Constrained Federated Reinforcement Learning for Secure Offloading in Vehicular Edge Computing

Abstract

Vehicular Edge Computing (VEC) reduces latency for computation-intensive vehicular applications by offloading tasks to roadside edge servers. Real vehicular workloads, however, consist of dependent subtasks with stringent reliability needs, while the cross-region collaborative learning used to adapt offloading policies must stay robust to heterogeneous data and secure over open wireless channels. We propose CRAFT, a Constrained Reinforcement-based Adaptive Federated Task Offloading framework that casts Directed Acyclic Graph (DAG)-structured offloading and resource allocation as a Constrained Markov Decision Process. CRAFT couples centralized-critic policy learning with Lagrangian relaxation to explicitly regulate the long-term task-completion constraint, and uses two-tier federated aggregation with proximal regularization to jointly mitigate non-Independent and Identically Distributed (non-IID) drift and stabilize constraint adaptation. Confidentiality and integrity for federated parameter exchange are provided by post-quantum homomorphic encryption and lattice-based signatures with modest amortized overhead. On a Virtual Machine-based testbed over a realistic Melbourne Central Business District network with real computation kernels, CRAFT cuts weighted cost by up to 42% over state-of-the-art baselines, maintains a completion rate of around 96%, and adds 6.4% security overhead.