Using co-evolution to produce robust robot control

Genetic programming is used to co-evolve robots and their test environments. Simulated mobile robots evolved based on their ability to navigate test courses. Test courses evolve based on their ability to cause robots to crash. Coupling robots to their test environments produces robust results. Co-ev...

Full description

Saved in:
Bibliographic Details
Main Author: McNutt, G.
Format: Conference Proceeding
Language:English
Subjects:
Computer crashes
Genetic programming
Mobile robots
Navigation
Robot control
Robot sensing systems
Robust control
Robustness
Sun
Vehicle crash testing
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Genetic programming is used to co-evolve robots and their test environments. Simulated mobile robots evolved based on their ability to navigate test courses. Test courses evolve based on their ability to cause robots to crash. Coupling robots to their test environments produces robust results. Co-evolution simultaneously evolves the robot controller, verifies the robot physical design and tests the robots in increasingly difficult environments. The robots produced can routinely navigate over 98% of these highly evolved, difficult test courses. In several cases, robots are found that navigate 100% of the courses.
ISSN:0191-2216
DOI:10.1109/CDC.1997.657684