MADRL-Based Multiobjective Joint Task Offloading and Resource Allocation in Heterogeneous IIoT

Abstract

In Industrial Internet of Things systems, heterogeneous devices generate tasks with diversified quality-of-service (QoS), including latency-sensitive tasks and energy-sensitive tasks. However, most existing task offloading approaches optimize a single performance metric or convert multi-objective problem into weighted single-objective problem. It is challenging to meet different QoS of the heterogeneous tasks. To address this challenge, a hierarchical solution framework is proposed. For the inner layer, the resource allocation subproblem is formulated as a separable convex program and solved via the Lagrange multiplier method and Karush-Kuhn-Tucker conditions. For the outer layer, an attention-masked multi-agent proximal policy optimization algorithm is proposed, in which a latency agent and an energy agent are deployed to make offloading decisions based on task types. Furthermore, an attention masking mechanism is incorporated to suppress inactive devices and mitigate the interference of zeropadded inputs. Simulation results demonstrate that the proposed algorithm achieves superior performance in terms of latency and energy consumption compared with benchmark approaches.