Spectrally Efficient Channel Estimation for High Mobility Communications

IMT-2030 (“6G") defines the next generation of digital communication systems with aims to operate in high-velocity environments such as high-speed trains and non-terrestrial networks using low-Earth orbit satellites. High mobile terminal speeds create difficulties for receivers with respect to high Doppler shifts and rapidly-changing channel distortion conditions. High Doppler shifts in multipath environments destroy subcarrier orthogonality in current LTE/5G communication systems that use Orthogonal Frequency Division Multiplexing (OFDM) modulation. Time-varying channels make channel distortion measurements stale and require more frequent channel estimates that lowers data throughput and spectral efficiency (SE). Our research focuses on the challenges of channel estimation in high mobility environments with solutions that minimize degradation in SE. 

We first solve the problem of channel estimation in time-varying channels. Rather than increasing the frequency of pilot symbol transmissions to refresh stale channel state information (CSI), we propose using machine learning (ML) with Gaussian Process Regression (GPR) to infer the channel distortion without direct measurement. Using ML can increase SE by spacing pilots farther apart in time to allow for more data throughput without sacrificing performance. We apply GPR to OFDM in high mobility scenarios, run system level simulations, and show that the performance of the learned channel exceeds traditional channel estimation methods. 

Next we mitigate interference from extreme Doppler shifts by introducing a new Orthogonal Time Frequency Space (OTFS) modulation operating in the delay-Doppler domain that is resilient to Doppler shift and characterizes time-varying channels in a quasi time-invariant space. We present an exemplary OTFS framework for aeronautical mobile telemetry (AMT) with parameters optimized for mobile velocities exceeding twice the speed of sound. Following system design and proof-of-concept, we focus on two distinct areas to improve OTFS performance for IMT-2030. First, we estimate the channel in the delay-time domain using GPR to decode in the time domain and avoid the problem of sub-optimal delay-Doppler domain decoding performance when in the presence of fractional Doppler. Better performance is seen over existing delay-Doppler domain decoding methods. Second, we solve a problem unique to AMT and Integrated Sensing and Communications (ISAC) where large path delay spreads exist due to reflections from distant geographic features. Large path delays can significantly worsen SE because traditional OTFS channel sounding requires data dropouts proportional to the length of the channel delay spread. We propose a new channel estimation technique using a low-power pilot signal superimposed over data that can measure large delay spread channels with no data dropouts, and show that spectral efficiency is better than traditional channel sounding measurements.