A Direct Position-Determination Approach for Multiple Sources Based on Neural Network Computation.

Area of Science:

• Signal Processing
• Wireless Communication
• Artificial Intelligence

Background:

• Traditional two-step localization methods rely on measuring positioning parameters.
• Direct Position Determination (DPD) offers superior localization but faces computational challenges with existing algorithms like Maximum Likelihood (ML) and Multiple Signal Classification (MUSIC).
• Real-time demands in localization necessitate computationally efficient algorithms.

Purpose of the Study:

• To propose a novel modular neural network for computationally efficient multiple-source Direct Position Determination (DPD).
• To evaluate the performance of the proposed neural network against conventional DPD algorithms.
• To demonstrate the suitability of the proposed method for real-time localization applications.

Main Methods:

• A modular neural network architecture was developed for DPD.
• The area of interest was divided into sub-areas, with Multilayer Perceptron (MLP) networks used for source detection and filtering.
• Radial Basis Function (RBF) networks were employed for precise position estimation within sub-areas.

Main Results:

• Simulation results confirmed the successful application of trained neural networks for DPD.
• The proposed MLP-MLP-RBF method demonstrated performance comparable to the MUSIC-based DPD algorithm across various signal conditions.
• The MLP-MLP-RBF network exhibited significantly lower computational intensity compared to classical DPD algorithms.

Conclusions:

• Modular neural networks provide a viable and efficient solution for multiple-source DPD.
• The proposed MLP-MLP-RBF method achieves high localization accuracy while being computationally less intensive.
• This approach is well-suited for real-time localization applications requiring fast and accurate transmitter positioning.