Home
RETINA is an R&D project, started circa 2013, aimed at developing and implementing a specialized processor allowing the reconstruction of events with hundreds of charged-particle tracks in pixel and silicon strip detectors at 40 MHz, thus suitable for processing HL-LHC events at the full crossing frequency.
We designed and tested a massively parallel pattern-recognition algorithm, the so called “artificial retina algorithm”, inspired by studies of the processing of visual images by the brain as it happens in nature. We find that high-quality tracking in large detectors is possible with sub-microseconds latencies when the algorithm is implemented in modern, high-speed, high-bandwidth FPGA devices, programmed with hardware-level languages to optimally exploit their features. This opens a possibility of making track reconstruction happen transparently as part of the detector readout ("embedded reconstruction").
The project received funding as a 3-year R&D project from INFN under CSN5, and become a part of the LHCb-RTA project in 2018 under WP 6.
The main applications that have been developed up to now are:
- Cluster finding in the VELO pixel detector for the LHCb Upgrade (currently in final implementation stage)
- Track reconstruction in VELO pixel detector for the LHCb Upgrade (currently in simulation stage)
- Downstream track reconstruction for the forward tracking system of LHCb Upgrade-2 (currently in simulation stage)
Alternative configurations have also been studied to various degrees of detail: "generic" silicon strip detectors, LHCb VELO+UT tracking, Muon detectors, etc..The system is flexible enough to be adaptable to the reconstruction of nearly any tracking system, and possibly also to some non-tracking detectors.
For more information look at the talks and publications listed under the links in the left column.