Digital pre-distorsion
Implementation fo state-of-the-art digital pre-distorsion algorithms on the DRRA2 fabric for 5G and 6G applications
Motivation
Point clouds have become a major player in the world of robots and autonomous vehicles. The processing of point clouds has typically been mapped to GPUs for their efficient parallel processing of large amounts of data. However, the power inefficiency of GPUs limits the deployment of advanced algorithms in edge devices where low power consumption is a must. This project will explore and develop RTL components that can be integrated into the DRRA2 fabric to efficiently process point clouds.
Supervisor and Examiner
- Supervisor: Jordi Altayo jordiag@kth.se
- Examiner: Ahmed Hemani hemani@kth.se
Tasks
- Study and understand the basics of point cloud processing.
- Perform a state-of-the-art evaluation of current point cloud accelerators such as PointAcc from MIT.
- Study the DRRA2 architecture and its expandability possibilities using custom application-specific resources.
- Propose and develop compute and memory resources that can efficiently handle the basic requirements of point cloud algorithms.
Required Skills
- Solid understanding of digital design.
- Basics of computer architecture.
- Solid knowledge of HDL such as Verilog, SystemVerilog, or VHDL.
- Basics of logic synthesis flows such as DC or Genus.