Kilani, Asma (2018) Improving the efficiency of the Pre-Optimization Plan Techniques. Masters thesis, University of Huddersfield.

Automated planning is an important research area of Artificial Intelligence (AI). In classical planning, which is a sub-area of automated planning, attention is given to ‘agile’ planning, i.e., solving planning problems as quickly as possible regardless of the quality of solution plans. Obtaining solutions quickly is important for real-time applications as well as in situations of imminent danger. Post-planning optimisation techniques for improving the quality of solution plans are a good option for improving poor quality plans. Since such techniques are run as post-processing, this avoids situations where there is a risk of not having solution plans in time. This thesis focuses on an important sub-area of post-planning optimisation; that is, on identifying and removing redundant actions from solution plans. In particular, this study extends the existing Action Elimination and Greedy Action Elimination algorithms by introduce two approaches to improve their efficiency. The AE and GAE algorithms are thereby developed into the UAIAE and UGAIAE systems respectively. The key to our approaches is based on optimise the process while keeping the same elimination power’ (identifying and removing the same number of redundant actions). First approach improves the algorithms by considering situations where inverse actions are redundant, while the other identifies a subset of actions that cannot be present in any redundant actions set. This subset is named justified unique actions. The study’s approach to identifying this subset has been motivated by a promising heuristic approach called ‘landmarks’, which are facts or actions that cannot be eliminated to achieve the goal.

The approaches in this study have been empirically evaluated using several benchmark domains, as well as several planning engines that participated in the Agile track of the International Planning Competition 2014. In addition, they have been evaluated against state-of-the-art optimal and satisficing planners, as well as they are evaluated against a plan repair technique.

The methods of AE family can be understood as polynomial methods that improve the quality of a plan by removing redundant actions, or as tools to complement more sophisticated plan optimisation techniques.

Asma Kilani FINAL THESIS.PDF - Accepted Version
Available under License Creative Commons Attribution Non-commercial No Derivatives.

Download (1MB) | Preview


Downloads per month over past year

Add to AnyAdd to TwitterAdd to FacebookAdd to LinkedinAdd to PinterestAdd to Email