A traffic management system for controlling traffic flow in an area is provided. The system has sensors (10) positioned at respective junctions (2) in the area, forming a network of nodes (2). Software (25) processes the sensor signals to derive vehicle signatures indicative of a particular vehicle detected at the node at a particular time. This is used to track the progress of vehicles traveling across the network and derive vehicle statistics (60) as to traffic volumes (64), routes (66) and journey times (62) at various times. Artificial intelligence (AI) (80) trained on historical vehicle statistics arranged to predict traffic arriving at plural junctions at a future time based on receiving a current count of vehicles sensed at nodes in the network. Based on the prediction, a control plan for traffic lights (3) is determined to optimise traffic flow.