Constrained Contextual Bandit Learning for Adaptive Radar Waveform Selection

A sequential decision process in which an adaptive radar system repeatedly interacts with a finite-state target channel is studied. The capable of passively sensing the spectrum at regular intervals, provides side information for waveform selection process. transmitter uses sequence observations as well feedback from collocated receiver to select waveforms accurately estimate parameters. It shown that problem can be effectively addressed using linear contextual bandit formulation manner both computationally feasible and sample efficient. Stochastic adversarial models are introduced, allowing achieve effective performance broad classes physical environments. Simulations radar-communication coexistence scenario, radar-jammer demonstrate proposed substantial improvement detection when Thompson sampling EXP3 algorithms used drive Further, it harmful impacts pulse-agile behavior on coherently processed data mitigated by adopting time-varying constraint radar’s catalog.