AI models and techniques for muons identification in the ATLAS Phase-II L0 Muon Barrel trigger.

In order to fully exploit the potential of High Luminosity LHC, the ATLAS Muon Barrel Spectrometer will undergo some structural modifications and the L0 trigger algorithm will run on FPGAs. In this contribution new AI models have been studied to identify the muon tracks, to distinguish the tracks from random noise hits and to reconstruct the physical quantities on which to perform the trigger selection. Different techniques have been adopted in order to satisfy the strict requirements given by ATLAS and compress the algorithm to successfully run on a FPGA with minimal loss in performance.