TY - JOUR
T1 - Agent-based project scheduling
AU - Knotts, Gary
AU - Dror, Moshe
AU - Hartman, Bruce C.
N1 - Funding Information:
This work was financially supported by PRONEX–UFRGS, FAPERGS (00/0173.9, 01/ 1103.5, and BIC 00512023), and CNPq (4700181/ 01-5). The authors are grateful to CPGq-UFRGS and Brazilian Geological Survey–CPRM (SUREG PA) for logistical support.
PY - 2000/5
Y1 - 2000/5
N2 - Agent technology offers a new way of thinking about many of the classic problems in operations research. Among these are problems such as project scheduling subject to resource constraints. In this paper, we develop and experimentally evaluate eight agent-based algorithms for solving the multimode, resource-constrained project scheduling problem. Our algorithms differ in the priority rules used to control agent access to resources. We apply our approach to a 51-activity project originally published by Maroto and Tormos [1] We solve the problem using two types of agent-based systems: (i) a system of simple, reactive agents that we call basic agents; and (ii) a system of more complex, deliberative agents that we call enhanced agents. Of the eight priority rules tested, we find that priority based on shortest processing time performs best in terms of schedule quality when applied by basic agents while the priority based on earliest due date performs best when applied by enhanced agents. In comparing agents across priority rules, we find that enhanced agents generate much better schedules (with makespans up to 66% shorter in some cases) and require only slightly more computation time.
AB - Agent technology offers a new way of thinking about many of the classic problems in operations research. Among these are problems such as project scheduling subject to resource constraints. In this paper, we develop and experimentally evaluate eight agent-based algorithms for solving the multimode, resource-constrained project scheduling problem. Our algorithms differ in the priority rules used to control agent access to resources. We apply our approach to a 51-activity project originally published by Maroto and Tormos [1] We solve the problem using two types of agent-based systems: (i) a system of simple, reactive agents that we call basic agents; and (ii) a system of more complex, deliberative agents that we call enhanced agents. Of the eight priority rules tested, we find that priority based on shortest processing time performs best in terms of schedule quality when applied by basic agents while the priority based on earliest due date performs best when applied by enhanced agents. In comparing agents across priority rules, we find that enhanced agents generate much better schedules (with makespans up to 66% shorter in some cases) and require only slightly more computation time.
UR - http://www.scopus.com/inward/record.url?scp=80051523912&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=80051523912&partnerID=8YFLogxK
U2 - 10.1080/07408170008963915
DO - 10.1080/07408170008963915
M3 - Article
AN - SCOPUS:80051523912
SN - 0740-817X
VL - 32
SP - 387
EP - 401
JO - IIE Transactions (Institute of Industrial Engineers)
JF - IIE Transactions (Institute of Industrial Engineers)
IS - 5
ER -