Share Email Print

Proceedings Paper

AltiVec performance increases for autonomous robotics for the MARSSCAPE architecture program
Author(s): Benny M. Gothard
Format Member Price Non-Member Price
PDF $17.00 $21.00

Paper Abstract

One of the main tall poles that must be overcome to develop a fully autonomous vehicle is the inability of the computer to understand its surrounding environment to a level that is required for the intended task. The military mission scenario requires a robot to interact in a complex, unstructured, dynamic environment. Reference A High Fidelity Multi-Sensor Scene Understanding System for Autonomous Navigation The Mobile Autonomous Robot Software Self Composing Adaptive Programming Environment (MarsScape) perception research addresses three aspects of the problem; sensor system design, processing architectures, and algorithm enhancements. A prototype perception system has been demonstrated on robotic High Mobility Multi-purpose Wheeled Vehicle and All Terrain Vehicle testbeds. This paper addresses the tall pole of processing requirements and the performance improvements based on the selected MarsScape Processing Architecture. The processor chosen is the Motorola Altivec-G4 Power PC(PPC) (1998 Motorola, Inc.), a highly parallized commercial Single Instruction Multiple Data processor. Both derived perception benchmarks and actual perception subsystems code will be benchmarked and compared against previous Demo II-Semi-autonomous Surrogate Vehicle processing architectures along with desktop Personal Computers(PC). Performance gains are highlighted with progress to date, and lessons learned and future directions are described.

Paper Details

Date Published: 18 February 2002
PDF: 10 pages
Proc. SPIE 4573, Mobile Robots XVI, (18 February 2002); doi: 10.1117/12.457454
Show Author Affiliations
Benny M. Gothard, Science Applications International Corp. (United States)

Published in SPIE Proceedings Vol. 4573:
Mobile Robots XVI
Douglas W. Gage; Howie M. Choset, Editor(s)

© SPIE. Terms of Use
Back to Top
Sign in to read the full article
Create a free SPIE account to get access to
premium articles and original research
Forgot your username?