Activity Detection in Grant-Free Systems

This thesis investigates various methods that aim to solve the activity detection problem in grant free wireless access. First, a grant-free wireless system model is established considering Gaussian and iterative collision-based packing (ICBP) codebooks. The methods include two exact algorithms, multiple signal classification (MUSIC) and orthogonal matching pursuit (OMP), which are compared in performance and the influence of the wireless system parameters are further analyzed. In addition, two machine learning models are developed that aim tosolve the activity detection problem as a multi-label classification task.