Machine learning-based beamforming for wireless communication

This technology is a deep learning platform that uses curriculum learning-based strategies to train neural networks for real-time optimization of beamforming, thus improving communication in 5G and wireless networks.

Unmet Need: Real-time, low-latency beamforming for 5G systems

Current beamforming optimization methods rely on iterative algorithms that are computationally expensive and introduce significant latency. These approaches are too slow for real-time use in modern wireless systems such as 5G, which require rapid, adaptive responses to changing network conditions. The complexity of these algorithms limits their scalability and makes them impractical for widespread use. Addressing these shortcomings is critical to enabling fast, efficient, wireless communication in next-generation networks.

The Technology: Real-time beamforming optimization for faster wireless systems

This technology uses artificial intelligence (AI) and machine learning approaches to train a neural network to compute optimal beamforming solutions in real time. This improves wireless signal direction, replacing slow, iterative, and complex methods with improved performance. With the AI model, this technology adapts to changing environments and outperforms current heuristic methods in multiple-input-single-output (MISO) and multiple-input-multiple-output (MIMO) systems, making it ideal for fast, reliable 5G and future communication networks.

Applications:

• Wireless networks
• Manufacturing and industrial automation
• Autonomous vehicles
• Military and defense
• Airplane connectivity and communication
• Cybersecurity
• Smart electric energy
• Sustainable computing
• Information technology systems

Advantages:

• Less computationally intensive than current methods
• Scalable and efficient for deployment
• Reduces latency and processes data in real-time
• Outperforms current heuristic methods in MISO and MIMO systems
• Lower power consumption because of reduced iterative solvers
• More robust to errors and noisy data

Lead Inventor:

Xiadong Wang, Ph.D.

Patent Information:

Patent Pending (US20250165780)

Related Publications:

• Johnston J, Wang X, “RNN Beamforming Optimizer for Rate-Splitting Multiple Access and Cell-Free Massive MIMO.” IEEE Transactions on Communications. 2024 January

• Johnston J, Liu X, Wu S, Wang X, “A Curriculum Learning Approach to Optimization with Application to Downlink Beamforming.” IEEE Transactions on Signal Processing. 2023 December

Tech Ventures Reference:

• IR CU24175

• Licensing Contact: Greg Maskel