Machine Learning based tracking at the trigger level

Triggers at high luminosity colliders play an important role in ensuring high sensitivity to new physics and signatures while keeping the data storage requirements at acceptable levels. This will be especially crucial in the next runs of the LHC and at the HL-LHC. Rather than simply imposing stricter pT or isolation requirements to keep trigger rates low, we explore the use of tracking requirements at trigger level. We present a NN based approach in identifying hits coming from a track. In this bottom-up, the complexity and input information are minimized to allow the NN to be implemented on a FPGA. This hardware based approach allows for increased data throughput, shorter latency and, crucially, flexibility to improve the algorithm in the future.