In this paper, we present a continuous-time optimal control framework for generating reference trajectories for autonomous vehicles. The method developed involves chance constraints for modelling uncertainty. A previous work presented such a model in discrete time and designed for urban driving scenarios only. We extend those results in continuous time. It generates reference trajectories on urban driving scenarios with faster computation and better capacity to capture uncertainty. Our model is less likely to violate the problem's constraints in risky scenarios, and is also robust for optimal control on long term horizons in national roads and highways.