Mohammed Nassim Lacheheub

In today’s rapidly evolving digital landscape, the development of environmentally sustainable and adaptive software systems is gaining prominence. This article proposes a novel spatio-temporal layered framework designed to enhance self-adaptive system capabilities by structuring adaptation along both spatial and temporal dimensions. The contribution introduces a multi-layer model that integrates modular system decomposition (spatial layer) with time-aware reasoning across continuous, short-term, and long-term horizons (temporal layer). Through the use of change point detection, machine learning-based forecasting, rule-based inference via lookup tables, and probabilistic process models, the system supports both reactive and proactive decision-making guided by multi-objective optimization criteria. A case study focused on analyzing the energy consumption, carbon footprint, and execution time of native Python data structures demonstrates the framework’s ability to guide informed and context-aware adaptation strategies. The results showcase a principled and extensible methodology for constructing energy-aware, self-managing systems under uncertainty, with practical implications for green computing and runtime optimization.