Multiagent Genetic Optimisation to Solve the Project Scheduling Problem under Uncertainty

Abstract—This paper considers a project scheduling problem under uncertainty, which belongs to a class of multiobjective problems of complex systems control whose decision search time grows exponentially depending on the problem dimension. In this paper, we propose a multiagent genetic optimisation method based on evolutionary and multiagent modelling by implementing different decision searching strategies, including a simulation module and numerical methods application. The comparative analysis of the scheduling methods has shown that the proposed method supports all features that might be useful in effective decision searching of the stochastic scheduling problem. The proposed multiagent genetic optimisation method, the MS Project resource reallocation method, and a heuristic simulation method were compared whilst addressing a real-world deterministic scheduling problem. The comparison has shown: firstly, the unsuitability of the MS Project planning method for solving the formulated problem; and secondly, both the advantage of the multiagent genetic optimisation method in terms of economic effect and disadvantage in terms of performance. Experimental results in conditions of uncertainty demonstrate the effectiveness of the proposed method. Some techniques to reduce the impact of the method’s disadvantage are proposed in the conclusion, as well as the aims of future work. Keywords-project scheduling; genetic algorithms; simulation; subcontract work optimisation; problem under uncertainty