I. Introduction
Last decade has witnessed an increasing number of vehicles that are anticipated to reach 2 billions by 2035 [1]. Many of which are equipped with state-of-the-art on-board computers and sensors to offer advanced features such as autopilot, self-parking, radar, automatic safety systems. Notably, studies have shown that 70% of individual vehicles almost spend 95% of time for parking in parking lots, home garages or street parking spaces [1] [2]. Those on-board computers with level-four autonomous driving support can cost tens of thousands of dollars but the expected utilization of vehicles is extremely low. For example, the average driving time per day in America was only 50.6 minutes according to the data survey of AAA Foundation for Traffic Safety in 2016 [3]. These statistics show that on-board computers are idle in most of the time, suggesting that significant computation resources can be exploited for provisioning additional services. The overlooked resources of PVs can become ideal candidates for transient resources leveraged by Mobile Edge Computing (MEC). By edge computing, the conventional computation and storage services offered by the remote cloud are now extended to the edge of the network. With the introduction of PVs, the strength of edge is now doubled, but they complicates the offloading problem through determining the proper network resources to execute a given task. Moreover, incoming tasks can be classified into delay-sensitive (e.g. augmented reality, video surveillance, mobile gaming, transportation, live streaming, robot collaboration, autopilot) and delay-insensitive with rigid computational requirements [4].
Edge computing architecture integrated with parked vehicles (pvs) enabled by kubernetes.