Dynamic Task Offloading in Distributed VEC Networks: An Exploration and Exploitation Assisted Contract-Theoretic Approach
Rujing Shen, Jinglin Shi, Wei Li, Yonghui Li
Abstract
In vehicular edge computing (VEC) networks, task offloading is a prospective approach to alleviate the computing burden of vehicles. Most works studied task offloading in centralized scenarios, where centralized infrastructures host the offloading and force vehicles into providing required private information. However, considering the high costs of infrastructure deployment, distributed scenarios without infrastructures are more practical. This paper focuses on a distributed VEC network, where a task vehicle (TaV) offloads tasks to surrounding vehicles (SuVs) with energy harvesting capabilities. In this context, SuVs are self-interested, and how to motivate self-interested vehicles to process offloaded tasks and provide the required information is challenging. To tackle these issues, an exploration and exploitation assisted contract-theoretic (EEACT) scheme is developed, where each SuV is recruited to participate in task offloading, and to accept the contract item designed for its type. The SuV type is related with the SuV's private information, and is partially monotonic. As a result, traditional simplification methods of incentive compatible (IC) and individual rational (IR) constraints in contract theory are not applicable. To address this problem, novel simplification methods are proposed and proved theoretically. Based on the contract design, the TaV determines the task offloading decisions by minimizing the processing cost of tasks, including the computation delay and the energy consumption. A low-complexity algorithm based on SuV types is proposed to obtain the approximately optimal offloading strategy. Numerical results demonstrate the superiority of the proposed scheme over the baseline schemes in terms of optimality.