Towards priority-awareness in autonomous intelligent systems

Abstract

In Autonomous and Intelligent systems (AIS), the decision-making process can be divided into two parts: (i) the priorities of the requirements are determined at design-time; (ii) design selection follows where alternatives are compared, and the preferred alternatives are chosen autonomously by the AIS. Runtime design selection is a trade-off analysis between non-functional requirements (NFRs) that uses optimisation methods, including decision-analysis and utility theory. The aim is to select the design option yielding the highest expected utility. A problem with these techniques is that they use a uni-scalar cumulative utility value to represent a combined priority for all the NFRs. However, this uni-scalar value doesn't give information about the varying impacts of actions under uncertain environmental contexts on the satisfaction priorities of individual NFRs. In this paper, we present a novel use of Multi-Reward Partially Observable Markov Decision Process (MR-POMDP) to support reasoning of separate NFR priorities. We discuss the use of rewards in MR-POMDPs as a way to support AIS with (a) priorityaware decision-making; and (b) maintain service-level agreement, by autonomously tuning NFRs' priorities to new contexts and based on data gathered at runtime. We evaluate our approach by applying it to a substantial Network case. CCS CONCEPTS• Computing methodologies → Probabilistic reasoning; • Theory of computation → Sequential decision making; • Software and its engineering → Extra-functional properties; Designing software;

Efficient Difference Analysis of Guaranteeable Requirements for Fault-tolerant Self-adaptation

“…Consider the case in which all states belongs to (2 R−1 − 1) regions. In this case, (1) there are M states, (2) the region removal will happen at most (2 R−1 − 1) times in a state (lines 12-17), and (3) it needs to check the state's destination transition times after each removal (lines [20][21][22]. Therefore, the worst computational complexity of the former part is O(N a (2 R−1 − 1)).…”

RDMSim: An Exemplar for Evaluation and Comparison of Decision-Making Techniques for Self-Adaptation

“…To the best of our knowledge, RDMSim is the first simulator to be implemented for this domain. Using RDMSim, researchers can compare their self-adaptive decision-making solutions with other techniques, including ours [21]. We have executed experiments for each scenario presented here, using our own decision-making technique, called MR-POMDP++ [21].…”

“…Using RDMSim, researchers can compare their self-adaptive decision-making solutions with other techniques, including ours [21]. We have executed experiments for each scenario presented here, using our own decision-making technique, called MR-POMDP++ [21]. The results are provided in the RDMSimExemplar repository, ready to be used for comparison purposes.…”

“…For further validation purposes, we have applied reinforcement learning based decision-making to the RDMSim. We provide our initial evaluation results for the RDMSim using MR-POMDP++ [7], [21] as part of the RDMSimExemplar repository. MR-POMDP++ is based on Multi-Reward Partially Observable Markov Decision Process (MR-POMDP).…”

“…A Python version is also publicly available. Researchers can use these scenarios to test their specific decision-making techniques based on Reinforcement Learning [21], Multi-Criteria Decision Analysis [22] and Evolutionary Computation [3] among others. Researchers can also design their own scenarios by modifying the different parameter ranges.…”