Quantifying controllability in temporal networks with uncertainty

Akmal, Shyan; Ammons, Savana; Li, Hemeng; Gao, Michael C.; Popowski, Lindsay; Boerkoel, James C.

doi:10.1016/j.artint.2020.103384

Cited by 9 publications

(13 citation statements)

References 6 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Dynamic controllability is much less restrictive than strong controllability, since information about previous, contingent events can be leveraged in decisions about the timing of executable events. Both strong and dynamic controllability were originally defined in the context of STNUs as binary properties of temporal networks, though recent work has proposed a degree of controllability metric for determining how close to controllable a network is (Akmal et al 2019).…”

Section: Simple Temporal Network With Uncertaintymentioning

confidence: 99%

“…Dynamic Control Dispatch (DC-Dispatch) The basic dynamic dispatch algorithm that we use is the early execution dispatch with inferred edges designed by Nilsson, Kvarnström, and Doherty (2014) to dispatch the uncontrollable PSTNs. This dispatcher works on controllable STNUs, and as shown by Akmal et al (2019), can, but is not guaranteed to, work on uncontrollable ones. The strategy works by generating additional (wait) constraints and using them to help decide when a contingent timepoint is enabled and can therefore be executed.…”

Section: Related Workmentioning

confidence: 99%

“…First, it extracts a bounded contingent edge from each probabilistic contingent edge in the original PSTN using a procedure we call EXTRACT-STNU-EDGES(E P , α). EXTRACT-STNU-EDGES takes in a set of probabilistic edges and a risk budget α, which represents the maximum risk of violating any constraint that we are willing to accept, and returns a set of bounded STNU edges, where each (potentially unbounded) probabilistic contingent edge has been converted to a bounded STNU edge by truncating Akmal et al (2019), which returns the STNU that corresponds to minimally "relaxing" the problem, in this case by squeezing the intervals associated with the contingent edges involved in the conflict. The terminology "relax" is used to reflect that we reduce the amount of uncertainty we need to control (by contracting, rather than expanding, contingent interval bounds), which makes the controllability problem easier to solve.…”

Section: Min-loss DCmentioning

confidence: 99%

“…Of this dataset, 85 had no success under any dispatch strategy, and were therefore omitted from the rest of our data analysis. The second is the adapted version of the CAR-SHARING dataset originally from Santana et al (2016), which was edited by Akmal et al (2019) to be to be consistent but not fully controllable. The original CAR-SHARING dataset contained a mix of PSTN-and STNU-style contingent edges.…”

Section: Empirical Evaluationmentioning

confidence: 99%

“…The original CAR-SHARING dataset contained a mix of PSTN-and STNU-style contingent edges. Akmal et al (2019) translated this into a benchmark containing pure STNUs by converting PSTN-style contingent edges to STNU-style contingent edges whose bounds encompassed two standard deviations on either side of the mean. We create an equivalent benchmark of pure PSTNs by running this conversion process in reverse.…”

Section: Empirical Evaluationmentioning

confidence: 99%

See 4 more Smart Citations

Dynamic Control of Probabilistic Simple Temporal Networks

Gao¹,

Popowski²,

Boerkoel³

2020

AAAI

Self Cite

View full text Add to dashboard Cite

The controllability of a temporal network is defined as an agent's ability to navigate around the uncertainty in its schedule and is well-studied for certain networks of temporal constraints. However, many interesting real-world problems can be better represented as Probabilistic Simple Temporal Networks (PSTNs) in which the uncertain durations are represented using potentially-unbounded probability density functions. This can make it inherently impossible to control for all eventualities. In this paper, we propose two new dynamic controllability algorithms that attempt to maximize the likelihood of successfully executing a schedule within a PSTN. The first approach, which we call Min-Loss DC, finds a dynamic scheduling strategy that minimizes loss of control by using a conflict-directed search to decide where to sacrifice the control in a way that optimizes overall success. The second approach, which we call Max-Gain DC, works in the other direction: it finds a dynamically controllable schedule and then attempts to progressively strengthen it by capturing additional uncertainty. Our approaches are the first known that work by finding maximally dynamically controllable schedules. We empirically compare our approaches against two existing PSTN offline dispatch approaches and one online approach and show that our Min-Loss DC algorithm outperforms the others in terms of maximizing execution success while maintaining competitive runtimes.

show abstract

Section: Simple Temporal Network With Uncertaintymentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Section: Min-loss DCmentioning

confidence: 99%

Section: Empirical Evaluationmentioning

confidence: 99%

Section: Empirical Evaluationmentioning

confidence: 99%

See 3 more Smart Citations