Reasoning and Agents Coursework 2
This coursework is about designing and evaluating symbolic planning domains and problems using the Planning Domain Definition Language (PDDL). It is marked out of 100 and worth 15% of the overall course grade. It consists of three components: • Modelling: Creating PDDL problem and domain files for a given specification (35 marks)
ShoppEdTM, a successful supermarket chain, is seeking to expand its business via technical innovation! For this, they develop a robotic assistant called ShopBot, which has to travel within the supermarket to collect items designated by a shopping list and bring them to a checkout stand. Items of certain categories (e.g., fresh produce) may have to be weighed before checkout. The objective of this task is to design and test the PDDL domain and problem files for this scenario.
• Experiment: Designing an experiment to evaluate a planner (15 marks);
The current problem files are not challenging enough for the planner. For this subtask, design a harder problem called problem-1-hard.pddl, whilst keeping domain-1.pddl fixed. This problem instance will be used to benchmark the planner’s performance. Using domain-1.pddl and problem-1-hard.pddl, design an experiment to evaluate the effect of different values of wg and wh on the planner’s performance. Give your experiment results in the report. Include the analysis which should be brief and have both quantitative and qualitative elements.
• Extensions: Extending the domain to deal with real-world challenges (50 marks). The objective of this task is to extend the domain to make it closer to the real world. Inf2D_cw2.pdf
All code can be found in the reasoning coursework folder.
Completed coursework by Tanisha. Feel free to reach out if you have any questions or suggestions!