A Support Vector Learning-Based Particle Filter Scheme for Target Localization in Communication-Constrained...

Area of Science:

• Underwater Acoustics
• Sensor Networks
• Signal Processing
• Machine Learning

Background:

• Target localization is crucial for underwater acoustic sensor networks (UASNs).
• Challenges include limited communication capacity and sensing noise in underwater environments.
• Existing methods struggle with accuracy and efficiency under these constraints.

Purpose of the Study:

• To develop an improved target localization algorithm for communication-constrained UASNs.
• To enhance localization accuracy and network communication efficiency.
• To address challenges posed by sensing noise and limited bandwidth.

Main Methods:

• Utilized fractional sensor nodes with a support vector learning-based particle filter algorithm.
• Implemented a node-selection strategy to choose short-distance sensor nodes per time frame.
• Proposed a least-square support vector regression (LSSVR)-based observation function to handle noisy data and improve observation accuracy.

Main Results:

• The proposed algorithm significantly improved target localization accuracy in tested noise scenarios.
• The node-selection strategy effectively selected subsets of sensor nodes, boosting communication efficiency.
• Enhanced particle effectiveness avoided the 'particle degeneracy' problem, leading to better localization.

Conclusions:

• The developed localization algorithm offers a robust solution for UASNs facing communication constraints and noise.
• The integration of LSSVR and a particle filter effectively mitigates sensing noise.
• The node-selection strategy optimizes network resource utilization and communication performance.