Software Systems and Middleware

Computational foundation of modern cyber-physical and autonomous systems

Programmable hardware, operating systems, real-time execution platforms, communication middleware, and software supply chains have evolved from conventional information technology into safety-critical engineered products. Software is no longer a secondary implementation detail: in autonomous systems, it coordinates sensing, perception, decision-making, communication, actuation, diagnostics, updates, and fault handling. As a result, the behavior of the overall system increasingly depends on software architecture, timing predictability, update discipline, and the dependability of third-party components.

Software systems combine both general principles and domain-specific constraints. General aspects include:

• Abstraction;
• Modularity;
• Operating systems;
• Toolchains;
• Networking;
• Open-source ecosystems;
• Configuration management;
• Software lifecycle processes;
• Middleware services.

These concepts are common across information technology and cyber-physical systems. Domain-specific aspects appear when software interacts directly with the physical world. Ground systems, aircraft, marine systems, and spacecraft all rely on software, but they differ in timing constraints, certification culture, operational lifetime, redundancy strategy, update frequency, environmental uncertainty, and acceptable risk.

• Automotive systems combine large-scale production with over-the-air updates and functional-safety expectations;
• Airborne systems emphasise rigorous certification and partitioning;
• Marine systems often have long lifecycles and heterogeneous equipment;
• Space systems require remote autonomy, fault tolerance, and operation under extreme constraints.

Middleware is especially important because it connects application logic to hardware, networks, sensors, actuators, and other software components. In autonomous systems, middleware must support both flexibility and assurance. It enables modular development and reuse, but it also introduces dependencies, timing effects, integration risk, cybersecurity exposure, and possible failure propagation across components. Therefore, middleware architecture must separate safety-critical real-time functions from less critical IT-style services while still allowing communication, monitoring, diagnostics, and updates.

The main verification and validation challenges arise from complexity, interaction, and uncertainty. Autonomous systems operate in open environments, making it difficult to define complete requirements and exhaustive test scenarios. Real-time behaviour must be validated together with functional correctness, because a correct decision made too late may still be unsafe. Software also depends increasingly on machine-learning components, open-source packages, vendor libraries, compilers, operating systems, and communication stacks, making supply-chain assurance essential. Physical testing alone is insufficient, expensive, and sometimes unsafe, so V&V must combine model-in-the-loop, software-in-the-loop, processor-in-the-loop, hardware-in-the-loop, simulation-based testing, field monitoring, and structured safety cases. Autonomy shifts assurance from proving that a fixed controller works in known conditions to demonstrating that the system remains acceptably safe across operational design domains, degraded modes, updates, faults, and unexpected interactions.

A simplified architecture for software systems and middleware in autonomous systems is:

Software and middleware act as the central integration layer of autonomous cyber-physical systems. Their importance lies not only in enabling advanced functionality but also in determining whether autonomous behaviour can be maintained, updated, verified, secured, and trusted throughout the system lifecycle.