Edge Learning With Unmanned Ground Vehicle: Joint Path, Energy, and Sample Size Planning

Edge learning (EL), which uses edge computing as a platform to execute machine algorithms, is able fully exploit the massive sensing data generated by Internet of Things (IoT). However, due limited transmit power at IoT devices, collecting in EL systems challenging task. To address this challenge, article proposes integrate unmanned ground vehicle (UGV) with EL. With such scheme, UGV could improve communication quality approaching various devices. different devices may for jobs and fundamental question how jointly plan path, devices’ energy consumption, number samples jobs? This further graph-based path planning model, network consumption sample size model that characterizes F-measure function minority class size. these models, joint (JPESP) problem formulated large-scale mixed-integer nonlinear programming (MINLP) problem, nontrivial solve high-dimensional discontinuous variables related movement. end, it proved each device should be served only once along thus dimension significantly reduced. Furthermore, handle variables, tabu search (TS)-based algorithm derived, converges expectation optimal solution JPESP problem. Simulation results under task scenarios show our optimization schemes outperform fixed full schemes.