Title :
Agent-based routing algorithms on a LAN
Author :
Yang, Y. ; Zincir-Heywood, A.N. ; Heywood, M.I. ; Srinivas, S.
Author_Institution :
Fac. of Comput. Sci., Dalhousie Univ., Halifax, NS, Canada
Abstract :
The AntNet algorithm for adaptive routing is implemented on a LAN using the TCP/IP protocol. A study is made of the relative merits of different reinforcement parameters central to the stable operation of the algorithm. The case of a constant reinforcement leads to slow but dependable performance whereas adaptive reinforcement appears to be sensitive to the window over which statistics are estimated.
Keywords :
adaptive estimation; learning (artificial intelligence); local area networks; parameter estimation; routing protocols; software agents; statistical analysis; transport protocols; AntNet algorithm; LAN; TCP/IP; adaptive reinforcement; adaptive routing; agent-based routing algorithms; constant reinforcement; parameter estimation; performance; protocol; stable operation; statistics; Algorithm design and analysis; Chemicals; Computer networks; Computer science; Data structures; Insects; Local area networks; Routing; Testing; Traffic control;
Conference_Titel :
Electrical and Computer Engineering, 2002. IEEE CCECE 2002. Canadian Conference on
Print_ISBN :
0-7803-7514-9
DOI :
10.1109/CCECE.2002.1012965
