In this paper, we develop an improved optimization algorithm based on genetic algorithm (GA) approach for the bandwidth allocation of ATM networks. The ATM switches can be connected with multiples of DS3 trunks via digital cross connect systems (DCS). One of the advantages of DCS is its ability to reconfigure a customer network dynamically. We utilize this advantage in the design and dynamic reconfiguration of ATM networks. The problem is formulated as a network optimization problem where a congestion measure based on the average packet delay is minimized, subject to capacity constraints posed by the underlying facility trunks. We choose the traffic routing on the express pipes and the allocation of the bandwidth on these pipes as the variables in this problem. The previous GA algorithm is not practical because (1) the number of the traffic distribution patterns is huge, and (2) the values of offered traffic are continuous. A new representation of the chromosome, Net- Chro, and the reproduction operator are presented. We show that the previous algorithm cannot guarantee full usage of trunk capacities in the solutions it generates. We also discuss open-loop control to overcome the congestion caused by a trunk failure.