[Defense] Resource-Efficient Hardware-Software Co-Design for Voxelized 3D Point-Cloud’s Indexing and Storage
Thursday, April 21, 2022
1:00 pm - 3:00 pm
will defend his dissertation
Resource-Efficient Hardware-Software Co-Design for Voxelized 3D Point-Cloud’s Indexing and Storage
3D maps, with many millions to billions of points, are now constructed and used in an increasing number of applications, with processing rates in the hundreds of thousands to millions of points per second. In mobile applications, power and energy consumption for managing such data and extracting useful information thereof, are critical concerns, especially to keep a system running for a long period of time before requiring battery recharging or swapping. A hardware-software computer co-design has the potential to better both energy and performance for this type of application through specialization. In this dissertation, I present the software and hardware components of such a co-design for indexing and storage of voxelized 3D point-cloud streams. Beginning from minimizing dynamic memory footprint as a first step towards achieving energy efficiency, I present and investigate novel voxel point-cloud representations. These are based on tree data structures with integer values as input coordinates, however, when used on byte-oriented computers, data access performance drawbacks may emerge. To negate them, I also present the hardware component of the co-design; a novel memory model and a novel instruction set architecture family customized for this memory model. The hardware implementation on FPGA exhibited significant power and energy savings, with the potential to improve the uptime of mobile survey systems.
1:00PM - 3:00PM CT
Location via MS Teams
Dr. Lennart Johnsson, dissertation advisor
Faculty, students and the general public are invited.