Peer Review
The proliferation of data-intensive mobile applications has necessitated efficient computation offloading techniques to mitigate resource constraints in mobile devices (MDs). Existing approaches often fail to address multi-objective optimization challenges effectively. This paper proposes an
Data-intensive applications such as augmented reality and IoT services have transformed mobile devices (MDs) into critical computational platforms. However, resource constraints—including limited battery life and processing power—hinder their performance. Mobile Edge Computing (MEC) mitigates these issues by enabling task offloading to nearby edge servers, thereby enhancing computational efficiency and reducing latency.
Despite these advancements, MEC faces challenges in:
Optimizing energy, delay, and resource utilization simultaneously.
Scaling solutions across heterogeneous environments.
Adapting real-time decision-making under dynamic workloads.
This research introduces the Enhanced Adaptive Cat Hunt Optimization (EACHO) algorithm to address these challenges. The primary contributions of this study include:
Leveraging Directed Acyclic Graphs (DAGs) for effective task dependency modeling.
Developing adaptive optimization strategies for dynamic MEC environments.
Demonstrating superior performance through comparative analysis with existing methods.
Numerous studies have explored optimization methods for computation offloading. Key contributions include the following:
Zhang et al.
Wang et al.
Zhang et al.
Hu et al.
Despite significant advancements, existing studies often face challenges:
Scalability Issues: Algorithms may not adapt well to large-scale heterogeneous environments.
High Computational Overhead: Many models lack efficiency in dynamic MEC scenarios.
Single Objective Focus: Few approaches balance multiple objectives such as EC, delay, and resource utilization effectively.
The proposed EACHO algorithm addresses these gaps by integrating advanced multi-objective optimization strategies tailored for diverse MEC environments.
The system model for this research focuses on the computation offloading process in heterogeneous MEC environments. The model encompasses user devices, MEC servers, and cloud servers, with optimization objectives including energy efficiency, reduced latency, and balanced resource utilization.
The MEC framework comprises three core components (
MDs: These are resource-constrained devices generating computational tasks. The MDs decide whether to execute tasks locally or offload them to MEC servers based on task requirements, device capacity, and network conditions.
Edge Servers: Located closer to MDs, MEC servers handle the offloaded tasks to reduce latency. They also perform resource management and decision-making processes for task scheduling and execution.
Cloud Servers: Cloud servers support resource-intensive tasks that cannot be efficiently managed by MEC servers or MDs. These servers provide additional computational power but at the cost of higher latency and EC due to the distance from MDs.
The system is modeled to maximize computational efficiency while minimizing EC and execution delays. A DAG is employed to represent task dependencies and scheduling constraints, allowing a structured approach to manage resource allocation and execution sequences.
The task execution model uses DAGs to represent computational workflows. Each node in the DAG represents a task, while edges denote dependencies between tasks. The DAG model ensures that tasks are executed in a sequence that respects their dependencies, improving scheduling efficiency and reducing execution bottlenecks.
Task Definition: Each task is characterized by parameters such as data size, computational requirements, and priority levels. Tasks are categorized into independent, sequential, or parallel tasks.
Dependency Constraints: The execution of a task depends on the completion of its predecessor task. This ensures task dependencies are respected.
Optimization Goals: The DAG-based model optimizes for latency, EC, and resource utilization by minimizing total execution time.
The communication model evaluates the transmission of tasks between MDs, MEC servers, and cloud servers. The model accounts for factors such as bandwidth availability, channel interference, and signal-to-noise ratio.
Transmission Rate (γ): The transmission rate is calculated using the Shannon-Hartley theorem:
where is the bandwidth, the signal strength, and the noise level.
Communication Delay: It is the time taken to transmit data from an MD to an MEC server. The energy model computes the EC for local execution and task offloading.
EACHO algorithm is designed to address multi-objective optimization in MEC environments. The algorithm incorporates adaptive mechanisms for efficient task scheduling and resource allocation.
Initialization:
Define the search space, including task parameters, resource availability, and network conditions.
Initialize solution vectors representing task schedules.
Fitness Evaluation:
Compute fitness based on latency, EC, and resource utilization: where are weights assigned to objectives.
Search Mechanism:
Exploration Phase: Generate new solutions using random perturbations: where is a random factor.
Exploitation Phase: Refine solutions within promising regions using adaptive parameters.
Adaptive Parameter Adjustment:
Adjust exploration and exploitation parameters.
Convergence Criteria:
Terminate the algorithm when the maximum number of iterations is reached or when fitness scores stabilize.
The EACHO algorithm demonstrates robustness and adaptability, achieving significant performance improvements in MEC task offloading scenarios.
In a mobile computation offloading system, an MD can either process a task locally or offload it to the cloud. The decision to offload depends on factors such as EC, latency, and computation time.
The decision to offload task iii to the cloud is represented by the binary variable
where
EC for local and cloud computation is calculated as follows:
where
where
The computation time for a task is modeled for both local and cloud computation as:
where
where
The transmission delay for offloading the task to the cloud is given by:
where
The total latency
where
In heterogeneous environments, mobile computation offloading aims to optimize multiple objectives simultaneously, such as EC and latency. The goal is to find a Pareto-optimal solution that balances these objectives.
We define two objectives to be minimized:
where
where
The optimization problem is a multi-objective minimization problem:
where x = [
A solution x∗ is Pareto optimal if no other solution improves one objective without degrading another. Formally, a solution is Pareto optimal if:
The Cat Hunt Optimization (CHO) algorithm is used to find Pareto-optimal solutions by iteratively adjusting the offloading decisions x. The objective function for the optimization process is:
where ω1 and ω2 are the weights that balance the EC and latency objectives. The algorithm simulates the behavior of a cat hunting for the optimal solution.
The optimization process is subject to the following constraints:
where
where Cmax is the maximum computational capacity of either the MD or the cloud server.
Simulations were conducted using MATLAB R2022a on a system with the following specifications:
Processor: Intel Core i7, 2.6 GHz
Memory: 16 GB RAM
Software Environment: MATLAB R2022a
The MEC environment included:
Number of tasks: 1000 tasks distributed across MDs.
Bandwidth: 10 MHz for each communication link.
Transmission Power: 1.5 W per device.
Task Sizes: Ranged between 0.5 and 10 MB.
The simulation compared the proposed EACHO algorithm with benchmark techniques such as NSGAIII, DRLCO, and MOWOA.
The performance of EACHO was evaluated based on the following metrics:
EC: Total energy consumed during task execution, including local execution and offloading.
Task Completion Delay: The sum of local processing times and transmission delays for all tasks.
Resource Utilization: Efficiency of MEC resource allocation.
The following results were observed:
EC: EACHO achieved an EC reduction of 18% compared to NSGAIII and 12% compared to DRLCO.
Task Completion Delay: EACHO reduced delay by 15% compared to MOWOA.
Cost Efficiency: The cost metric showed improvements of 10% compared to benchmark algorithms.
| Algorithm | Energy Consumption (J) | Task Delay (s) | Cost Efficiency |
|---|---|---|---|
| EACHO | 0.251 × 10−3 | 0.0172 | 0.387 |
| NSGAIII | 0.305 × 10−3 | 0.0198 | 0.425 |
| DRLCO | 0.287 × 10−3 | 0.0185 | 0.412 |
| MOWOA | 0.275 × 10−3 | 0.0201 | 0.408 |
The proposed EACHO algorithm demonstrated superior performance across all metrics. The reduction in EC and task delay highlights its ability to optimize resource allocation effectively. Additionally, the algorithm’s adaptive parameter tuning enabled it to outperform static heuristic approaches under varying workloads.
The incorporation of DAG-based modeling further enhanced the scheduling efficiency by ensuring proper task dependencies. Experimental results validate that EACHO is robust, scalable, and well-suited for real-world MEC applications.
| Methods | Delay (s) | EC (10− |
Cost |
|---|---|---|---|
| DRLCO |
1.002 | 0.467 | 1.469 |
| DDQN |
2.325 | 0.652 | – |
| MOIA |
1.897 | 0.982 | – |
| MOWOA |
2.271 | 0.295 | – |
| MOWOA2 |
2.265 | 0.342 | – |
| DECO (Type 1) |
0.198 | 0.752 | – |
| DECO (Type 2) |
3.992 | 0.482 | – |
| NSGAIII-TOMEC |
0.0198 | 0.351 | 0.451 |
| EACHO (Proposed) | 0.0172 | 0.251 | 0.387 |
This performance makes EACHO highly effective for mobile computation offloading under diverse scenarios, surpassing existing approaches in terms of delay, EC, and cost.
This research introduces the EACHO algorithm, designed to address multi-objective optimization by simultaneously targeting key objectives such as EC, delay reduction, and cost optimization. These factors are critical in evaluating the Quality of Experience (QoE) for mobile users. Initially, the research formulates the offloading problem associated with DAGs to highlight the intricate relationships between offloading and resource allocation strategies in heterogeneous environments, with constraints on delay and the need to minimize EC. The proposed EACHO method incorporates a balance factor with adaptive values to improve the global search capabilities of partial populations, leading to enhanced performance. Comparative analyses with baseline schemes and simulation results demonstrate the EACHO algorithm’s effectiveness in significantly reducing task delays, optimizing resource utilization, and minimizing the EC of MDs. Future work will extend this research to explore more complex MEC systems, where offloaded tasks can be segmented into multiple partitions. Additionally, the algorithm will be applied to enhance decision-making processes, optimize offloading strategies, and adapt to dynamic network conditions, further improving system performance.
Additional material is published online only. To view please visit the journal online.
N/a
N/a
No industry funding
N/a
Jyothirmai Munjam, Kesavan Gopal and M. Sailaja – Conceptualization, Writing – original draft, review and editing
Jyothirmai Munjam
Unsolicited and externally peer-reviewed
N/a