Home   Research Publications Members Related Software
IndexBrowse   BibliographiesMy selection
 Search: in   (word length ≥ 3)
Publication no #604   Download bibtex file Type :   Html | Bib | Both
Add to my selection
Fast and Slow Planning: An Empirical Study of Learning-based Hybrid Planning

Ashutosh Pandey, Ivan Ruchkin, Bradley Schmerl and David Garlan.

2019. Submitted for publication.

Online links:   Bibtex entry   Plain Text

Software systems increasingly rely on planning to make adaptation decisions autonomously. Planning is required to produce high-quality adaptation plans in a timely manner, and these two qualities are often in conflict. A promising approach, called hybrid planning, reconciles timeliness and quality by combining reactive planning approaches that quickly provide emergency responses, and a (deliberative) approach that takes time but determines a high-quality plan. While being effective at times, reactive planning sometimes risk making the situation worse. Hence, a major challenge in hybrid planning is to pick an appropriate reactive approach until the deliberative planning is ready with a high-quality plan. This paper introduces a novel learning-based approach to choose an appropriate reactive approach from the given set of reactive approaches. The new approach (on average) outperforms existing approaches that is based on specifying fixed conditions to invoke reactive approaches. Furthermore, to simplify the adoption of hybrid planning among engineers, we perform an empirical study of hybrid planning for a self-adaptive cloud-based system, and derive evidence-based guidelines instantiate hybrid planning effectively.

Keywords: Machine Learning, Self-adaptation.  
    Created: 2019-04-01 10:19:23
Feedback: ABLE Webmaster
Last modified: Mon February 12 2018 11:21:51