NeuroWav: Toward Real-Time Waveform Design for VANETs using Neural Networks

Vehicular Ad-Hoc networks depend on clear communication between vehicles using radio frequency in order to operate effectively. Interference from existing technologies using the RF spectrum, e.g. IoT devices, UAV, mobile systems, calls into question the feasibility of future VANET systems without an ability to cut through the noise. One approach to overcome interference is to use waveform design to provide this capability. Regrettably, most traditional algorithms are too computationally complex to perform efficiently in real-time. In this paper, we present early work on NeuroWav: a neural network based approach to waveform design to combat the effects of interference at low latency. NeuroWav is low size, weight, and power, executes 10X faster than the fastest extant waveform design algorithms, and provides performance results comparable with a high fidelity waveform design algorithm. Simulation results are provided that corroborate the theoretical expectations.