CPFT-MOSA: A Comprehensive Parallel Fault-Tolerant Multi-Objective Simulated Annealing Framework for UAV-Assisted Edge Computing in Smart City Traffic Management

UAV-assisted edge computing has emerged as a critical technology for real-time monitoring and control in smart city traffic management systems. A key challenge in these systems is efficiently distributing computational tasks across available edge nodes while ensuring system reliability and performance, particularly for resource-intensive applications like object detection. This paper aims to address these challenges by proposing CPFT-MOSA (Comprehensive Parallel Fault-Tolerant Multi-Objective Simulated Annealing), a novel task allocation optimization framework specifically designed for UAV-assisted edge computing in traffic management scenarios. Our approach makes three primary contributions: 1) a multi-objective optimization framework that minimizes active nodes, optimizes energy distribution, and reduces execution time while ensuring fault tolerance through parallel task execution; 2) an integrated YOLO-based real-time task prioritization system that dynamically adapts to traffic conditions under resource and bandwidth constraints; and 3) a feasibility-driven task assignment strategy that maintains computational balance while meeting system constraints. The proposed method decouples YOLO-based object detection tasks into several parallel modules within each UAV processing block and implements a dynamic resource management mechanism to balance processing efficacy and reliability. An experimental evaluation conducted across 17 nodes managing 45 full layers (comprising 270 total layer assignments) demonstrates that CPFT-MOSA performs superiorly to traditional methods. Our results show quantitative improvements in three key metrics: efficient resource utilization with an average of 16.8 active nodes; a 21% improvement in energy distribution efficiency with a value of 0.063; and an 8.5% reduction in execution time to 1.6 units compared to conventional approaches. The framework’s ability to balance multiple competing objectives while maintaining fault tolerance makes it particularly suitable for real-world traffic management applications where reliability and performance are crucial for time-sensitive decision-making in dynamic urban environments.