Mining assumptions for synthesis

Li, Wenchao; Dworkin, Lili; Seshia, Sanjit A.

doi:10.1109/memcod.2011.5970509

Cited by 88 publications

(123 citation statements)

References 28 publications

Supporting

Mentioning

123

Contrasting

Order By: Relevance

“…Acacia+ [9] also solves LTL realizability problems encoded as safety games. Another approach to realizability checking [35] builds upon RATSY using a template-based specification mining approach to identify situations of an underconstrained environment or an over-constrained system. This approach is complimented by work on detecting unrealizability due to overly-strong system guarantees or overly restricted signals [29].…”

Section: Specification Qualitymentioning

confidence: 99%

Specification: The Biggest Bottleneck in Formal Methods and Autonomy

Rozier

2016

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Advancement of AI-enhanced control in autonomous systems stands on the shoulders of formal methods, which make possible the rigorous safety analysis autonomous systems require. An aircraft cannot operate autonomously unless it has design-time reasoning to ensure correct operation of the autopilot and runtime reasoning to ensure system health management, or the ability to detect and respond to off-nominal situations. Formal methods are highly dependent on the specifications over which they reason; there is no escaping the "garbage in, garbage out" reality. Specification is difficult, unglamorous, and arguably the biggest bottleneck facing verification and validation of aerospace, and other, autonomous systems.This VSTTE invited talk and paper examines the outlook for the practice of formal specification, and highlights the on-going challenges of specification, from design-time to runtime system health management. We exemplify these challenges for specifications in Linear Temporal Logic (LTL) though the focus is not limited to that specification language. We pose challenge questions for specification that will shape both the future of formal methods, and our ability to more automatically verify and validate autonomous systems of greater variety and scale. We call for further research into LTL Genesis. Abstract. Advancement of AI-enhanced control in autonomous systems stands on the shoulders of formal methods, which make possible the rigorous safety analysis autonomous systems require. An aircraft cannot operate autonomously unless it has design-time reasoning to ensure correct operation of the autopilot and runtime reasoning to ensure system health management, or the ability to detect and respond to off-nominal situations. Formal methods are highly dependent on the specifications over which they reason; there is no escaping the "garbage in, garbage out" reality. Specification is difficult, unglamorous, and arguably the biggest bottleneck facing verification and validation of aerospace, and other, autonomous systems. This VSTTE invited talk and paper examines the outlook for the practice of formal specification, and highlights the on-going challenges of specification, from design-time to runtime system health management. We exemplify these challenges for specifications in Linear Temporal Logic (LTL) though the focus is not limited to that specification language. We pose challenge questions for specification that will shape both the future of formal methods, and our ability to more automatically verify and validate autonomous systems of greater variety and scale. We call for further research into LTL Genesis.

show abstract

Section: Specification Qualitymentioning

confidence: 99%

Specification: The Biggest Bottleneck in Formal Methods and Autonomy

Rozier

2016

Lecture Notes in Computer Science

View full text Add to dashboard Cite

show abstract

“…It computes a set P of candidate LTL formulas which can be used to refine guarantees of φ 1 . Then at each iteration, a formula ψ ∈ P is chosen, and it is checked if the counter-strategy for φ 2 satisfies ¬ψ (similar to assumption mining in [5]). If it does and ψ is consistent with φ 2 , it is checked if ψ is an assumption refinement for φ 2 , in which case ψ can be used to refine the guarantees (assumptions) of φ 1 (φ 2 , respectively), and ψ is returned as a suggested refinement.…”

Section: Compositional Refinementmentioning

confidence: 99%

“…The problem of refining the environment assumptions is also considered in [5,6]. Synthesizing distributed systems from global specification is a hard problem [7].…”

Section: Introductionmentioning

confidence: 99%

Pattern-Based Refinement of Assume-Guarantee Specifications in Reactive Synthesis

Alur

Moarref

Topcu

2015

Tools and Algorithms for the Construction and Analysis of Systems

View full text Add to dashboard Cite

Abstract. We consider the problem of compositional refinement of components' specifications in the context of compositional reactive synthesis. Our solution is based on automatic refinement of assumptions and guarantees expressed in linear temporal logic (LTL). We show how behaviors of the environment and the system can be inferred from counterstrategies and strategies, respectively, as formulas in special forms called patterns. Instantiations of patterns are LTL formulas which hold over all runs of such strategies, and are used to refine the specification by adding new input assumptions or output guarantees. We propose three different approaches for compositional refinement of specifications, based on how much information is shared between the components, and demonstrate and compare the methods empirically.

show abstract

“…Li et al [11] focus on mining environment assumptions for GR(1) specifications. We base our work on this paper, and use a similar approach for human-in-the-loop control systems.…”

Section: Related Workmentioning

confidence: 99%

Synthesis for Human-in-the-Loop Control Systems

Sadigh

Sastry

et al. 2014

Lecture Notes in Computer Science

Self Cite

View full text Add to dashboard Cite

Abstract-Several control systems in safety-critical applications involve the interaction of an autonomous controller with one or more human operators. Examples include pilots interacting with an autopilot system in an aircraft, and a driver interacting with automated driver-assistance features in an automobile. The correctness of such systems depends not only on the autonomous controller, but also on the actions of the human controller. In this paper, we present a formalism for human-inthe-loop control systems. Particularly, we focus on the problem of synthesizing a semi-autonomous controller from high-level temporal specifications that expect occasional human intervention for correct operation. We present an algorithm for this problem, and demonstrate its operation on problems related to driver assistance in automobiles.

show abstract

Mining assumptions for synthesis

Cited by 88 publications

References 28 publications

Specification: The Biggest Bottleneck in Formal Methods and Autonomy

Specification: The Biggest Bottleneck in Formal Methods and Autonomy

Pattern-Based Refinement of Assume-Guarantee Specifications in Reactive Synthesis

Synthesis for Human-in-the-Loop Control Systems

Contact Info

Product

Resources

About