A HYBRID PLATFORM FOR CONTEXT-AWARE V2X COMMUNICATIONS

Abstract

This report presents a new paradigm for Mobile Edge Learning (“MEL”) that enables the implementation of realistic distributed machine learning (DML) tasks on wireless edge nodes while taking into consideration heterogeneous computing and networking environments. A heterogeneity aware (HA) scheme was designed to solve the problem of dynamic task allocation for MEL in a way that maximizes the DML accuracy over wireless heterogeneous nodes or “learners” while respecting time constraints. This will enable context aware vehicle to everything (V2X) communication. The problem was first formulated as a quadratically constrained integer linear program (QCILP). Being non-deterministic polynomial-time (NP)-hard, it was relaxed into a non-convex problem over real variables that could be solved using commercially available numerical solvers. The relaxation also allowed us to propose a solution based on deriving the analytical upper bounds of the optimal solution using Lagrangian analysis and Karush-Kuhn-Tucker (KKT) conditions. The merits of the proposed analytical solution were demonstrated by comparing its performance to numerical approaches and comparing the validation accuracy of the proposed HA scheme to a baseline heterogeneity unaware (HU) equal task allocation approach. Simulation results showed that the HA schemes decreased convergence time up to 56 percent and increased the final validation accuracy up to 8 percent.