ESAT: A Tool for Animating Logic-Based Specifications of Evolvable Component Systems

Introduction

An increasingly important area of runtime monitoring is the incorporation of techniques for diagnosis and repair, for example, in autonomic control applications [9], in robotics, and in e-business process change [12]. In particular, a runtime monitor becomes a ’supervisor’ - a process which not only monitors but may evolve the running system dynamically. In [4], a framework for the logical modelling of hierarchically structured supervised component systems was set out. The modelling captures the following key behavioural concepts: at runtime, a supervisory component can (i) monitor its supervisee to ensure conformance against desired behaviour, (ii) analyse reasons for non-conformance, should that arise, (iii) evolve its supervisee in a pre-programmed way following diagnosis, or via external stimulus received from higher-level supervisory components. Structurally, components may contain sub-components, actions over the state of the component, and programs over the actions. In this logical framework, components are specified by first-order logic theories. Actions are either basic revisions to the state of the component or combinations of actions. Crucially, a supervisory component is treated as a logical theory meta to its supervisee, thus providing access to all facets of the supervisee’s structure. A supervisory component program is executed meta to its supervisee’s program. Synchronisation between the two may occur through a variety of schemes, from lock-step synchronisation to asynchronous execution with defined synchronisation points. A supervisory program action may evolve its supervisee by making changes to its state, to its actions, to its sub-components, or to its program. This occurs in the logical framework via a theory change induced from the meta-level.