Multi-Domain Coincidence Processing and Memory Architecture for Real-Time Geiger Mode LiDAR
Name
McGuire-jtmcg-meng-eecs-2022-thesis.pdf
Description
Thesis PDF
Size
51.7 MB
Format
Adobe PDF
Checksum (MD5)
b4dda75207030e68f336fea446c9ace3
Author(s)
McGuire, Jacob T.
Advisor(s)
Rowe, Gregory
Steinmeyer, Joe
Vasile, Alexandru
Date Issued
May 2022
Publisher
Massachusetts Institute of Technology
Abstract
Geiger-Mode LiDAR is a powerful time-of-flight range sensing technology that enables rapid, wide area three-dimensional mapping with the unique capability of foliage penetration. These sensor arrays produce very high data rates on the order of 5 Gbps, requiring high-bandwidth motion compensation and coincidence processing to correlate the range returns and locate the modes in three-dimensional space. This paper proposes a multi-processor system architecture and memory management techniques for performing orientation-compensated histogram generation and peak detection to filter the LiDAR data stream, removing redundancy and spurious outputs. The multi-processor design, employing custom logic in concert with multiple CPUs, offers a reduction in system size, weight, and power [SWaP] by several orders of magnitude when compared to existing CPU-only real time coincidence processor designs. Behavioral simulations and hardware-in-the-loop testing offer partial proof of functionality for this design, which is capable of reducing the data rate by a factor of approximately 300 with output in the form of Cartesian coordinates, which can be directly integrated into a point cloud data structure for viewing. This promising result warrants further development work on LiDAR system designs incorporating these concepts.
MIT Department
Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science
Terms of Use
In Copyright - Educational Use Permitted
Copyright MIT
Persistent DSpace Link
