Ubiquitous Sensor Network (USN) computing is a useful technology forautonomic integrating in different environments which can be available anywhere.Managing USN plays an important role on the availability of nodes and paths. Inorder to manage nodes there is a cyclic route starts from manager, passing nodes,and come back to manager as feedback. In this paper, a new, self-optimizing methodpresented for finding this cyclic path by combining epsilon greedy and geneticalgorithm and then it is compared with other well-known methods in terms of cost ofthe route they find and the power consumption. The results show that the route thatis found by our new method costs at least 53% less than other methods. However insome cases, it uses 32% more energy for finding the route which can be compensatein traversing the shorter route. The overall simulation results in prototype data showthe effectiveness of the proposed method.