Real-time geometry-based wireless channel emulation

Connected autonomous vehicles and industry 4.0 pro- duction scenarios require ultrareliable low-latency communication links. The varying positions of transmitter, reflecting objects, and receiver cause a nonstationary time- and frequency-selective fad- ing process. In this paper, we present the necessary hardware architecture and signal processing algorithms for a real-time geometry-based channel emulator, that is needed for testing of wireless control systems. We partition the nonstationary fading process into a sequence of local stationarity regions and model the channel impulse response as sum of propagation paths with time-varying attenuation, delay, and Doppler shift. We implement a subspace projection of the propagation path parameters, to com- press the time-variant channel impulse response. This enables a low data-rate link from the host computer, which computes the geometry-based propagation paths, to the software defined radio unit, that implements the convolution on a field programmable gate array (FPGA). With our new architecture, the complexity of the FPGA implementation becomes independent of the number of propagation paths. Our channel emulator can be parametrized by all known channel models. Without loss of generality, we use a parameterization by a geometry-based stochastic channel model, due to its nonstationary nature. We provide channel impulse re- sponse measurements of the channel emulator, using the RUSK Lund channel sounder for a vehicular scenario with 617 propaga- tion paths. A comparison of the time-variant power delay profile and Doppler spectral density of simulated and emulated channel impulse response showed a close match with an error smaller than −35 dB. The results demonstrate that our channel emulator is able to accurately emulate nonstationary fading channels with contin- uously changing path delays and Doppler shifts.