Project Details
Projekt Print View

GRK 1765:  System Correctness under Adverse Conditions - SCARE

Subject Area Computer Science
Term from 2012 to 2021
Project identifier Deutsche Forschungsgemeinschaft (DFG) - Project number 190214496
 
Final Report Year 2022

Final Report Abstract

The Research Training Group SCARE addressed computerised systems that are placed in an environment with which they cooperate, i.e., sense, control, and equip with unprecedented functionality. System correctness means that the cooperation between environment and system satisfies desired behavioural properties. This relationship depends on certain assumptions about the environment and the components of the system. The research training group systematically investigated the problem of system correctness under adverse, only partially predictable conditions which can influence the behaviour of the system, the system context, and the assumptions made for verifying correctness. SCARE studied three aspects of adverse conditions: • limited knowledge. • unpredictable behaviour. • changing structure of environment and system. SCARE pursued the following research themes: formal modeling, verification and analysis as well as constructive techniques, where formal methods are combined with engineering techniques, in particular the use of machine learning. SCARE achieved various advances of the state-of-the art: • novel correctness results were obtained for various types of systems: (graph) programs; stochastic, analog and hybrid systems; networked control systems; wireless sensor networks; automotive ad-hoc networks; controllers for traffic manoeuvres; concurrent systems; neural networks. • new means of system specifications were introduced: variants of logics for traffic manoeuvres and graph transformation systems; new automata model for hybrid systems; models representing interacting autonomous agents following rational strategies; retarded differential and hybrid systems. • novel automatic system verification algorithms were deployed: the first ever model checker for retarded differential equations; constraint solvers for stochastic satisfiability modulo theory over undecidable arithmetic theories; model checkers and synthesis algorithms for distributed systems represented by Petri nets and Petri games. We list highlights of the achievements: verification of dynamically typed programs; verification of graph programs against non-local graph specifications; a spatial logic for the safety of urban traffic manoeuvres; stochastic satisfiability modulo theory using random quantifiers over continuous quantifier domains; quantitative verification of stochastic hybrid systems; the first-ever model-checking algorithm handling retarded differential dynamics; reduction of Hilbertian-state dynamical systems to finite-dimensional hybrid state; Bayesian hybrid automata, a precise model of interacting rational agents under safety constraints; reliable routing in automotive ad-hoc networks subject to rapid topology changes; energy-efficient wireless sensing; applying genetic programming to configure and design replication strategies; improved understanding and modelling of aging effects in electronic systems; a methodology for HW/SW-optimization for predictable timing; model checking of Petri nets with transits against temporal flow properties; over-approximation in Petri net synthesis; exploiting symmetries of high-level Petri games in distributed synthesis; detecting adversarial input to neural networks with internal attacks; automatic verification of non-linear neural networks embedded into cyber-physical environments; nature-inspired dynamic optimization; machine learning methods applied to develop drugs against the coronavirus.

Publications

 
 

Additional Information

Textvergrößerung und Kontrastanpassung