Advanced channel estimation in OTFS and NOMA using deep bayesian gaussian processes and compressive sensing

Area of Science:

• Wireless Communications
• Signal Processing
• Machine Learning

Background:

• Accurate channel estimation (CE) is vital for high-mobility wireless systems (OTFS, NOMA).
• Conventional pilot-based CE methods (LS, MMSE) struggle with accuracy in vehicular environments due to Doppler shifts and multipath propagation.
• Existing methods lack scalability and precise estimation.

Purpose of the Study:

• To develop an advanced channel estimation model for OTFS-NOMA systems operating in high-mobility vehicular environments.
• To enhance estimation accuracy and reduce pilot overhead compared to conventional methods.
• To leverage deep learning and probabilistic modeling for robust channel estimation.

Main Methods:

• Proposed a Deep Bayesian Gaussian Process-Compressive Sensing (DBGP-CS) model.
• Utilized Deep Neural Networks (DNNs) for non-linear delay-Doppler (DD) feature learning.
• Incorporated Gaussian Processes (GP) for uncertainty quantification and compressive sensing for channel sparsity exploitation.
• Simulated performance with 100 users at 120 km/h using the Extended Typical Urban (ETU) channel model.

Main Results:

• Achieved a 50% reduction in pilot overhead.
• Reported a Normalized Mean Squared Error (NMSE) of 0.01447 at 12 dB, a 90% improvement over MMSE-CE.
• Obtained a Bit Error Rate (BER) of 0.021159 at 12 dB.
• Demonstrated robust performance across varying mobility speeds (60-120 km/h) and multipath conditions (3-9 paths).

Conclusions:

• The DBGP-CS model offers a significant advancement in channel estimation for high-mobility OTFS-NOMA systems.
• The model effectively addresses challenges posed by Doppler shifts and multipath propagation.
• DBGP-CS provides superior accuracy and efficiency compared to traditional CE techniques.