Real-Time Edge Computing for Cyber-Physical Systems: A Meta-Analysis of Energy Efficiency and Fault Tolerance

Abstract

Real-time edge computing has become an important computational approach for cyber-physical systems (CPS) requiring energy-efficient processing, low-latency responses, continuous service availability, and rapid recovery from hardware, software, and communication failures. This study conducted a quantitative systematic review and meta-analysis to evaluate energy efficiency, real-time performance, fault tolerance, and associated energy–reliability trade-offs across edge-enabled CPS environments. From 2,846 initially identified records, 162 studies were retained for quantitative synthesis, collectively representing 7,846 edge nodes, 128,460 connected CPS devices, 9,735 processing units, 1,926 edge servers, 41,736 communication links, 18,524 experimental repetitions, and 6,738 controlled fault-injection events. The pooled results showed that edge-enabled configurations reduced total energy consumption by 24.8%, average power demand by 18.6%, communication energy by 31.7%, and energy per completed task by 26.4%, while battery operational duration improved by 22.1%. Processing latency decreased by 35.8%, end-to-end response time by 32.6%, transmission delay by 41.2%, and deadline-miss rates by 38.5%, while throughput increased by 29.4% and deadline satisfaction improved from 81.6% to 93.2%. Across 118 studies reporting fault-tolerance outcomes, service availability increased from 94.1% to 98.3%, reliability from 92.8% to 97.6%, and successful recovery from 82.7% to 95.1%, while recovery time and service downtime decreased by 39.7% and 51.3%, respectively. Analysis of 91 studies simultaneously reporting energy and reliability outcomes showed that enhanced fault tolerance required an average 12.4% additional energy. Balanced adaptive configurations reduced energy consumption by 18.3%, maintained 98.8% availability, and reduced recovery time by 46.7%. Meta-regression explained 38.6% of between-study variation in energy outcomes, 44.2% in real-time performance, and 47.8% in fault-tolerance outcomes. All major pooled effects remained statistically significant at p < 0.001, while sensitivity and publication-bias analyses supported the robustness of the findings. The results demonstrated that adaptive edge configurations can effectively balance energy efficiency, real-time responsiveness, and resilient CPS operation.