Architectural modeling and performance analysis of event-driven real-time mobile systems

Abstract

Purpose. Development of a formal methodology for architectural modeling and performance analysis of event-driven real-time mobile systems, addressing bounded response times, energy minimization, and predictable behavior under resource constraints and dynamic network conditions. Method. Integration of algebraic modeling, graph-theoretical analysis, and WCET estimation. System formalized as a tuple ⟨𝐸𝐸𝐸𝐸, 𝐻𝐻𝐻𝐻, 𝐶𝐶𝐶𝐶, 𝑇𝑇𝑇𝑇, 𝑅𝑅𝑅𝑅⟩ representing events, handlers, mappings, temporal constraints, and execution locations. WCRT analysis establishes temporal guarantees: WCRT eWCETh h ij comm j queue . AdaptiveEventOffloading algorithm implements multi-criteria optimization for dynamic execution placement. Results. A comparative analysis of four architectural strategies was conducted: local processing, cloud offloading, edge computing, and adaptive hybrid distribution. Local processing provides the highest determinism but increases energy consumption; cloud offloading reduces device energy but introduces latency variability; edge computing achieves intermediate trade-offs; adaptive hybrid strategies achieve balanced multi-dimensional performance. The WCRT-based analytical evaluation provides provable upper bounds on response time at the design stage without profiling. Scientific novelty. The study integrates algebraic system specification, WCRT analysis, and graph-based dependency modeling into a unified architectural evaluation framework. An explicit multi-dimensional trade-off model is developed between latency, energy, and determinism within device–edge–cloud hybrid systems. Adaptive execution placement is formalized as a constrained multi-criteria optimization problem. The work proposes a structural shift from implementation-level profiling to architectural-level predictive modeling. Practical importance. The methodology enables systematic design of event-driven mobile real-time systems through early-stage architectural decisions and supports temporal prediction without full implementation. Application domains include autonomous navigation, augmented reality, intelligent transportation, and telemedicine.