Reference no: EM131279875 , Length: word count:2000

Assignment Title: An Affective Intelligent System

1. General description

For this ICA, you need to develop an interactive affective intelligent systemin educational, health-care or any other application domains of your choice. You have learned several machine learning approaches for decision making in games and general AI applications. Emotion and personality modelling have also made anintelligent agent interesting and unique. For this ICA, you will require to practise your skills on AI and affective computing on the following aspects.

1.1 The requirement of the interface

- An interactive user interface needs to be developed. You may borrow any environment you developed previously as the basicplatform for the AI agents to live in. Or you may employ any existing engine (e.g. Slick2D) to plug your AI development in. If you haveproblems with the development of a sophisticated or an advanced application interface, then a text or console based simple user interface is also acceptable as long as you indeed show a good thinking and development on AI and affective components.

1.2 The requirements of core AI components

- To provide at least one learning functionality for the AI agent so that it is capable of learning from the environment and performing decision making.
- To develop a simple autonomous agent behaviour system for the AI agent to enable it to assist other agents or the human user better (e.g. in e-learning or healthcare applications), or oppose the human player effectively (e.g. in combat game applications).
- To equip the AI agent with emotion and personality modelling so that the agent is also able to respond to situations emotionally and accordingly. To construct a simple rule-based system for this component would be fine.
- To develop a simple package of the overall application development (e.g. simple AI functionalities with a simple graphical or text-based user interface).

1.3 Extra AI features to enhance your development

- To employ two machine learning approaches for decision-making. The chosen approaches are able to effectively cooperate with each other to perform advanced decision making.
- To develop an advanced autonomous agent behaviour system for the AI agentto enable it to assist other agents or the human player/user better, or oppose the human player more effectively;
- To implement a complex rule-based fuzzy logic or other statistical based emotion modelling component (e.g. the level of nervousness, interest, excitement or panic emotion developed);
- To provide an advanced speech or gesture based interaction;
- To develop an advanced package of the overall application development (e.g. using a client/server architecture, multi-threading, advanced character and application environment design and development with good graphical presentation, etc). But this feature will be awarded with a small score allocation since the main focus will be on the AI and affective features development.

1.4 Approaches to choose from for your implementation

- Rule-based Inference using expert systems
- Fuzzy logic
- Decision tree learning
- Naive Bayes classifier
- Neural networks
- Emotion modelling using OCC
- Five factors (OCEAN) for personality modelling
You should choose at least one machine learning method among the above mentioned several potential approaches (including rule-based inference, fuzzy logic, decision tree learning, naive Bayes classifier and neural networks) combined with other basic AI feature development such as a simple behaviour tree and a simple emotion and personality modelling for a passing level development. E.g. if you have developed a basic rule-based expert system with a simple emotion and personality modelling for a basic product development, you may expect a passing score.

2. Assessment allocation

Report: The report length will be approximately 2000 words. You need to mention the background research carried out, core features of your own development (design and implementation), evaluation, conclusion and future work. Citations and references should be used and presented properly.

Product : A software product should be bug free and able to be demonstrated in a suitable University lab. Video demos are also required to be included in your submission to illustrate your system's best AI-related performances. You may implement the ICA using any programming language of your choice.

3. The individual report

The individual report will be marked based on the following components:
- The research problem addressed
- The knowledge of relevant research context
- The originality of the work
- Product development stages (methodology, design and implementation)
- Critical evaluation
- The overall presentation of the work

Report
(40%) - No critical analysis of relevant work or it is poorly presented
- Points raised about approach taken and implementation are confusing and very limited
- There is very little evidence of testing, and most problems identified are left outstanding
- Poor originality
- Limited discussion on relevant work but it is well presented
- Discussion on approach taken and implementation stages is reasonable but limited
- Limited evidence of testing, and testing appears to be more about proving things work than finding what causes it to fail
- Limited originality - Reasonable level of discussion on relevant work and it is well presented
- Discussion on approach taken and implementation stages is sufficient and clear
- Some evidence of testing and some discussion on identified problems
- Some originality - Good sufficient critical analysis of relevant work and it is well presented
- Good and sufficient discussion on approach taken and implementation stages
- Good evidence of testing including using normal testing strategies and third party testing, and potential solutions for identified problems are provided
- Good originality - Excellent knowledge and critical analysis on relevant work
- Excellent, clear and concise discussion of approach taken and implementation stages
- Excellent evidence of testing including using various testing strategies and third party testing, and potential solutions for identified problems are provided and well discussed
- Excellent originality

Examples :

List of previous assignment ideas:

Disaster and rescue simulation based on injury assessment
A wool cutting game demo with an ally NPC, sheep NPCs and wolf NPCs
Zombie attack (good team AI)
A rule-based system to identify types of animals (e.g. a bird or a fish)
A pirate and rescue game demo with teammate NPC ships and pirate ships
A martial arts fighting game demo (an AI player vs a human player)
Spam email detection
Emotion detection from twitter messages
Football match results prediction
Bioinformatics applications (e.g. diabetes, flu, heart or other disease detection)
Face recognition
Facial emotion recognition (for 6 basic emotions)
Gesture recognition (e.g. command or emotional gestures)
Object recognition
Characters and digits recognition
Behaviour pattern prediction in games
Tennis match winning/losing prediction
Death round with one human player and two AI players