Ïã¸ÛÁùºÏ²Ê

Aims:

This module will explore how an intelligent system is to use its knowledge representation and to translate this into actions to achieve goals. We will consider situations in which behaviours are pre-programmed, and dynamic situations in which systems learn and adapt online using reinforcement learning. Students will be asked to implement their own intelligent systems which will bring together elements of learning and adaptation to support goal-seeking behaviours.

Intended learning outcomes:

On successful completion of the module, a student will be able to:

1. Understand and explain the underlying theory and challenges associated with motion planning and action taking with robotics systems.
2. Understand the theory and be able to implement common search-based planning algorithms (Dijkstra, A*).
3. Understand and explain the underlying principles of the exploitation vs. exploration trade-offs in unknown environments.
4. Develop a basic understanding of online and reinforcement learning.
5. Develop experience applying these algorithms to understand both their strengths and limitations.

Indicative content:

The following are indicative of the topics the module will typically cover:

• Formulation of motion planning problems.
• Derivation of discrete grid-based search problems for path planning including Dijkstra and A*.
• Methods of exploring unknown environments and to avoid obstacles.
• Online learning methods.

Students will develop a good understanding of how the science and engineering of intelligent systems can be applied to the design and control of robotic systems. The module will have a strong practical element where students will develop and apply software-based solutions for a range of problems.

Requisites:

To be eligible to select this module as optional or elective, a student must: (1) be registered on a programme and year of study for which it is a formally available; (2) have passed Cognitive Systems and Intelligent Technologies (COMP0014); and (3) have also selected Machine Learning and Neural Computation (COMP0036).