Improving the Efficiency of Self-Organizing Emergent Systems by an Advisor

Dependable Risk-Aware Efficiency Improvement for Self-Organizing Emergent Systems

“…Using a limited period k allows for the recurring tasks to change. In [26], as in this paper, the extract action determines the recurring set of tasks by data mining the last k run instances with a clustering algorithm, precisely, Sequential Leader Clustering [12]. The clustering algorithm also provides, for each task, an identifier that can be used by exception rules.…”

“…This These actions are abstractions and obviously the concrete realizations of them will be application dependent. The details of the EIA are discussed in [26] along with an implementation for PDP. The remaining discussion will primarily be focused on the important aspects of each action.…”

“…In [26], we described a concept that improves the operational efficiency of self-organizing emergent systems while almost completely preserving their earlier described beneficial properties. This concept is called an efficiency improvement advisor (EIA).…”

“…Similar to previous work in [26], [13], and [27], we use pickup and delivery problems (PDP) [23] as the application domain for the experimental evaluation of this concept. Our results show that the RA-EIA, by taking advantage of the time available through independent offline execution, is able to provide minimal risk-averse advice achieving improved performance compared to the original EIA.…”

“…the type of advice), presented in [26], were termed ignore rules, which suggest to an agent that it should ignore a particular task. Ignore rules act as a minor influence on agent behavior, as demonstrated in [13], where they showed low potential for exterior abuse (i.e.…”