The paper addresses the inventory control problem in logistic networks with complex, mesh-type topologies. The goods are shipped with non-negligible lead-time delay and an uncertain, arbitrary demand may be imposed on any node in the system. Excess demand is lost. Single-item periodic-review distribution process is governed by the (r, Q) policy. In order to adjust the policy parameters, a continuous genetic algorithm is used. In the optimization procedures, three objectives – holding and transportation costs reduction and customer satisfaction maximization – are considered. The paper shows how one can effectively find the reorder point and order quantity for each node when the policy is implemented in a distributed mode, as desired in complex systems. Two approaches to the crossover operation have been compared. The separate operator allows one to obtain more suitable solutions at the expense of more significant computational effort.