Computation task offloading involves transferring computational tasks, particularly those requiring significant processing power, from resource-constrained devices to more powerful and resource-rich remote servers. As applications evolve such as 3D video rendering, conventional offloading methods struggle with increased computational and communication demands. To address the challenges posed by the increasing volume of data

overwhelming available resources, this thesis introduces novel models, algorithms, and optimisation strategies aimed at reducing data redundancy to enhance the efficiency of computation offloading.

The thesis contributes to several key areas. First, a dynamic multi-task offloading model is proposed for low-performance mobile devices, where processing power and resources are limited. This model efficiently exploits time-domain and task-domain correlations to minimize redundant data transmissions and reduce energy consumption. The proposed algorithms dynamically adapt resource allocation based on real-time network and device conditions, ensuring that only the most critical data is offloaded for processing. In the realm of multi-user edge computing environments, this thesis investigates the problem of maximizing the swarm lifetime through exploiting correlations between robots’ data. A graph-based model and several subset selection methods are introduced to select optimal robot subsets for task offloading, balancing the energy consumption among different robots and improving energy efficiency. Theoretical analyses and simulations show that these methods achieve substantial gains in energy efficiency and communication performance. Finally, a novel progressive hybrid automatic-repeat-request (PH-ARQ) protocol is introduced. Leveraging the information bottleneck principle, this protocol prioritizes the data most relevant to the task, reducing the overall source coding rate and improving both energy efficiency and communication reliability. All these contributions provide new insights and approaches to addressing the complexities of computation task offloading in increasingly data-intensive environments. By refining models and optimizing resource allocation, it contributes to more effective and sustainable offloading strategies for modern wireless systems, potentially revolutionizing future wireless communication.