User's Guide for SAMUEL, Version 1.3.
Abstract
SAMUEL (Strategy Acquisition Method Using Empirical Learning) is a machine learning system designed to actively explore alternative behavior in a simulated environment, and to construct high performance rules from this experience. The learning method relies on the notion of competition and employs genetic algorithms to search the space of decision policies. The rule language in SAMUEL also makes it easier to incorporate existing knowledge, whether acquired from experts or by symbolic learning programs. The system includes a competition based production system interpreter, incremental strength updating procedures to measure the utility of rules, and genetic algorithms to modify strategies based on past performance. The current version includes a more convenient language for the expression of tactical control rules, better interfaces, and a number of new heuristics for rule modification. We have experimented with SAMUEL on a task involving learning control rules that enable a simulated robotic aircraft to evade an approaching missile. SAMUEL has been able to learn high performance strategies for this task. This manual should help the user to experiment with SAMUEL on other problems.
Document Details
- Document Type
- Technical Report
- Publication Date
- May 06, 1991
- Accession Number
- ADA235611
Entities
People
- Helen G. Cobb
- John J. Grefenstette
Organizations
- United States Naval Research Laboratory