Autonomous Agents
Assignment (Part III)
EMATM0042 – Intelligent Information Systems
Monday 18 March – Part 3
kevin.mcareavey@bristol.ac.uk
Assignment
Overview (1)
Groupings: each group has 5 students
Try to find your group members by yourselves first
Send me your group member names (up to 5) by one member of your group by …
I will merge groups where there are less than 5 students in a group
Assessments approaches
Part I as a written report. (20%)
Part II as a group presentation in class, each group will have 5 minutes for presentation plus 2 minutes questions (10%)
Part III as a written report, plus programming code, and video for your system (more detail to follow on this) (20%)
Peer-assignment
To evaluate contributions from group members, peer assignment will be used as weights to adjust group marks. Each member will evaluate other 4 members with a score out of 5. For example, full contribution of a member will be assigned a score 5. If all 5 members have equal contributions, each member will have a score 20 and the group report mark will be the result for every member.
Assignment
Overview (2)
Overall requirements: Create a real-world scenario, the scenario shall allow you to
Part I (Submit to SAFE)
Define a set of suitable rules (between 10 to 15) with reasonable complexity. Rules that could lead to inconsistency.
Use some specific facts (e.g., like Tom) to trigger rules and to illustrate how conflicting conclusions can be drawn.
Method(s) used for conflict resolution and why.
Identify data and attributes. Define a dataset (or multiple datasets) and populate your datasets. Use WEKA to analyse your data (e.g., Decision tree or other algorithms). Compare their performances.
Generate rules from the outputs of your data analytic algorithms and compare your rules from (1) with rules obtained in (5).
Part II (during the Lab session)
Class presentation
Part III (submit to SAFE)
Allow you to extend the scope to create multiple agents
Define agent’s beliefs/actions etc
Implement your scenario using Jason
Assignment
Part I: Rules
Overall requirements: Create a real-world scenario, the scenario shall allow you to
Part I (Deadline: 8th March 2019 midnight. Submit to SAFE)
Define a set of suitable rules (between 10 to 15) with reasonable complexity. Rules that could lead to inconsistency.
Use some specific facts (e.g., like Tom) to trigger rules and to illustrate how conflicting conclusions can be drawn.
Method(s) used for conflict resolution and why.
Identify data and attributes. Define a dataset (or multiple datasets) and populate your datasets. Use WEKA to analyse your data (e.g., Decision tree or other algorithms). Compare their performances.
Generate rules from the outputs of your data analytic algorithms and compare your rules from (1) with rules obtained in (5).
Hand-in: one copy per-group including: (maximum 15 pages per report)
2-page scenario (maximum), 1 or 2-page rules (English sentences, then IF-THEN rules),
1-page data/facts, 1 or 2 pages on working examples (Point 2 above), 1-page on conflict resolution.
1-page sample of your dataset. 1 or 2 pages of WEKA output of your decision tree (J48). 1 or 2 pages of output of another algorithm (with 10-fold validation to generate accuracy and error rate. 1 page for comparisons of these two algorithms on which algorithm is better and why.
1-page of analysis how close or different your original rule set is from the rule set obtained from J48 Decision Tree.
NOTE: scenario complexity, rules and exemplar instances are all taken into account when marking.
Assignment
Part II: Presentation
Overall requirements: Create a real-world scenario, the scenario shall allow you to
Part II (Date: 2nd May, during the Lab session)
Class presentation
Format
Each group has 5 minutes to present, every member needs to present something using 1 minute.
Then there are 2minutes for questions.
Note: we have 70 students, so roughly 14 groups in total. 14*7=98 minutes (net) time required.
Assignment
Part III: Agents & Multi-Agent Systems
Overall requirements: Create a real-world scenario, the scenario shall allow you to
Part III
Implement an AgentSpeak-style agent program in Jason (i.e. initial belief base, initial goals, plan library).
Implement a virtual environment in Jason with Java.
Implement custom environment actions and make use of those actions in your plan library.
Implement multiple agents within the same virtual environment (i.e. implement a Jason multi-agent project).
Implement agent communication within your virtual environment and make use of communication in your plan library.
Implement a graphical interface for your virtual environment.
Hand-in: one copy per-group including:
Written report (maximum 4 pages): maximum 2 pages to explain scenario, plus maximum 2 pages to highlight key features according to criteria outlined above.
Programming code: a ZIP file containing everything required to run your system, and a short README text file with instructions for how to do so (e.g. https://github.com/kevinmcareavey/secproblog/blob/master/README.md)
Video (screencast): a short video (e.g. 3-minutes) to demonstrate your scenario and the key features.
NOTE: every project should address criteria 1-3 as a minimum (implementation complexity will be taken into account when marking).
Can be the same scenario as before, or can be a completely new scenario
Examples
Industrial control
Smart trains
Smart grid
Demos
Smart Grid
Multi-agent system
Solar park agent
Wind farm agent
Battery park agent
Distribution sub-station agent
Distribution transformer agent
House agent
Watch out for…
User input (i.e. simulated events)
Belief change
Uncertainty modelling
Information fusion
User input
Beliefs
Events
Sensor data
Demos
Smart Trains
Multi-agent system
Train agent 1
Train agent 2
Watch out for…
User input (i.e. simulated events)
Belief change
Uncertainty modelling
Decision-making under uncertainty
User input
Beliefs
Demos
Industrial Control
Multi-agent system
Boiler agent
Human (user)
Watch out for…
User input (i.e. requests)
Belief change
Uncertainty modelling
Decision-making under uncertainty
User input
Actions
Sensor data
Bibliography
Michael Wooldridge. An Introduction to Multiagent Systems. John Wiley & Sons, 2nd edition, 2009.
Stuart J. Russell & Peter Norvig. Artificial Intelligence: A Modern Approach. Prentice Hall, 3rd edition, 2009.
Hector Geffner & Blai Bonet. A Concise Introduction to Models and Methods for Automated Planning. Morgan & Claypool, 2013.
Anand S. Rao. AgentSpeak(L): BDI agents speak out in a logical computable language. In Proceedings of the 7th European Workshop on Modelling Autonomous Agents in a Multi-Agent World (MAAMAW’96), pages 42-55, 1996.
Rafael H. Bordini, Jomi Fred Hübner, & Michael Wooldridge. Programming Multi-Agent Systems in AgentSpeak Using Jason. John Wiley & Sons, 2007.
Suggested reading
HandBrake 1.2.2 2019022300
HandBrake 1.2.2 2019022300
/docProps/thumbnail.jpeg