This paper discusses a number of research areas concerned with dynamic traffic management, including the simulation, forecasting and automatic control of motorway traffic flows. The research is based on detailed non-linear Statistical Traffic Model (STM) simulations of motorways in The Netherlands. However, we have found that the behaviour of the traffic system can be approximated by simple, reduced order linearised time variable parameter transfer function models, that are identified and recursively estimated directly from simulated or measured data. Such linear models are employed in a Proportional-Integral-Plus, Linear Quadratic (PIP-LQ) control system for multivariable ramp metering (i.e. traffic lights on the on-ramp entrances). The paper also proposes a novel non-parametric approach to the estimation of the ubiquitous non-linear fundamental traffic diagram. This technique successfully handles the discontinuity between free flow and congested conditions and so provides an objective estimate of the required control set point.