System modelling for run-time verification and validation of autonomous systems

Abstract

The verification and validation of autonomous systems, and their relationship with systems engineering, is a challenging problem. They are tasked with achieving goals with reduced (or no) human interaction, and will be verified or validated with respect to the task they should achieve, rather than traditional systems which are assessed as whether they have the necessary capabilities to support an operator in achieving the task. This paper describes ongoing work to develop a systems modelling approach to allow design-time system models to be reused by an autonomous system at runtime. We begin by identifying the properties associated with the engineering of autonomous systems that differentiate them from other types of complex system. A framework to support the verification and validation of aspects of autonomous systems at runtime is then presented, and we develop a specific aspect of this framework -a run-time Computation Independent Model (CIM), using a language from the automated planning domain, the Planning Domain Definition Language (PDDL).

Towards Run-time Assurance of Advanced Propulsion Algorithms

“…22 Increasingly, run-time monitoring is being viewed as an important part of an integrated approach to the V&V and certification of future safety-critical systems-in conjunction with design-time verification and testing. 23 Some researchers have gone so far as to propose that the creation of monitors for a system to guarantee formally specified safety properties will allow certification (at least in part) to be performed at run-time. 24,25 …”