Adaptive Task Offloading Framework for Serverless Edge Computing

Serverless Edge Computing (SEC) has emerged as a transformative paradigm to meet the computational demands of latency-sensitive applications in the Internet of Things (IoT). By processing tasks closer to data sources., SEC minimizes latency and improves system responsiveness. However., the heterogeneity of edge resources., dynamic workloads., and the need to prioritize high-urgency tasks introduce significant challenges in resource allocation and scheduling. This paper presents a hybrid task offloading framework tailored for SEC environments., integrating centralized task classification., task-specific server clustering., and multiagent Proximal Policy Optimization (PPO) based decision-making to manage resource allocation and task execution dynamically. By balancing deterministic strategies for task specialization and adaptive learning for real-time decision making., the proposed framework achieves significant reductions in latency while enhancing task completion rates. Experimental results validate the framework's superiority over traditional deep reinforcement learning (DRL) approaches., by achieving latency reduction of about 15% and improving average completion time by 29.46% demonstrating its effectiveness and scalability in complex., distributed computing environments.