Artificial Intelligence / Machine Learning Based Communication and Spectrum Sensing Applications

HISAR Lab specializes in artificial intelligence (AI) and machine learning (ML) methods for spectrum sensing and modulation classification. Several machine learning and deep learning architectures using advanced object detection methods such as YOLO and Detectron have been developed to tackle various challenges in this area, such as analog and digital modulation classification, cellular signal classification for technologies such as GSM, UMTS, LTE and 5GNR. In addition, HISAR Laboratory also focuses on wireless interference detection and classification.

While our work is not limited to these applications, we envision that in the near future, AI/ML-based solutions will be key components of complex physical layer blocks. Instead of cascading new blocks on the transmitter and receiver sides, which increases computational and system complexity, AI/ML can replace or complement these blocks, and adding models that perform multiple operations with a single block by reducing system complexity through learned models will be a more efficient and effective solution for communication systems. Our research includes the development of various communication blocks such as symbol mapper, encoder, beamformer, channel estimator, demodulator, symbol decoder and many other transceiver components based on AI/ML to improve accuracy and reduce computational burden.