Spark-Based Parallel Genetic Algorithm for Simulating a Solution of Optimal Deployment of an Underwater Sensor...

Area of Science:

• Computer Science
• Engineering
• Distributed Systems

Background:

• Large-scale underwater sensor networks face challenges in node deployment due to energy constraints, delays, and disconnections.
• Genetic algorithms (GA) can optimize deployment but suffer from long computation times for large datasets.
• Existing parallel frameworks like Hadoop offer some improvement, but Spark provides greater parallel processing capabilities.

Purpose of the Study:

• To propose and evaluate a Spark-based parallel genetic algorithm (GA) for optimizing underwater sensor network (UWSN) node deployment.
• To address the limitations of traditional GA in terms of computation time and premature convergence in large-scale UWSN scenarios.
• To leverage Spark's parallel processing power for efficient GA operations like crossover and mutation.

Main Methods:

• Developed a Spark-based parallel genetic algorithm (GA) tailored for UWSN environments.
• Utilized the Shubert multi-peak function to calculate the extremum for optimal sensor deployment.
• Compared the performance of the Spark-based GA against single-node and Hadoop-based implementations for large-scale UWSN deployment.

Main Results:

• The Spark-based GA significantly reduced the running time for large-scale UWSN deployment compared to single-node and Hadoop frameworks.
• The proposed method effectively avoided premature convergence, a common issue in GA, due to enhanced randomness.
• Optimal deployment of underwater sensor nodes was achieved with improved efficiency.

Conclusions:

• A Spark-based parallel GA is a highly effective approach for optimizing large-scale underwater sensor network deployment.
• This method overcomes the computational bottlenecks and convergence issues associated with traditional GA in UWSNs.
• The Spark implementation offers a practical and efficient solution for real-world UWSN applications.