You are required to read and agree to the below before accessing a full-text version of an article in the IDE article repository.

The full-text document you are about to access is subject to national and international copyright laws. In most cases (but not necessarily all) the consequence is that personal use is allowed given that the copyright owner is duly acknowledged and respected. All other use (typically) require an explicit permission (often in writing) by the copyright owner.

For the reports in this repository we specifically note that

  • the use of articles under IEEE copyright is governed by the IEEE copyright policy (available at http://www.ieee.org/web/publications/rights/copyrightpolicy.html)
  • the use of articles under ACM copyright is governed by the ACM copyright policy (available at http://www.acm.org/pubs/copyright_policy/)
  • technical reports and other articles issued by M‰lardalen University is free for personal use. For other use, the explicit consent of the authors is required
  • in other cases, please contact the copyright owner for detailed information

By accepting I agree to acknowledge and respect the rights of the copyright owner of the document I am about to access.

If you are in doubt, feel free to contact webmaster@ide.mdh.se

Formal Methods for Scalable Synthesis and Verification of Autonomous Systems

Fulltext:


Authors:


Publication Type:

Doctoral Thesis

Publisher:

Mälardalen University

ISRN:

1651-4238


Abstract

Autonomous systems (a.k.a., agents) are often designed to move and execute tasks, without or with little human intervention. As the agents are often involved in safety- or mission-critical scenarios, ensuring the correctness of mission planning (i.e., path finding and task scheduling) and collision avoidance is crucial for such systems. However, traditional verification approaches, such as testing, are not sufficient to provide such assurance.Formal methods such as model checking are well known for their rigorous verification based on mathematical models and logic rules, which provide guarantees of the absence of errors in system models. However, employing them entails tackling many challenges such as the complicated formal modeling and the scalability of the algorithmic methods. Additionally, the mission planning concerns the static and predictable factors in the working environment of the agents, such as stationary obstacles and predefined tasks, whereas the collision avoidance focuses on the dynamic and unpredictable factors, such as pedestrians. Consequently, certain questions arise in this context: (i) How can formal methods be applied in providing correctness-guaranteed solutions within a holistic framework that handles both the static mission planning and the dynamic collision avoidance?, and (ii) When the methods for realizing the agents' artificial intelligence, such as machine learning, involve large amounts of data, how to improve the scalability of formal methods when verifying the results of such methods? In this dissertation, we offer answers to the questions by developing solutions in form of new frameworks and algorithms targeting the mentioned problems, implementing the solutions in software tools, and evaluating their performance on real-world applications. We propose a two-layer framework for formal modeling and verification of agents. The framework separates the discrete mission planning from the continuous movement of agents, which is needed for collision avoidance verification. Additionally, different formal modeling and verification techniques are adopted in the two layers of the framework respectively.For mission planning, we design two types of tool-supported approaches, one based on graphic search, and one based on learning. The former is sound and complete, and supported by the tools UPPAAL and UPPAAL TIGA. However, the graphic-search approach is not scalable for large numbers of agents. The learning-based solution complements the graphic-search one, by handling more agents, being supported by UPPAAL STRATEGO. As a trade-off, the learning-based method is sound but not complete.For the verification of collision avoidance, we propose two solutions, the first one based on statistical model checking in UPPAAL SMC, and second one based on the symbolic model checking of UPPAAL STRATEGO. In the second solution, we transform the hybrid agent models, whose verification is undecidable, into a conservative over-approximation as a discrete-time model whose model checking is decidable. These results are proven as theorems in the dissertation.To support our methods, we develop a toolset named MALTA that enables the automation of model construction and mission planning, and provides a visualization of environment configuration and the resulting mission plans. By using MALTA, we experiment with our novel methods in an industrial use case: an autonomous quarry. The experiment results demonstrate the advantages and weaknesses of different methods used in different types of environments, as well as the applicability of our methods and tool in complex systems.

Bibtex

@phdthesis{Gu6448,
author = {Rong Gu},
title = {Formal Methods for Scalable Synthesis and Verification of Autonomous Systems},
isbn = {978-91-7485-552-4},
month = {May},
year = {2022},
school = {M{\"a}lardalen University},
url = {http://www.es.mdh.se/publications/6448-}
}