Computation Offloading and Task Scheduling on Network Edge

The Fifth-Generation (5G) networks facilitate the evolution of communication systems and accelerate a revolution in the Information Technology (IT) field. In the 5G era, wireless networks are anticipated to provide connectivity for billions of Mobile User Devices (MUDs) around the world and to support a variety of innovative use cases, such as autonomous driving, ubiquitous Internet of Things (IoT), and Internet of Vehicles (IoV). The novel use cases, however, usually incorporate compute-intensive applications, which generate enormous computing service demands with diverse and stringent service requirements. In particular, autonomous driving calls for prompt data processing for the safety-related applications, IoT nodes deployed in remote areas need energy-efficient computing given limited on-board energy, and vehicles require low-latency computing for IoV applications in a highly dynamic network. To support the emerging computing service demands, Mobile Edge Computing (MEC), as a cutting-edge technology in 5G, utilizes computing resources on network edge to provide computing services for MUDs within a radio access network. The primary benefits of MEC can be elaborated from two perspectives. From the perspective of MUDs, MEC enables low-latency and energy-efficient computing by allowing MUDs to offload their computation tasks to proximal edge servers, which are installed in access points such as cellular base stations, Road-Side Units (RSUs), and Unmanned Aerial Vehicles (UAVs). On the other hand, from the perspective of network operators, MEC allows a large amount of computing data to be processed on network edge, thereby alleviating backhaul congestion. {MEC is a promising technology to support computing demands for the novel 5G applications within the RAN. The interesting issue is to maximize the computation capability of network edge to meet the diverse service requirements arising from the applications in dynamic network environments. However, the main technical challenges are: 1) how an edge server schedules its limited computing resources to optimize the Quality-of-Experience (QoE) in autonomous driving; 2) how the computation loads are balanced between the edge server and IoT nodes in computation loads to enable energy-efficient computing service provisioning; and 3) how multiple edge servers coordinate their computing resources to enable seamless and reliable computing services for high-mobility vehicles in IoV. In this thesis, we develop efficient computing resource management strategies for MEC, including computation offloading and task scheduling, to address the above three technical challenges. First, we study computation task scheduling to support real-time applications, such as localization and obstacle avoidance, for autonomous driving. In our considered scenario, autonomous vehicles periodically sense the environment, offload sensor data to an edge server for processing, and receive computing results from the edge server. Due to mobility and computing latency, a vehicle travels a certain distance between the instant of offloading its sensor data and the instant of receiving the computing result. Our objective is to design a scheduling scheme for the edge server to minimize the above traveled distance of vehicles. The idea is to determine the processing order according to the individual vehicle mobility and computation capability of the edge server. We formulate a Restless Multi-Armed Bandit (RMAB) problem, design a Whittle index-based stochastic scheduling scheme, and determine the index using a Deep Reinforcement Learning (DRL) method. The proposed scheduling scheme can avoid the time-consuming policy exploration common in DRL scheduling approaches and makes effectual decisions with low complexity. Extensive simulation results demonstrate that, with the proposed index-based scheme, the edge server can deliver computing results to the vehicles promptly while adapting to time-variant vehicle mobility. Second, we study energy-efficient computation offloading and task scheduling for an edge server while provisioning computing services {for IoT nodes in remote areas}. In the considered scenario, a UAV is equipped with computing resources and plays the role of an aerial edge server to collect and process the computation tasks offloaded by ground MUDs. Given the service requirements of MUDs, we aim to maximize UAV energy efficiency by jointly optimizing the UAV trajectory, the user transmit power, and computation task scheduling. The resulting optimization problem corresponds to nonconvex fractional programming, and the Dinkelbach algorithm and the Successive Convex Approximation (SCA) technique are adopted to solve it. Furthermore, we decompose the problem into multiple subproblems for distributed and parallel problem solving. To cope with the case when the knowledge of user mobility is limited, we apply a spatial distribution estimation technique to predict the location of ground users so that the proposed approach can still be valid. Simulation results demonstrate the effectiveness of the proposed approach to maximize the energy efficiency of the UAV. Third, we study collaboration among multiple edge servers in computation offloading and task scheduling to support computing services {in IoV}. In the considered scenario, vehicles traverse the coverage of edge servers and offload their tasks to their proximal edge servers. We develop a collaborative edge computing framework to reduce computing service latency and alleviate computing service interruption due to the high mobility of vehicles: 1) a Task Partition and Scheduling Algorithm (TPSA) is proposed to schedule the execution order of the tasks offloaded to the edge servers given a computation offloading strategy; and 2) an artificial intelligence-based collaborative computing approach is developed to determine the task offloading, computing, and result delivery policy for vehicles. Specifically, the offloading and computing problem is formulated as a Markov decision process. A DRL technique, i.e., deep deterministic policy gradient, is adopted to find the optimal solution in a complex urban transportation network. With the developed framework, the service cost, which includes computing service latency and service failure penalty, can be minimized via the optimal computation task scheduling and edge server selection. Simulation results show that the proposed AI-based collaborative computing approach can adapt to a highly dynamic environment with outstanding performance. In summary, we investigate computing resource management to optimize QoE of MUDs in the coverage of an edge server, to improve energy efficiency for an aerial edge server while provisioning computing services, and to coordinate computing resources among edge servers for supporting MUDs with high mobility. The proposed approaches and theoretical results contribute to computing resource management for MEC in 5G and beyond