Reflex and healing architecture for software health management

Large-scale embedded systems require autonomous fault detection and recovery mechanisms since manual intervention is often infeasible, particularly in space missions where communication latency and system inaccessibility constrain responses. The reflex and healing approach provides a systematic framework for integrating fault management into system architecture, enabling systems to maintain reliable operation despite failures. This work is innovative in demonstrating hierarchical fault management that can scale to complex distributed systems.

This paper presents a reflex and healing architecture for software health management in complex embedded systems such as those used in space missions. The framework employs hierarchical reflex engines that detect discrepancies through monitoring, diagnose faults using Timed Fault Propagation Graphs, and execute fault mitigation strategies using state machines. The architecture supports both reactive and proactive healing actions coordinated across multiple hierarchical levels.

The paper demonstrates a three-level hierarchical management structure (global, regional, and local) with reflex engines at each level that can detect and respond to faults autonomously. The framework successfully integrates with the ARINC-653 avionics standard, enabling applicability to safety-critical real-time systems. Case studies show how the architecture enables both fault isolation and coordinated recovery actions across system components.