Modeling and Performance Analysis of Task Offloading Schemes for Intelligent Transportation Systems

Abstract

The rise of Internet of Things (IoT) devices in smart cities has necessitated real-time processing that is often beyond the capabilities of remote cloud infrastructures due to high latency. Vehicular Fog Computing (VFC), where Roadside Units (RSUs) offload tasks to computationally capable passing vehicles (R2V task offloading), emerges as a promising Mobile Edge Computing (MEC) solution. However, the high mobility of vehicles and the finite capacity of the RSU’s buffer introduce signif- icant challenges, mainly the occurance of Head-of-Line (HoL) blocking. This occurs when a task at the head of the queue cannot be released because no in-range vehicle has sufficient processing capacity to complete it before leaving the RSU’s range, leading to task accumulation and denial of admission.

Prior literature modeled this blocking using an abstract probabilistic parameter, the task release probability p. This thesis addresses this limitation by deriving a closed-form analytical expression for p, explicitly correlating it with concrete ve- hicular traffic dynamics (flow rate, speed, residence time) and on-board computing resource availability. We develop a comprehensive discrete-event simulation frame- work, using M/M/1 and M/M/1/K queuing models, to quantify the impact of the HoL blocking probability (1−p) on system performance metrics, including the prob- ability of tail blocking when the buffer is full and the probability of HoL blocking due to incapable vehicles.

Building on this foundation, we focus on time-sensitive applications by incorporating task deadlines. We introduce a detailed comparative analysis between the widely used First-In-First-Out (FIFO) policy and the Earliest-Deadline-First (EDF) policy. This study rigorously evaluates the performance gains of EDF in minimizing the probability of deadline mismatch, average queuing delay, and overall average number of tasks in the queue under varying traffic load and system capacity constraints.