Technology Category
Planning, Routing, and Scheduling: Optimization of Spatio-temporal SystemsMultiagent Systems: Mechanism DesignMachine Learning: Efficient ML / Green AI
Carbon-aware decentralized dynamic task offloading in MIMO-MEC networks via multi-agent reinforcement learning
🤖 AI Summary
This work addresses the challenges of high carbon emissions and inefficient scheduling in large-scale IoT microservices over MIMO-MEC networks, caused by spatiotemporal mismatches between task arrivals and renewable energy availability, as well as multi-user interference. To tackle these issues, the paper proposes a decentralized dynamic task offloading framework based on Multi-Agent Proximal Policy Optimization (MAPPO), which operates solely on local observations. By integrating a carbon-aware reward mechanism with a Decentralized Execution with Parameter Sharing (DEPS) architecture, the framework jointly optimizes carbon emissions, buffer delay, and energy waste, enabling green-slot adaptive scheduling and decoupling system throughput from grid carbon footprint. Experimental results demonstrate that, under extreme load conditions, the proposed method achieves near-zero packet loss and the lowest carbon intensity compared to DDPG and Lyapunov-based baselines, while maintaining constant O(1) inference complexity, offering significant advantages for lightweight deployment.
Technology Category
Application Category
📝 Abstract
Massive internet of things microservices require integrating renewable energy harvesting into mobile edge computing (MEC) for sustainable eScience infrastructures. Spatiotemporal mismatches between stochastic task arrivals and intermittent green energy along with complex inter-user interference in multi-antenna (MIMO) uplinks complicate real-time resource management. Traditional centralized optimization and off-policy reinforcement learning struggle with scalability and signaling overhead in dense networks. This paper proposes CADDTO-PPO, a carbon-aware decentralized dynamic task offloading framework based on multi-agent proximal policy optimization. The multi-user MIMO-MEC system is modeled as a Decentralized Partially Observable Markov Decision Process (DEC-POMDP) to jointly minimize carbon emissions and buffer latency and energy wastage. A scalable architecture utilizes decentralized execution with parameter sharing (DEPS), which enables autonomous IoT agents to make fine-grained power control and offloading decisions based solely on local observations. Additionally, a carbon-first reward structure adaptively prioritizes green time slots for data transmission to decouple system throughput from grid-dependent carbon footprints. Finally, experimental results demonstrate CADDTO-PPO outperforms deep deterministic policy gradient (DDPG) and lyapunov-based baselines. The framework achieves the lowest carbon intensity and maintains near-zero packet overflow rates under extreme traffic loads. Architectural profiling validates the framework to demonstrate a constant $O(1)$ inference complexity and theoretical lightweight feasibility for future generation sustainable IoT deployments.
Problem
Research questions and friction points this paper is trying to address.
carbon-aware
task offloading
MIMO-MEC
renewable energy
multi-agent reinforcement learning
Innovation
Methods, ideas, or system contributions that make the work stand out.
carbon-aware
decentralized task offloading
multi-agent reinforcement learning
MIMO-MEC
DEC-POMDP
🔎 Similar Papers
No similar papers found.
