Traditional analog-to-digital convertors(ADCs) based on the Shannon sampling theorem can hardly satisfy the sampling requirements for high-frequency and wideband signals. For the purpose of sampling rate reducing, data transmission, storage and process relaxing, new analog-to-information convertors(AICs) based on sub-Nyquist sampling methods have drawn intensive attentions in past few years. Compressed sensing (CS) theory, single vector space (SVS) and union of subspaces(UoS) sampling theory are firstly introduced. Based on the UoS signal model, an emphasis on AIC sampling architectures and recovery algorithms are summarized and compared afterwards. Then, a multiple-antenna sub-Nyquist sampling architecture and its corresponding augmented recovery method based on the subspace decomposition are proposed. Finally, future research directions are given.