Neural Networks for Real-Time Terrain Typing.
Abstract
Many robotics tasks require an ability to determine quickly the nature of the terrain surrounding the robot. In cross country navigation in particular, the robot needs to know where the vegetation is and where the hard obstacles are. I have developed a general system which has successfully allowed real-time terrain typing in the NavLab II autonomous vehicle. This system and training paradigm are based on standard neural network technology and allow the robot to learn arbitrary non-linear mappings from color and texture space to terrain space.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 01, 1995
- Accession Number
- ADA293569
Entities
People
- Ian L. Davis
Organizations
- Carnegie Mellon University