labs: 20 hours
Homework / Private Study: 200
AssessmentTwo individual assignments worth 50% | 2-hour written exam 50%
LecturerMohamed Elawady

Aims and Objectives

The aim of this module is to endow students with:

  • Mathematical concepts and numerical reasoning.
  • Algorithmic programming.
  • Problem analysis and modelling.

Learning Outcomes

After completing this module participants will be able to:

  • Define the problem of AI as it relates to autonomous systems in fully-observable, partially-observable, adversarial, and real-time environments.
  • Understand key AI techniques for deliberative reasoning, such as Monte-Carlo Tree Search and Reinforcement Learning, and how to apply them.
  • Understand AI algorithms for the integrated reasoning and action of autonomous agents.
  • Become proficient with common tools and libraries used for autonomous robotics.
  • Program in Python to build autonomous agents that act efficiently in challenging environments to optimise reward, achieve goals, and collaborate with others.

Syllabus

  • What is AI for autonomous systems? Foundations, history and related disciplines. The state of the art in modern AI including notable applications and successes.
  • Solving problems by searching: formulating problems as a search for a solution, uniformed search algorithms, informed (heuristic) search algorithms, formulating heuristic functions.
  • Introduction to the problem of AI as it relates to autonomous systems. Defining, understanding, and modelling:
    • Uncertainty: Partial-observability and non-determinism.
    • Numeric problems: Infinite state-space search, continuous change, and general heuristic functions.
    • Temporal problems: Durations, deadlines, time-windows, and online deliberation.
    • Adversarial problems.
  • Beyond classical search: optimisation algorithms, partially observable Markov decision processes, risk-aware planning, integrating learning and search.
  • Techniques for integrated deliberation and execution: defining data-structures that describe behaviours, analytical algorithms, and constraint satisfaction. Implementing key algorithms that perform and connect search, repair, and autonomous action.
  • Distributed autonomous systems and composition of behaviours: understanding how autonomous systems can collaborate with limited communication to act efficiently as a collective system.
  • Architectures of autonomous agents: introduction to software tools and libraries for building large-scale autonomous systems and practical exercises implementing search and execution algorithms on both simulated and physical platforms.

Recommended Reading

This list is indicative only – the class lecturer may recommend alternative reading material. Please do not purchase any of the reading material listed below until you have confirmed with the class lecturer that it will be used for this class.

Artificial Intelligence: A Modern Approach. Russell, S. and Norvig, P. 4th Edition, Pearson, 2020. | Stocked at Andersonian Library (Other retailers are available)

Last updated: 2023-09-29 19:15:48