Resource Allocation in Construction Scheduling based on Multi- Agent Negotiation

Computerized project management in construction is traditionally based on precedence diagrams such as PERT or CPM. Resources required for the execution of activities are usually scarce and therefore provide additional constraints to the predecessor-successor relationships. The objective of this NPhard problem, which is also known as Resource Constrained Project Scheduling Problem (RCPSP), is to minimize the makespan considering all existing constraints. Since larger project plans in construction contain thousands of individual activities, an exact solution of the latter is hardly possible within a reasonable amount of time. For this reason various approaches such as branch-and-bound methods or heuristics like simulated annealing, particle swarm optimization and genetic algorithms were applied. The procedures are based on the determination of processes, where resources are available and predecessors completed. Hence resources are allocated immediately after the completion of prior tasks and will not be reserved for more critical activities. In contrast to prior works in Multi-Agent-Systems (MAS) the approach introduced in this paper uses agent-technology both for processes and resources. Autonomous process agents register their activities on a central blackboard, where all agents may negotiate on resource allocation depending on the available proposals offered by resource agents. The latter calculate their qualification for the corresponding activity using implemented utility functions. Hence resources are allocated to critical processes and identical resources are distributed to support ideal construction flow. This concept for agent based project scheduling is implemented in Discrete-Event-Simulation (DES) and evaluated by multiple experiments. Therefore the quality of results were validated using standardized project plans from the field of Operations Research (OR). Additionally the results are compared with the results from Monte-Carlo-Simulations integrated in the same system, so that quality and computing time could be evaluated. The overall concept was proven as stable and reliable for the scheduling of construction projects.