Though most existing signal systems include some optimization schemes, their performance suffers from inaccurate route travel prediction due to the limitations of data. In the Persistent Traffic Cookies (PTC) system being researched in UC Irvine, the path-based variables including path flow and path travel time can be obtained from the historical trip tables and current movement information stored in individual vehicles. The information is automatically updated by intersection wireless hardware every time the drivers return to those locations, and can be read by the same hardware, unless the drivers choose to withhold it. With such path variables data diary, the improved traffic control schemes can be introduced for a group of intersections called a sub-network. For network-level control, the path flows are estimated from the inference of individual vehicle movement to capture a movement along several intersections. The future path flow is predicted based on the current path flow and historical data. Based on it, we present a scheme to group a series of intersections as a sub-network for signal optimization. First, the interaction between any two intersections is estimated by the path flow between them. After choosing a Critical Intersection in a network, a group of intersections having a certain interaction with it is selected and formed a sub-network. The signal optimization in a sub-network is accomplished by a Mixed Integer Linear Problem with the objective to minimize the total delay and a set of constraints. To our knowledge, this is the first real-time traffic control optimization scheme developed using travel diaries.